Showing 1–50 of 85 results for author: Malo, L

Scoreformer: A Surrogate Model For Large-Scale Prediction of Docking Scores

Authors: Álvaro Ciudad, Adrián Morales-Pastor, Laura Malo, Isaac Filella-Mercè, Victor Guallar, Alexis Molina

Abstract: In this study, we present ScoreFormer, a novel graph transformer model designed to accurately predict molecular docking scores, thereby optimizing high-throughput virtual screening (HTVS) in drug discovery. The architecture integrates Principal Neighborhood Aggregation (PNA) and Learnable Random Walk Positional Encodings (LRWPE), enhancing the model's ability to understand complex molecular struct… ▽ More In this study, we present ScoreFormer, a novel graph transformer model designed to accurately predict molecular docking scores, thereby optimizing high-throughput virtual screening (HTVS) in drug discovery. The architecture integrates Principal Neighborhood Aggregation (PNA) and Learnable Random Walk Positional Encodings (LRWPE), enhancing the model's ability to understand complex molecular structures and their relationship with their respective docking scores. This approach significantly surpasses traditional HTVS methods and recent Graph Neural Network (GNN) models in both recovery and efficiency due to a wider coverage of the chemical space and enhanced performance. Our results demonstrate that ScoreFormer achieves competitive performance in docking score prediction and offers a substantial 1.65-fold reduction in inference time compared to existing models. We evaluated ScoreFormer across multiple datasets under various conditions, confirming its robustness and reliability in identifying potential drug candidates rapidly. △ Less

Submitted 25 June, 2024; v1 submitted 13 June, 2024; originally announced June 2024.

Comments: Accepted at the 1st Machine Learning for Life and Material Sciences Workshop at ICML 2024

An Unsupervised Approach to Achieve Supervised-Level Explainability in Healthcare Records

Authors: Joakim Edin, Maria Maistro, Lars Maaløe, Lasse Borgholt, Jakob D. Havtorn, Tuukka Ruotsalo

Abstract: Electronic healthcare records are vital for patient safety as they document conditions, plans, and procedures in both free text and medical codes. Language models have significantly enhanced the processing of such records, streamlining workflows and reducing manual data entry, thereby saving healthcare providers significant resources. However, the black-box nature of these models often leaves heal… ▽ More Electronic healthcare records are vital for patient safety as they document conditions, plans, and procedures in both free text and medical codes. Language models have significantly enhanced the processing of such records, streamlining workflows and reducing manual data entry, thereby saving healthcare providers significant resources. However, the black-box nature of these models often leaves healthcare professionals hesitant to trust them. State-of-the-art explainability methods increase model transparency but rely on human-annotated evidence spans, which are costly. In this study, we propose an approach to produce plausible and faithful explanations without needing such annotations. We demonstrate on the automated medical coding task that adversarial robustness training improves explanation plausibility and introduce AttInGrad, a new explanation method superior to previous ones. By combining both contributions in a fully unsupervised setup, we produce explanations of comparable quality, or better, to that of a supervised approach. We release our code and model weights. △ Less

Submitted 13 June, 2024; originally announced June 2024.

NIRPS first light and early science: breaking the 1 m/s RV precision barrier at infrared wavelengths

Authors: Étienne Artigau, François Bouchy, René Doyon, Frédérique Baron, Lison Malo, François Wildi, Franceso Pepe, Neil J. Cook, Simon Thibault, Vladimir Reshetov, Xavier Dumusque, Christophe Lovis, Danuta Sosnowska, Bruno L. Canto Martins, Jose Renan De Medeiros, Xavier Delfosse, Nuno Santos, Rafael Rebolo, Manuel Abreu, Guillaume Allain, Romain Allart, Hugues Auger, Susana Barros, Luc Bazinet, Nicolas Blind , et al. (89 additional authors not shown)

Abstract: The Near-InfraRed Planet Searcher or NIRPS is a precision radial velocity spectrograph developed through collaborative efforts among laboratories in Switzerland, Canada, Brazil, France, Portugal and Spain. NIRPS extends to the 0.98-1.8 $μ$m domain of the pioneering HARPS instrument at the La Silla 3.6-m telescope in Chile and it has achieved unparalleled precision, measuring stellar radial velocit… ▽ More The Near-InfraRed Planet Searcher or NIRPS is a precision radial velocity spectrograph developed through collaborative efforts among laboratories in Switzerland, Canada, Brazil, France, Portugal and Spain. NIRPS extends to the 0.98-1.8 $μ$m domain of the pioneering HARPS instrument at the La Silla 3.6-m telescope in Chile and it has achieved unparalleled precision, measuring stellar radial velocities in the infrared with accuracy better than 1 m/s. NIRPS can be used either stand-alone or simultaneously with HARPS. Commissioned in late 2022 and early 2023, NIRPS embarked on a 5-year Guaranteed Time Observation (GTO) program in April 2023, spanning 720 observing nights. This program focuses on planetary systems around M dwarfs, encompassing both the immediate solar vicinity and transit follow-ups, alongside transit and emission spectroscopy observations. We highlight NIRPS's current performances and the insights gained during its deployment at the telescope. The lessons learned and successes achieved contribute to the ongoing advancement of precision radial velocity measurements and high spectral fidelity, further solidifying NIRPS' role in the forefront of the field of exoplanets. △ Less

Submitted 13 June, 2024; v1 submitted 12 June, 2024; originally announced June 2024.

Comments: Proceeding at the SPIE Astronomical Telescopes + Instrumentation conference [Yokohama,Japan; June 2024]

New Mass and Radius Constraints on the LHS 1140 Planets -- LHS 1140 b is Either a Temperate Mini-Neptune or a Water World

Authors: Charles Cadieux, Mykhaylo Plotnykov, René Doyon, Diana Valencia, Farbod Jahandar, Lisa Dang, Martin Turbet, Thomas J. Fauchez, Ryan Cloutier, Collin Cherubim, Étienne Artigau, Neil J. Cook, Billy Edwards, Tim Hallatt, Benjamin Charnay, François Bouchy, Romain Allart, Lucile Mignon, Frédérique Baron, Susana C. C. Barros, Björn Benneke, B. L. Canto Martins, Nicolas B. Cowan, J. R. De Medeiros, Xavier Delfosse , et al. (21 additional authors not shown)

Abstract: The two-planet transiting system LHS 1140 has been extensively observed since its discovery in 2017, notably with $Spitzer$, HST, TESS, and ESPRESSO, placing strong constraints on the parameters of the M4.5 host star and its small temperate exoplanets, LHS 1140 b and c. Here, we reanalyse the ESPRESSO observations of LHS 1140 with the novel line-by-line framework designed to fully exploit the radi… ▽ More The two-planet transiting system LHS 1140 has been extensively observed since its discovery in 2017, notably with $Spitzer$, HST, TESS, and ESPRESSO, placing strong constraints on the parameters of the M4.5 host star and its small temperate exoplanets, LHS 1140 b and c. Here, we reanalyse the ESPRESSO observations of LHS 1140 with the novel line-by-line framework designed to fully exploit the radial velocity content of a stellar spectrum while being resilient to outlier measurements. The improved radial velocities, combined with updated stellar parameters, consolidate our knowledge on the mass of LHS 1140 b (5.60$\pm$0.19 M$_{\oplus}$) and LHS 1140 c (1.91$\pm$0.06 M$_{\oplus}$) with unprecedented precision of 3%. Transits from $Spitzer$, HST, and TESS are jointly analysed for the first time, allowing us to refine the planetary radii of b (1.730$\pm$0.025 R$_{\oplus}$) and c (1.272$\pm$0.026 R$_{\oplus}$). Stellar abundance measurements of refractory elements (Fe, Mg and Si) obtained with NIRPS are used to constrain the internal structure of LHS 1140 b. This planet is unlikely to be a rocky super-Earth as previously reported, but rather a mini-Neptune with a $\sim$0.1% H/He envelope by mass or a water world with a water-mass fraction between 9 and 19% depending on the atmospheric composition and relative abundance of Fe and Mg. While the mini-Neptune case would not be habitable, a water-abundant LHS 1140 b potentially has habitable surface conditions according to 3D global climate models, suggesting liquid water at the substellar point for atmospheres with relatively low CO$_2$ concentration, from Earth-like to a few bars. △ Less

Submitted 18 December, 2023; v1 submitted 23 October, 2023; originally announced October 2023.

Comments: 31 pages, 18 figures, accepted for publication in ApJL

Comprehensive High-resolution Chemical Spectroscopy of Barnard's Star with SPIRou

Authors: Farbod Jahandar, René Doyon, Étienne Artigau, Neil J. Cook, Charles Cadieux, David Lafrenière, Thierry Forveille, Jean-François Donati, Pascal Fouqué, Andrés Carmona, Ryan Cloutier, Paul Cristofari, Eric Gaidos, João Gomes da Silva, Lison Malo, Eder Martioli, J. -D. do Nascimento Jr., Stefan Pelletier, Thomas Vandal, Kim Venn

Abstract: Determination of fundamental parameters of stars impacts all fields of astrophysics, from galaxy evolution to constraining the internal structure of exoplanets. This paper presents a detailed spectroscopic analysis of Barnard's star that compares an exceptionally high-quality (an average signal-to-noise ratio of $\sim$1000 in the entire domain), high-resolution NIR spectrum taken with CFHT/SPIRou… ▽ More Determination of fundamental parameters of stars impacts all fields of astrophysics, from galaxy evolution to constraining the internal structure of exoplanets. This paper presents a detailed spectroscopic analysis of Barnard's star that compares an exceptionally high-quality (an average signal-to-noise ratio of $\sim$1000 in the entire domain), high-resolution NIR spectrum taken with CFHT/SPIRou to PHOENIX-ACES stellar atmosphere models. The observed spectrum shows thousands of lines not identified in the models with a similar large number of lines present in the model but not in the observed data. We also identify several other caveats such as continuum mismatch, unresolved contamination and spectral lines significantly shifted from their expected wavelengths, all of these can be a source of bias for abundance determination. Out of $>10^4$ observed lines in the NIR that could be used for chemical spectroscopy, we identify a short list of a few hundred lines that are reliable. We present a novel method for determining the effective temperature and overall metallicity of slowly-rotating M dwarfs that uses several groups of lines as opposed to bulk spectral fitting methods. With this method, we infer $T_{eff}$ = 3231 $\pm$ 21 K for Barnard's star, consistent with the value of 3238 $\pm$ 11 K inferred from the interferometric method. We also provide abundance measurements of 15 different elements for Barnard's star, including the abundances of four elements (K, O, Y, Th) never reported before for this star. This work emphasizes the need to improve current atmosphere models to fully exploit the NIR domain for chemical spectroscopy analysis. △ Less

Submitted 1 April, 2024; v1 submitted 18 October, 2023; originally announced October 2023.

Comments: 29 pages, 23 figures, accepted for publication in ApJ

Optimizing Drug Design by Merging Generative AI With Active Learning Frameworks

Authors: Isaac Filella-Merce, Alexis Molina, Marek Orzechowski, Lucía Díaz, Yang Ming Zhu, Julia Vilalta Mor, Laura Malo, Ajay S Yekkirala, Soumya Ray, Victor Guallar

Abstract: Traditional drug discovery programs are being transformed by the advent of machine learning methods. Among these, Generative AI methods (GM) have gained attention due to their ability to design new molecules and enhance specific properties of existing ones. However, current GM methods have limitations, such as low affinity towards the target, unknown ADME/PK properties, or the lack of synthetic tr… ▽ More Traditional drug discovery programs are being transformed by the advent of machine learning methods. Among these, Generative AI methods (GM) have gained attention due to their ability to design new molecules and enhance specific properties of existing ones. However, current GM methods have limitations, such as low affinity towards the target, unknown ADME/PK properties, or the lack of synthetic tractability. To improve the applicability domain of GM methods, we have developed a workflow based on a variational autoencoder coupled with active learning steps. The designed GM workflow iteratively learns from molecular metrics, including drug likeliness, synthesizability, similarity, and docking scores. In addition, we also included a hierarchical set of criteria based on advanced molecular modeling simulations during a final selection step. We tested our GM workflow on two model systems, CDK2 and KRAS. In both cases, our model generated chemically viable molecules with a high predicted affinity toward the targets. Particularly, the proportion of high-affinity molecules inferred by our GM workflow was significantly greater than that in the training data. Notably, we also uncovered novel scaffolds significantly dissimilar to those known for each target. These results highlight the potential of our GM workflow to explore novel chemical space for specific targets, thereby opening up new possibilities for drug discovery endeavors. △ Less

Submitted 4 May, 2023; originally announced May 2023.

Automated Medical Coding on MIMIC-III and MIMIC-IV: A Critical Review and Replicability Study

Authors: Joakim Edin, Alexander Junge, Jakob D. Havtorn, Lasse Borgholt, Maria Maistro, Tuukka Ruotsalo, Lars Maaløe

Abstract: Medical coding is the task of assigning medical codes to clinical free-text documentation. Healthcare professionals manually assign such codes to track patient diagnoses and treatments. Automated medical coding can considerably alleviate this administrative burden. In this paper, we reproduce, compare, and analyze state-of-the-art automated medical coding machine learning models. We show that seve… ▽ More Medical coding is the task of assigning medical codes to clinical free-text documentation. Healthcare professionals manually assign such codes to track patient diagnoses and treatments. Automated medical coding can considerably alleviate this administrative burden. In this paper, we reproduce, compare, and analyze state-of-the-art automated medical coding machine learning models. We show that several models underperform due to weak configurations, poorly sampled train-test splits, and insufficient evaluation. In previous work, the macro F1 score has been calculated sub-optimally, and our correction doubles it. We contribute a revised model comparison using stratified sampling and identical experimental setups, including hyperparameters and decision boundary tuning. We analyze prediction errors to validate and falsify assumptions of previous works. The analysis confirms that all models struggle with rare codes, while long documents only have a negligible impact. Finally, we present the first comprehensive results on the newly released MIMIC-IV dataset using the reproduced models. We release our code, model parameters, and new MIMIC-III and MIMIC-IV training and evaluation pipelines to accommodate fair future comparisons. △ Less

Submitted 21 April, 2023; originally announced April 2023.

Comments: 11 pages, 6 figures, to be published in Proceedings of the 46th International ACM SIGIR Conference on Research and Development in Information Retrieval (SIGIR '23), July 23--27, 2023, Taipei, Taiwan

ACM Class: H.3.0

The SPIRou Legacy Survey Rotation period of quiet M dwarfs from circular polarization in near-infrared spectral lines: I. The SPIRou APERO analysis

Authors: P. Fouqué, E. Martioli, J. -F. Donati, L. T. Lehmann, B. Zaire, S. Bellotti, E. Gaidos, J. Morin, C. Moutou, P. Petit, S. H. P. Alencar, L. Arnold, É. Artigau, T. -Q. Cang, A. Carmona, N. J. Cook, P. Cortés-Zuleta, P. I. Cristofari, X. Delfosse, R. Doyon, G. Hébrard, L. Malo, C. Reylé, C. Usher

Abstract: Context. The rotation period of stars is an important parameter along with mass, radius, effective temperature. It is an essential parameter for any radial velocity monitoring, as stellar activity can mimic the presence of a planet at the stellar rotation period. Several methods exist to measure it, including long sequences of photometric measurements or temporal series of stellar activity indicat… ▽ More Context. The rotation period of stars is an important parameter along with mass, radius, effective temperature. It is an essential parameter for any radial velocity monitoring, as stellar activity can mimic the presence of a planet at the stellar rotation period. Several methods exist to measure it, including long sequences of photometric measurements or temporal series of stellar activity indicators. Aims. Here, we use the circular polarization in near-infrared spectral lines for a sample of 43 quiet M dwarfs and compare the measured rotation periods to those obtained with other methods. Methods. From Stokes V spectropolarimetric sequences observed with SPIRou at CFHT and the data processed with the APERO pipeline, we compute the least squares deconvolution profiles using different masks of atomic stellar lines with known Landé factor appropriate to the effective temperature of the star. We derive the longitudinal magnetic field to examine its possible variation along the 50 to 200 observations of each star. For determining the stellar rotation period, we apply a Gaussian process regression enabling us to determine the rotation period of stars with evolving longitudinal field. Results. Among the 43 stars of our sample, we were able to measure a rotation period for 27 stars. For 8 stars, the rotation period was previously unknown. We find a good agreement of our rotation periods with periods found in the literature based on photometry and activity indicators and confirm that near-infrared spectropolarimetry is an important tool to measure rotation periods, even for magnetically quiet stars. Furthermore, we compute ages for 20 stars of our sample using gyrochronology. △ Less

Submitted 8 February, 2023; v1 submitted 7 February, 2023; originally announced February 2023.

Journal ref: A&A 672, A52 (2023)

Self-Supervised Speech Representation Learning: A Review

Authors: Abdelrahman Mohamed, Hung-yi Lee, Lasse Borgholt, Jakob D. Havtorn, Joakim Edin, Christian Igel, Katrin Kirchhoff, Shang-Wen Li, Karen Livescu, Lars Maaløe, Tara N. Sainath, Shinji Watanabe

Abstract: Although supervised deep learning has revolutionized speech and audio processing, it has necessitated the building of specialist models for individual tasks and application scenarios. It is likewise difficult to apply this to dialects and languages for which only limited labeled data is available. Self-supervised representation learning methods promise a single universal model that would benefit a… ▽ More Although supervised deep learning has revolutionized speech and audio processing, it has necessitated the building of specialist models for individual tasks and application scenarios. It is likewise difficult to apply this to dialects and languages for which only limited labeled data is available. Self-supervised representation learning methods promise a single universal model that would benefit a wide variety of tasks and domains. Such methods have shown success in natural language processing and computer vision domains, achieving new levels of performance while reducing the number of labels required for many downstream scenarios. Speech representation learning is experiencing similar progress in three main categories: generative, contrastive, and predictive methods. Other approaches rely on multi-modal data for pre-training, mixing text or visual data streams with speech. Although self-supervised speech representation is still a nascent research area, it is closely related to acoustic word embedding and learning with zero lexical resources, both of which have seen active research for many years. This review presents approaches for self-supervised speech representation learning and their connection to other research areas. Since many current methods focus solely on automatic speech recognition as a downstream task, we review recent efforts on benchmarking learned representations to extend the application beyond speech recognition. △ Less

Submitted 27 October, 2022; v1 submitted 21 May, 2022; originally announced May 2022.

A Brief Overview of Unsupervised Neural Speech Representation Learning

Authors: Lasse Borgholt, Jakob Drachmann Havtorn, Joakim Edin, Lars Maaløe, Christian Igel

Abstract: Unsupervised representation learning for speech processing has matured greatly in the last few years. Work in computer vision and natural language processing has paved the way, but speech data offers unique challenges. As a result, methods from other domains rarely translate directly. We review the development of unsupervised representation learning for speech over the last decade. We identify two… ▽ More Unsupervised representation learning for speech processing has matured greatly in the last few years. Work in computer vision and natural language processing has paved the way, but speech data offers unique challenges. As a result, methods from other domains rarely translate directly. We review the development of unsupervised representation learning for speech over the last decade. We identify two primary model categories: self-supervised methods and probabilistic latent variable models. We describe the models and develop a comprehensive taxonomy. Finally, we discuss and compare models from the two categories. △ Less

Submitted 1 March, 2022; originally announced March 2022.

Comments: The 2nd Workshop on Self-supervised Learning for Audio and Speech Processing (SAS) at AAAI

Model-agnostic out-of-distribution detection using combined statistical tests

Authors: Federico Bergamin, Pierre-Alexandre Mattei, Jakob D. Havtorn, Hugo Senetaire, Hugo Schmutz, Lars Maaløe, Søren Hauberg, Jes Frellsen

Abstract: We present simple methods for out-of-distribution detection using a trained generative model. These techniques, based on classical statistical tests, are model-agnostic in the sense that they can be applied to any differentiable generative model. The idea is to combine a classical parametric test (Rao's score test) with the recently introduced typicality test. These two test statistics are both th… ▽ More We present simple methods for out-of-distribution detection using a trained generative model. These techniques, based on classical statistical tests, are model-agnostic in the sense that they can be applied to any differentiable generative model. The idea is to combine a classical parametric test (Rao's score test) with the recently introduced typicality test. These two test statistics are both theoretically well-founded and exploit different sources of information based on the likelihood for the typicality test and its gradient for the score test. We show that combining them using Fisher's method overall leads to a more accurate out-of-distribution test. We also discuss the benefits of casting out-of-distribution detection as a statistical testing problem, noting in particular that false positive rate control can be valuable for practical out-of-distribution detection. Despite their simplicity and generality, these methods can be competitive with model-specific out-of-distribution detection algorithms without any assumptions on the out-distribution. △ Less

Submitted 2 March, 2022; originally announced March 2022.

Comments: Accepted at the 25th International Conference on Artificial Intelligence and Statistics (AISTATS), 2022

Benchmarking Generative Latent Variable Models for Speech

Authors: Jakob D. Havtorn, Lasse Borgholt, Søren Hauberg, Jes Frellsen, Lars Maaløe

Abstract: Stochastic latent variable models (LVMs) achieve state-of-the-art performance on natural image generation but are still inferior to deterministic models on speech. In this paper, we develop a speech benchmark of popular temporal LVMs and compare them against state-of-the-art deterministic models. We report the likelihood, which is a much used metric in the image domain, but rarely, or incomparably… ▽ More Stochastic latent variable models (LVMs) achieve state-of-the-art performance on natural image generation but are still inferior to deterministic models on speech. In this paper, we develop a speech benchmark of popular temporal LVMs and compare them against state-of-the-art deterministic models. We report the likelihood, which is a much used metric in the image domain, but rarely, or incomparably, reported for speech models. To assess the quality of the learned representations, we also compare their usefulness for phoneme recognition. Finally, we adapt the Clockwork VAE, a state-of-the-art temporal LVM for video generation, to the speech domain. Despite being autoregressive only in latent space, we find that the Clockwork VAE can outperform previous LVMs and reduce the gap to deterministic models by using a hierarchy of latent variables. △ Less

Submitted 5 April, 2022; v1 submitted 22 February, 2022; originally announced February 2022.

Comments: Accepted at the 2022 ICLR workshop on Deep Generative Models for Highly Structured Data (https://deep-gen-struct.github.io)

Do We Still Need Automatic Speech Recognition for Spoken Language Understanding?

Authors: Lasse Borgholt, Jakob Drachmann Havtorn, Mostafa Abdou, Joakim Edin, Lars Maaløe, Anders Søgaard, Christian Igel

Abstract: Spoken language understanding (SLU) tasks are usually solved by first transcribing an utterance with automatic speech recognition (ASR) and then feeding the output to a text-based model. Recent advances in self-supervised representation learning for speech data have focused on improving the ASR component. We investigate whether representation learning for speech has matured enough to replace ASR i… ▽ More Spoken language understanding (SLU) tasks are usually solved by first transcribing an utterance with automatic speech recognition (ASR) and then feeding the output to a text-based model. Recent advances in self-supervised representation learning for speech data have focused on improving the ASR component. We investigate whether representation learning for speech has matured enough to replace ASR in SLU. We compare learned speech features from wav2vec 2.0, state-of-the-art ASR transcripts, and the ground truth text as input for a novel speech-based named entity recognition task, a cardiac arrest detection task on real-world emergency calls and two existing SLU benchmarks. We show that learned speech features are superior to ASR transcripts on three classification tasks. For machine translation, ASR transcripts are still the better choice. We highlight the intrinsic robustness of wav2vec 2.0 representations to out-of-vocabulary words as key to better performance. △ Less

Submitted 29 November, 2021; originally announced November 2021.

Comments: Under review as a conference paper at ICASSP 2022

TOI-1278 B: SPIRou unveils a rare Brown Dwarf Companion in Close-In Orbit around an M dwarf

Authors: Étienne Artigau, Guillaume Hébrard, Charles Cadieux, Thomas Vandal, Neil J. Cook, René Doyon, Jonathan Gagné, Claire Moutou, Eder Martioli, Antonio Frasca, Farbod Jahandar, David Lafrenière, Lison Malo, Jean-François Donati, Pia Cortes-Zuleta, Isabelle Boisse, Xavier Delfosse, Andres Carmona, Pascal Fouqué, Julien Morin, Jason Rowe, Giuseppe Marino, Riccardo Papini, David R. Ciardi, Michael B. Lund , et al. (17 additional authors not shown)

Abstract: We present the discovery of an $18.5\pm0.5$M$_{\rm Jup}$ brown dwarf (BD) companion to the M0V star TOI-1278. The system was first identified through a percent-deep transit in TESS photometry; further analysis showed it to be a grazing transit of a Jupiter-sized object. Radial velocity (RV) follow-up with the SPIRou near-infrared high-resolution velocimeter and spectropolarimeter in the framework… ▽ More We present the discovery of an $18.5\pm0.5$M$_{\rm Jup}$ brown dwarf (BD) companion to the M0V star TOI-1278. The system was first identified through a percent-deep transit in TESS photometry; further analysis showed it to be a grazing transit of a Jupiter-sized object. Radial velocity (RV) follow-up with the SPIRou near-infrared high-resolution velocimeter and spectropolarimeter in the framework of the 300-night SPIRou Legacy Survey (SLS) carried out at the Canada-France-Hawaii Telescope (CFHT) led to the detection of a Keplerian RV signal with a semi-amplitude of $2306\pm10$ m/s in phase with the 14.5-day transit period, having a slight but non-zero eccentricity. The intermediate-mass ratio ($M_\star/M_{\rm{comp}} \sim31$) is unique for having such a short separation ($0.095\pm0.001$ AU) among known M-dwarf systems. Interestingly, M dwarf-brown dwarf systems with similar mass ratios exist with separations of tens to thousands of AUs. △ Less

Submitted 8 June, 2021; originally announced June 2021.

Comments: accepted for publication in The Astronomical Journal

Do End-to-End Speech Recognition Models Care About Context?

Authors: Lasse Borgholt, Jakob Drachmann Havtorn, Željko Agić, Anders Søgaard, Lars Maaløe, Christian Igel

Abstract: The two most common paradigms for end-to-end speech recognition are connectionist temporal classification (CTC) and attention-based encoder-decoder (AED) models. It has been argued that the latter is better suited for learning an implicit language model. We test this hypothesis by measuring temporal context sensitivity and evaluate how the models perform when we constrain the amount of contextual… ▽ More The two most common paradigms for end-to-end speech recognition are connectionist temporal classification (CTC) and attention-based encoder-decoder (AED) models. It has been argued that the latter is better suited for learning an implicit language model. We test this hypothesis by measuring temporal context sensitivity and evaluate how the models perform when we constrain the amount of contextual information in the audio input. We find that the AED model is indeed more context sensitive, but that the gap can be closed by adding self-attention to the CTC model. Furthermore, the two models perform similarly when contextual information is constrained. Finally, in contrast to previous research, our results show that the CTC model is highly competitive on WSJ and LibriSpeech without the help of an external language model. △ Less

Submitted 17 February, 2021; originally announced February 2021.

Comments: Published in the proceedings of INTERSPEECH 2020, pp. 4352-4356

Hierarchical VAEs Know What They Don't Know

Authors: Jakob D. Havtorn, Jes Frellsen, Søren Hauberg, Lars Maaløe

Abstract: Deep generative models have been demonstrated as state-of-the-art density estimators. Yet, recent work has found that they often assign a higher likelihood to data from outside the training distribution. This seemingly paradoxical behavior has caused concerns over the quality of the attained density estimates. In the context of hierarchical variational autoencoders, we provide evidence to explain… ▽ More Deep generative models have been demonstrated as state-of-the-art density estimators. Yet, recent work has found that they often assign a higher likelihood to data from outside the training distribution. This seemingly paradoxical behavior has caused concerns over the quality of the attained density estimates. In the context of hierarchical variational autoencoders, we provide evidence to explain this behavior by out-of-distribution data having in-distribution low-level features. We argue that this is both expected and desirable behavior. With this insight in hand, we develop a fast, scalable and fully unsupervised likelihood-ratio score for OOD detection that requires data to be in-distribution across all feature-levels. We benchmark the method on a vast set of data and model combinations and achieve state-of-the-art results on out-of-distribution detection. △ Less

Submitted 18 January, 2022; v1 submitted 16 February, 2021; originally announced February 2021.

Comments: Appeared in Proceedings of the 38th International Conference on Machine Learning (ICML 2021). 18 pages, source code available at https://github.com/JakobHavtorn/hvae-oodd, https://github.com/vlievin/biva-pytorch and https://github.com/larsmaaloee/BIVA

On Scaling Contrastive Representations for Low-Resource Speech Recognition

Authors: Lasse Borgholt, Tycho Max Sylvester Tax, Jakob Drachmann Havtorn, Lars Maaløe, Christian Igel

Abstract: Recent advances in self-supervised learning through contrastive training have shown that it is possible to learn a competitive speech recognition system with as little as 10 minutes of labeled data. However, these systems are computationally expensive since they require pre-training followed by fine-tuning in a large parameter space. We explore the performance of such systems without fine-tuning b… ▽ More Recent advances in self-supervised learning through contrastive training have shown that it is possible to learn a competitive speech recognition system with as little as 10 minutes of labeled data. However, these systems are computationally expensive since they require pre-training followed by fine-tuning in a large parameter space. We explore the performance of such systems without fine-tuning by training a state-of-the-art speech recognizer on the fixed representations from the computationally demanding wav2vec 2.0 framework. We find performance to decrease without fine-tuning and, in the extreme low-resource setting, wav2vec 2.0 is inferior to its predecessor. In addition, we find that wav2vec 2.0 representations live in a low dimensional subspace and that decorrelating the features of the representations can stabilize training of the automatic speech recognizer. Finally, we propose a bidirectional extension to the original wav2vec framework that consistently improves performance. △ Less

Submitted 1 February, 2021; originally announced February 2021.

Comments: © 2021 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works

HIRES, the high-resolution spectrograph for the ELT

Authors: Alessandro Marconi, Manuel Abreu, Vardan Adibekyan, Matteo Aliverti, Carlos Allende Prieto, Pedro J. Amado, Manuel Amate, Etienne Artigau, Sergio R. Augusto, Susana Barros, Santiago Becerril, Bjorn Benneke, Edwin Bergin, Philippe Berio, Naidu Bezawada, Isabelle Boisse, Xavier Bonfils, Francois Bouchy, Christopher Broeg, Alexandre Cabral, Rocio Calvo-Ortega, Bruno Leonardo Canto Martins, Bruno Chazelas, Andrea Chiavassa, Lise B. Christensen , et al. (77 additional authors not shown)

Abstract: HIRES will be the high-resolution spectrograph of the European Extremely Large Telescope at optical and near-infrared wavelengths. It consists of three fibre-fed spectrographs providing a wavelength coverage of 0.4-1.8 mic (goal 0.35-1.8 mic) at a spectral resolution of ~100,000. The fibre-feeding allows HIRES to have several, interchangeable observing modes including a SCAO module and a small dif… ▽ More HIRES will be the high-resolution spectrograph of the European Extremely Large Telescope at optical and near-infrared wavelengths. It consists of three fibre-fed spectrographs providing a wavelength coverage of 0.4-1.8 mic (goal 0.35-1.8 mic) at a spectral resolution of ~100,000. The fibre-feeding allows HIRES to have several, interchangeable observing modes including a SCAO module and a small diffraction-limited IFU in the NIR. Therefore, it will be able to operate both in seeing and diffraction-limited modes. ELT-HIRES has a wide range of science cases spanning nearly all areas of research in astrophysics and even fundamental physics. Some of the top science cases will be the detection of bio signatures from exoplanet atmospheres, finding the fingerprints of the first generation of stars (PopIII), tests on the stability of Nature's fundamental couplings, and the direct detection of the cosmic acceleration. The HIRES consortium is composed of more than 30 institutes from 14 countries, forming a team of more than 200 scientists and engineers. △ Less

Submitted 24 November, 2020; originally announced November 2020.

Comments: to appear in the ESO Messenger No.182, December 2020

SPIRou: nIR velocimetry & spectropolarimetry at the CFHT

Authors: J. -F. Donati, D. Kouach, C. Moutou, R. Doyon, X. Delfosse, E. Artigau, S. Baratchart, M. Lacombe, G. Barrick, G. Hebrard, F. Bouchy, L. Saddlemyer, L. Pares, P. Rabou, Y. Micheau, F. Dolon, V. Reshetov, Z. Challita, A. Carmona, N. Striebig, S. Thibault, E. Martioli, N. Cook, P. Fouque, T. Vermeulen , et al. (41 additional authors not shown)

Abstract: This paper presents an overview of SPIRou, the new-generation near-infrared spectropolarimeter / precision velocimeter recently installed on the 3.6-m Canada-France-Hawaii Telescope (CFHT). Starting from the two main science goals, namely the quest for planetary systems around nearby M dwarfs and the study of magnetized star / planet formation, we outline the instrument concept that was designed t… ▽ More This paper presents an overview of SPIRou, the new-generation near-infrared spectropolarimeter / precision velocimeter recently installed on the 3.6-m Canada-France-Hawaii Telescope (CFHT). Starting from the two main science goals, namely the quest for planetary systems around nearby M dwarfs and the study of magnetized star / planet formation, we outline the instrument concept that was designed to efficiently address these forefront topics, and detail the in-lab and on-sky instrument performances measured throughout the intensive testing phase that SPIRou was submitted to before passing the final acceptance review in early 2019 and initiating science observations. With a central position among the newly started programmes, the SPIRou Legacy Survey (SLS) Large Programme was allocated 300 CFHT nights until at least mid 2022. We also briefly describe a few of the first results obtained in the various science topics that SPIRou started investigating, focusing in particular on planetary systems of nearby M dwarfs, transiting exoplanets and their atmospheres, magnetic fields of young stars, but also on alternate science goals like the atmospheres of M dwarfs and the Earth's atmosphere. We finally conclude on the essential role that SPIRou and the CFHT can play in coordination with forthcoming major facilities like the JWST, the ELTs, PLATO and ARIEL over the decade. △ Less

Submitted 20 August, 2020; originally announced August 2020.

Comments: MNRAS, in press (20 pages, 18 figures, 1 table)

MultiQT: Multimodal Learning for Real-Time Question Tracking in Speech

Authors: Jakob D. Havtorn, Jan Latko, Joakim Edin, Lasse Borgholt, Lars Maaløe, Lorenzo Belgrano, Nicolai F. Jacobsen, Regitze Sdun, Željko Agić

Abstract: We address a challenging and practical task of labeling questions in speech in real time during telephone calls to emergency medical services in English, which embeds within a broader decision support system for emergency call-takers. We propose a novel multimodal approach to real-time sequence labeling in speech. Our model treats speech and its own textual representation as two separate modalitie… ▽ More We address a challenging and practical task of labeling questions in speech in real time during telephone calls to emergency medical services in English, which embeds within a broader decision support system for emergency call-takers. We propose a novel multimodal approach to real-time sequence labeling in speech. Our model treats speech and its own textual representation as two separate modalities or views, as it jointly learns from streamed audio and its noisy transcription into text via automatic speech recognition. Our results show significant gains of jointly learning from the two modalities when compared to text or audio only, under adverse noise and limited volume of training data. The results generalize to medical symptoms detection where we observe a similar pattern of improvements with multimodal learning. △ Less

Submitted 12 May, 2020; v1 submitted 2 May, 2020; originally announced May 2020.

Comments: Accepted at ACL 2020

The magnetic field and accretion regime of CI Tau

Authors: JF Donati, J Bouvier, SH Alencar, C Moutou, L Malo, M Takami, F Menard, C Dougados, GA Hussain, the MaTYSSE collaboration

Abstract: This paper exploits spectropolarimetric data of the classical T Tauri star CI Tau collected with ESPaDOnS at the Canada-France-Hawaii Telescope, with the aims of detecting and characterizing the large-scale magnetic field that the star hosts, and of investigating how the star interacts with the inner regions of its accretion disc through this field. Our data unambiguously show that CI Tau has a ro… ▽ More This paper exploits spectropolarimetric data of the classical T Tauri star CI Tau collected with ESPaDOnS at the Canada-France-Hawaii Telescope, with the aims of detecting and characterizing the large-scale magnetic field that the star hosts, and of investigating how the star interacts with the inner regions of its accretion disc through this field. Our data unambiguously show that CI Tau has a rotation period of 9.0d, and that it hosts a strong, mainly poloidal large-scale field. Accretion at the surface of the star concentrates within a bright high-latitude chromospheric region that spatially overlaps with a large dark photospheric spot, in which the radial magnetic field reaches -3.7kG. With a polar strength of -1.7kG, the dipole component of the large-scale field is able to evacuate the central regions of the disc up to about 50% of the co-rotation radius (at which the Keplerian orbital period equals the stellar rotation period) throughout our observations, during which the average accretion rate was found to be unusually high. We speculate that the magnetic field of CI Tau is strong enough to sustain most of the time a magnetospheric gap extending to at least 70% of the co-rotation radius, which would explain why the rotation period of CI Tau is as long as 9d. Our results also imply that the 9d radial velocity (RV) modulation that CI Tau exhibits is attributable to stellar activity, and thus that the existence of the candidate close-in massive planet CI Tau b to which these RV fluctuations were first attributed needs to be reassessed with new evidence. △ Less

Submitted 28 November, 2019; originally announced November 2019.

Comments: MNRAS, in press (10 pages, 6 figures, 2 tables)

BIVA: A Very Deep Hierarchy of Latent Variables for Generative Modeling

Authors: Lars Maaløe, Marco Fraccaro, Valentin Liévin, Ole Winther

Abstract: With the introduction of the variational autoencoder (VAE), probabilistic latent variable models have received renewed attention as powerful generative models. However, their performance in terms of test likelihood and quality of generated samples has been surpassed by autoregressive models without stochastic units. Furthermore, flow-based models have recently been shown to be an attractive altern… ▽ More With the introduction of the variational autoencoder (VAE), probabilistic latent variable models have received renewed attention as powerful generative models. However, their performance in terms of test likelihood and quality of generated samples has been surpassed by autoregressive models without stochastic units. Furthermore, flow-based models have recently been shown to be an attractive alternative that scales well to high-dimensional data. In this paper we close the performance gap by constructing VAE models that can effectively utilize a deep hierarchy of stochastic variables and model complex covariance structures. We introduce the Bidirectional-Inference Variational Autoencoder (BIVA), characterized by a skip-connected generative model and an inference network formed by a bidirectional stochastic inference path. We show that BIVA reaches state-of-the-art test likelihoods, generates sharp and coherent natural images, and uses the hierarchy of latent variables to capture different aspects of the data distribution. We observe that BIVA, in contrast to recent results, can be used for anomaly detection. We attribute this to the hierarchy of latent variables which is able to extract high-level semantic features. Finally, we extend BIVA to semi-supervised classification tasks and show that it performs comparably to state-of-the-art results by generative adversarial networks. △ Less

Submitted 6 November, 2019; v1 submitted 6 February, 2019; originally announced February 2019.

On the Inductive Bias of Word-Character-Level Multi-Task Learning for Speech Recognition

Authors: Jan Kremer, Lasse Borgholt, Lars Maaløe

Abstract: End-to-end automatic speech recognition (ASR) commonly transcribes audio signals into sequences of characters while its performance is evaluated by measuring the word-error rate (WER). This suggests that predicting sequences of words directly may be helpful instead. However, training with word-level supervision can be more difficult due to the sparsity of examples per label class. In this paper we… ▽ More End-to-end automatic speech recognition (ASR) commonly transcribes audio signals into sequences of characters while its performance is evaluated by measuring the word-error rate (WER). This suggests that predicting sequences of words directly may be helpful instead. However, training with word-level supervision can be more difficult due to the sparsity of examples per label class. In this paper we analyze an end-to-end ASR model that combines a word-and-character representation in a multi-task learning (MTL) framework. We show that it improves on the WER and study how the word-level model can benefit from character-level supervision by analyzing the learned inductive preference bias of each model component empirically. We find that by adding character-level supervision, the MTL model interpolates between recognizing more frequent words (preferred by the word-level model) and shorter words (preferred by the character-level model). △ Less

Submitted 28 November, 2018; originally announced December 2018.

Comments: Accepted at the IRASL workshop at NeurIPS 2018

The magnetic propeller accretion regime of LkCa 15

Authors: J-F Donati, J Bouvier, SH Alencar, C Hill, A Carmona, CP Folsom, F Menard, SG Gregory, GA Hussain, K Grankin, C Moutou, L Malo, M Takami, GJ Herczeg, the MaTYSSE collaboration

Abstract: We present a spectropolarimetric study of the classical T Tauri star (cTTS) LkCa 15 investigating the large-scale magnetic topology of the central star and the way the field connects to the inner regions of the accretion disc. We find that the star hosts a strong poloidal field with a mainly axisymmetric dipole component of 1.35 kG, whereas the mass accretion rate at the surface of the star is… ▽ More We present a spectropolarimetric study of the classical T Tauri star (cTTS) LkCa 15 investigating the large-scale magnetic topology of the central star and the way the field connects to the inner regions of the accretion disc. We find that the star hosts a strong poloidal field with a mainly axisymmetric dipole component of 1.35 kG, whereas the mass accretion rate at the surface of the star is $10^{-9.2}$ $\hbox{${\rm M}_{\odot}$ yr$^{-1}$}$. It implies that the magnetic field of LkCa 15 is able to evacuate the central regions of the disc up to a distance of 0.07 au at which the Keplerian orbital period equals the stellar rotation period. Our results suggest that LkCa 15, like the lower-mass cTTS AA Tau, interacts with its disc in a propeller mode, a regime supposedly very efficient at slowing down the rotation of cTTSs hosting strong dipolar fields. △ Less

Submitted 12 November, 2018; originally announced November 2018.

Comments: MNRAS letter, in press (3 figures, 1 table)

Inner disk structure of the classical T Tauri star LkCa 15

Authors: S. H. P. Alencar, J. Bouvier, J. -F. Donati, E. Alecian, C. P. Folsom, K. Grankin, G. A. J. Hussain, C. Hill, A. -M. Cody, A. Carmona, C. Dougados, S. G. Gregory, G. Herczeg, F. Ménard, C. Moutou, L. Malo, M. Takami, the MaTYSSE collaboration

Abstract: Magnetospheric accretion has been thoroughly studied in young stellar systems with full non-evolved accretion disks, but it is poorly documented for transition disk objects with large inner cavities. We aim at characterizing the star-disk interaction and the accretion process onto the central star of LkCa 15, a transition disk system with an inner dust cavity. We obtained quasi-simultaneous photom… ▽ More Magnetospheric accretion has been thoroughly studied in young stellar systems with full non-evolved accretion disks, but it is poorly documented for transition disk objects with large inner cavities. We aim at characterizing the star-disk interaction and the accretion process onto the central star of LkCa 15, a transition disk system with an inner dust cavity. We obtained quasi-simultaneous photometric and spectropolarimetric observations of the system over several rotational periods. We analyzed the system light curve, as well as changes in spectral continuum and line profile to derive the properties of the accretion flow from the edge of the inner disk to the central star. We also derived magnetic field measurements at the stellar surface. We find that the system exhibits magnetic, photometric, and spectroscopic variability with a period of about 5.70 days. The light curve reveals a periodic dip, which suggests the presence of an inner disk warp that is located at the corotation radius at about 0.06 au from the star. Line profile variations and veiling variability are consistent with a magnetospheric accretion model where the funnel flows reach the star at high latitudes. This leads to the development of an accretion shock close to the magnetic poles. All diagnostics point to a highly inclined inner disk that interacts with the stellar magnetosphere. The spectroscopic and photometric variability of LkCa 15 is remarkably similar to that of AA Tau, the prototype of periodic dippers. We therefore suggest that the origin of the variability is a rotating disk warp that is located at the inner edge of a highly inclined disk close to the star. This contrasts with the moderate inclination of the outer transition disk seen on the large scale and thus provides evidence for a significant misalignment between the inner and outer disks of this planet-forming transition disk system. △ Less

Submitted 12 November, 2018; originally announced November 2018.

Comments: 16 pages, 22 figures, accepted by Astronomy & Astrophysics

Journal ref: A&A 620, A195 (2018)

WEIRD: Wide-orbit Exoplanet search with InfraRed Direct imaging

Authors: Frédérique Baron, Étienne Artigau, Julien Rameau, David Lafrenière, Jonathan Gagné, Lison Malo, Loïc Albert, Marie-Eve Naud, René Doyon, Markus Janson, Philippe Delorme, Charles Beichman

Abstract: We report results from the Wide-orbit Exoplanet search with InfraRed Direct imaging (WEIRD), a survey designed to search for Jupiter-like companions on very wide orbits (1000 to 5000 AU) around young stars ($<$120 Myr) that are known members of moving groups in the solar neighborhood ($<$70 pc). Sharing the same age, distance, and metallicity as their host while being on large enough orbits to be… ▽ More We report results from the Wide-orbit Exoplanet search with InfraRed Direct imaging (WEIRD), a survey designed to search for Jupiter-like companions on very wide orbits (1000 to 5000 AU) around young stars ($<$120 Myr) that are known members of moving groups in the solar neighborhood ($<$70 pc). Sharing the same age, distance, and metallicity as their host while being on large enough orbits to be studied as "isolated" objects make such companions prime targets for spectroscopic observations and valuable benchmark objects for exoplanet atmosphere models. The search strategy is based on deep imaging in multiple bands across the near-infrared domain. For all 177 objects of our sample, $z_{ab}^\prime$, $J$, [3.6] and [4.5] images were obtained with CFHT/MegaCam, GEMINI/GMOS, CFHT/WIRCam, GEMINI/Flamingos-2, and $Spitzer$/IRAC. Using this set of 4 images per target, we searched for sources with red $z_{ab}^\prime$ and $[3.6]-[4.5]$ colors, typically reaching good completeness down to 2Mjup companions, while going down to 1Mjup for some targets, at separations of $1000-5000$ AU. The search yielded 4 candidate companions with the expected colors, but they were all rejected through follow-up proper motion observations. Our results constrain the occurrence of 1-13 Mjup planetary-mass companions on orbits with a semi-major axis between 1000 and 5000 AU at less than 0.03, with a 95\% confidence level. △ Less

Submitted 23 July, 2018; originally announced July 2018.

Comments: 55 pages, 16 figures, accepted to AJ

The Solar Neighborhood. XLIII: Discovery of New Nearby Stars with pm < 0.18 arcsec yr-1 (TINYMO sample)

Authors: Adric R. Riedel, Michele L. Silverstein, Todd J. Henry, Wei-Chun Jao, Jennifer G. Winters, John P. Subasavage, Lison Malo, Nigel C. Hambly

Abstract: We have conducted a novel search of most of the southern sky for nearby red dwarfs having low proper motions, with specific emphasis on those with proper motion < 0.18 arcsec yr-1, the lower cutoff of Luyten's classic proper motion catalog. We used a tightly constrained search of the SuperCOSMOS database and a suite of photometric distance relations for photographic BRI and 2MASS JHKs magnitudes t… ▽ More We have conducted a novel search of most of the southern sky for nearby red dwarfs having low proper motions, with specific emphasis on those with proper motion < 0.18 arcsec yr-1, the lower cutoff of Luyten's classic proper motion catalog. We used a tightly constrained search of the SuperCOSMOS database and a suite of photometric distance relations for photographic BRI and 2MASS JHKs magnitudes to estimate distances to more than fourteen million red dwarf candidates. Here we discuss 29 stars in 26 systems estimated to be within 25 parsecs, all of which have pm < 0.18 arcsec yr-1, which we have investigated using milliarcsecond astrometry, VRI photometry, and low-resolution spectroscopy. In total, we present the first parallaxes of 20 star systems, nine of which are within 25 parsecs. We have additionally identified eight young M dwarfs, of which two are new members of the nearby young moving groups, and 72 new giants, including two new carbon stars. We also present the entire catalog of 1215 sources we have identified by this means. △ Less

Submitted 23 April, 2018; originally announced April 2018.

Comments: Submitted to the Astronomical Journal

BANYAN. XII. New Members of Nearby Young Associations from Gaia-Tycho Data

Authors: Jonathan Gagné, Olivier Roy-Loubier, Jacqueline K. Faherty, René Doyon, Lison Malo

Abstract: We present a search for stellar members of young associations within 150 pc of the Sun based on Gaia-Tycho and an updated version of the BANYAN $Σ$ software to determine Bayesian membership probabilities that includes Gaia-2MASS color-magnitude diagrams. We identify 32 new F0-M3-type bona fide members of the 10-200 Myr-old Sco-Cen, Carina, Tucana-Horologium, Columba and Octans associations and the… ▽ More We present a search for stellar members of young associations within 150 pc of the Sun based on Gaia-Tycho and an updated version of the BANYAN $Σ$ software to determine Bayesian membership probabilities that includes Gaia-2MASS color-magnitude diagrams. We identify 32 new F0-M3-type bona fide members of the 10-200 Myr-old Sco-Cen, Carina, Tucana-Horologium, Columba and Octans associations and the AB Doradus, $β$ Pictoris and Carina-Near moving groups. These new bona fide members have measurements of their full kinematics and literature data consistent with a young age. We also confirm the membership of 66 previously known candidate members using their Gaia-Tycho trigonometric distances or new literature radial velocities, and identify 219 additional new candidate members, most of which do not yet have a radial velocity measurement. This work is the first step towards a completeness-corrected survey of young association members based on Gaia-DR2 in the near future. △ Less

Submitted 9 April, 2018; originally announced April 2018.

Comments: Accepted for publication in ApJ. 34 pages, 11 figures and 4 tables

Optical and Near-Infrared Radial Velocity Content of M Dwarfs: Testing Models with Barnard's Star

Authors: Étienne Artigau, Lison Malo, René Doyon, Pedro Figueira, Xavier Delfosse, Nicola Astudillo-Defru

Abstract: High precision radial velocity (RV) measurements have been central in the study of exoplanets during the last two decades, from the early discovery of hot Jupiters, to the recent mass measurements of Earth-sized planets uncovered by transit surveys. While optical radial-velocity is now a mature field, there is currently a strong effort to push the technique into the near-infrared (nIR) domain (chi… ▽ More High precision radial velocity (RV) measurements have been central in the study of exoplanets during the last two decades, from the early discovery of hot Jupiters, to the recent mass measurements of Earth-sized planets uncovered by transit surveys. While optical radial-velocity is now a mature field, there is currently a strong effort to push the technique into the near-infrared (nIR) domain (chiefly $Y$, $J$, $H$ and $K$ band passes) to probe planetary systems around late-type stars. The combined lower mass and luminosity of M dwarfs leads to an increased reflex RV signal for planets in the habitable zone compared to Sun-like stars. The estimates on the detectability of planets rely on various instrumental characteristics, but also on a prior knowledge of the stellar spectrum. While the overall properties of M dwarf spectra have been extensively tested against observations, the same is not true for their detailed line profiles, which leads to significant uncertainties when converting a given signal-to-noise ratio to a corresponding RV precision as attainable on a given spectrograph. By combining archival CRIRES and HARPS data with ESPaDOnS data of Barnard's star, we show that state-of-the-art atmosphere models over-predict the $Y$ and $J$-band RV content by more than a factor of $\sim$$2$, while under-predicting the $H$ and $K$-band content by half. △ Less

Submitted 20 March, 2018; originally announced March 2018.

Comments: accepted for publication in AJ

BANYAN. XI. The BANYAN $Σ$ multivariate Bayesian algorithm to identify members of young associations within 150 pc

Authors: Jonathan Gagné, Eric E. Mamajek, Lison Malo, Adric Riedel, David Rodriguez, David Lafrenière, Jacqueline K. Faherty, Olivier Roy-Loubier, Laurent Pueyo, Annie C. Robin, René Doyon

Abstract: BANYAN $Σ$ is a new Bayesian algorithm to identify members of young stellar associations within 150 pc of the Sun. It includes 27 young associations with ages in the range ~1-800 Myr, modelled with multivariate Gaussians in 6-dimensional XYZUVW space. It is the first such multi-association classification tool to include the nearest sub-groups of the Sco-Cen OB star-forming region, the IC 2602, IC… ▽ More BANYAN $Σ$ is a new Bayesian algorithm to identify members of young stellar associations within 150 pc of the Sun. It includes 27 young associations with ages in the range ~1-800 Myr, modelled with multivariate Gaussians in 6-dimensional XYZUVW space. It is the first such multi-association classification tool to include the nearest sub-groups of the Sco-Cen OB star-forming region, the IC 2602, IC 2391, Pleiades and Platais 8 clusters, and the $ρ$ Ophiuchi, Corona Australis, and Taurus star-formation regions. A model of field stars is built from a mixture of multivariate Gaussians based on the Besançon Galactic model. The algorithm can derive membership probabilities for objects with only sky coordinates and proper motion, but can also include parallax and radial velocity measurements, as well as spectrophotometric distance constraints from sequences in color-magnitude or spectral type-magnitude diagrams. BANYAN $Σ$ benefits from an analytical solution to the Bayesian marginalization integrals that makes it more accurate and significantly faster than its predecessor BANYAN II. A contamination versus hit rate analysis is presented and demonstrates that BANYAN $Σ$ achieves a better classification performance than other moving group classification tools, especially in terms of cross-contamination between young associations. An updated list of bona fide members in the 27 young associations, augmented by the Gaia-DR1 release, are presented. This new tool will make it possible to analyze large data sets such as the upcoming Gaia-DR2 to identify new young stars. IDL and Python versions of BANYAN $Σ$ are made available with this publication. (shortened) △ Less

Submitted 3 February, 2018; v1 submitted 27 January, 2018; originally announced January 2018.

Comments: 52 pages, 12 figures, 11 tables. Accepted for publication in the Astrophysical Journal Supplement Series. Minor corrections and updated URLs

Strong dipole magnetic fields in fast rotating fully convective stars

Authors: D. Shulyak A. Reiners, A. Engeln, L. Malo, R. Yadav, J. Morin, O. Kochukhov

Abstract: M dwarfs are the most numerous stars in our Galaxy with masses between approximately 0.5 and 0.1 solar mass. Many of them show surface activity qualitatively similar to our Sun and generate flares, high X-ray fluxes, and large-scale magnetic fields. Such activity is driven by a dynamo powered by the convective motions in their interiors. Understanding properties of stellar magnetic fields in these… ▽ More M dwarfs are the most numerous stars in our Galaxy with masses between approximately 0.5 and 0.1 solar mass. Many of them show surface activity qualitatively similar to our Sun and generate flares, high X-ray fluxes, and large-scale magnetic fields. Such activity is driven by a dynamo powered by the convective motions in their interiors. Understanding properties of stellar magnetic fields in these stars finds a broad application in astrophysics, including, e.g., theory of stellar dynamos and environment conditions around planets that may be orbiting these stars. Most stars with convective envelopes follow a rotation-activity relationship where various activity indicators saturate in stars with rotation periods shorter than a few days. The activity gradually declines with rotation rate in stars rotating more slowly. It is thought that due to a tight empirical correlation between X-ray and magnetic flux, the stellar magnetic fields will also saturate, to values around ~4kG. Here we report the detection of magnetic fields above the presumed saturation limit in four fully convective M-dwarfs. By combining results from spectroscopic and polarimetric studies we explain our findings in terms of bistable dynamo models: stars with the strongest magnetic fields are those in a dipole dynamo state, while stars in a multipole state cannot generate fields stronger than about four kilogauss. Our study provides observational evidence that dynamo in fully convective M dwarfs generates magnetic fields that can differ not only in the geometry of their large scale component, but also in the total magnetic energy. △ Less

Submitted 25 January, 2018; originally announced January 2018.

Comments: Version accepted for publication by Nature Astronomy (13 June 2017; https://www.nature.com/articles/s41550-017-0184)

Journal ref: Nature Astronomy, Volume 1, id. 0184 (2017)

The First Post-Kepler Brightness Dips of KIC 8462852

Authors: Tabetha S. Boyajian, Roi Alonso, Alex Ammerman, David Armstrong, A. Asensio Ramos, K. Barkaoui, Thomas G. Beatty, Z. Benkhaldoun, Paul Benni, Rory Bentley, Andrei Berdyugin, Svetlana Berdyugina, Serge Bergeron, Allyson Bieryla, Michaela G. Blain, Alicia Capetillo Blanco, Eva H. L. Bodman, Anne Boucher, Mark Bradley, Stephen M. Brincat, Thomas G. Brink, John Briol, David J. A. Brown, J. Budaj, A. Burdanov , et al. (181 additional authors not shown)

Abstract: We present a photometric detection of the first brightness dips of the unique variable star KIC 8462852 since the end of the Kepler space mission in 2013 May. Our regular photometric surveillance started in October 2015, and a sequence of dip** began in 2017 May continuing on through the end of 2017, when the star was no longer visible from Earth. We distinguish four main 1-2.5% dips, named "Els… ▽ More We present a photometric detection of the first brightness dips of the unique variable star KIC 8462852 since the end of the Kepler space mission in 2013 May. Our regular photometric surveillance started in October 2015, and a sequence of dip** began in 2017 May continuing on through the end of 2017, when the star was no longer visible from Earth. We distinguish four main 1-2.5% dips, named "Elsie," "Celeste," "Skara Brae," and "Angkor", which persist on timescales from several days to weeks. Our main results so far are: (i) there are no apparent changes of the stellar spectrum or polarization during the dips; (ii) the multiband photometry of the dips shows differential reddening favoring non-grey extinction. Therefore, our data are inconsistent with dip models that invoke optically thick material, but rather they are in-line with predictions for an occulter consisting primarily of ordinary dust, where much of the material must be optically thin with a size scale <<1um, and may also be consistent with models invoking variations intrinsic to the stellar photosphere. Notably, our data do not place constraints on the color of the longer-term "secular" dimming, which may be caused by independent processes, or probe different regimes of a single process. △ Less

Submitted 2 January, 2018; originally announced January 2018.

Comments: 19 pages, 8 figures, accepted for publication in ApJL

Predictions of planet detections with near infrared radial velocities in the up-coming SPIRou Legacy Survey-Planet Search

Authors: Ryan Cloutier, Étienne Artigau, Xavier Delfosse, Lison Malo, Claire Moutou, René Doyon, Jean-Francois Donati, Andrew Cumming, Xavier Dumusque, Élodie Hébrard, Kristen Menou

Abstract: The SPIRou near infrared spectro-polarimeter is destined to begin science operations at the Canada-France-Hawaii Telescope in mid-2018. One of the instrument's primary science goals is to discover the closest exoplanets to the Solar System by conducting a 3-5 year long radial velocity survey of nearby M dwarfs at an expected precision of $\sim 1$ m s$^{-1}$; the SPIRou Legacy Survey-Planet Search… ▽ More The SPIRou near infrared spectro-polarimeter is destined to begin science operations at the Canada-France-Hawaii Telescope in mid-2018. One of the instrument's primary science goals is to discover the closest exoplanets to the Solar System by conducting a 3-5 year long radial velocity survey of nearby M dwarfs at an expected precision of $\sim 1$ m s$^{-1}$; the SPIRou Legacy Survey-Planet Search (SLS-PS). In this study we conduct a detailed Monte-Carlo simulation of the SLS-PS using our current understanding of the occurrence rate of M dwarf planetary systems and physical models of stellar activity. From simultaneous modelling of planetary signals and activity, we predict the population of planets detected in the SLS-PS. With our fiducial survey strategy and expected instrument performance over a nominal survey length of $\sim 3$ years, we expect SPIRou to detect $85.3^{+29.3}_{-12.4}$ planets including $20.0^{+16.8}_{-7.2}$ habitable zone planets and $8.1^{+7.6}_{-3.2}$ Earth-like planets from a sample of 100 M1-M8.5 dwarfs out to 11 pc. By studying mid-to-late M dwarfs previously inaccessible to existing optical velocimeters, SPIRou will put meaningful constraints on the occurrence rate of planets around those stars including the value of $η_{\oplus}$ at an expected level of precision of $\lesssim 45$%. We also predict a subset of $46.7^{+16.0}_{-6.0}$ planets may be accessible with dedicated high-contrast imagers on the next generation of ELTs including $4.9^{+4.7}_{-2.0}$ potentially imagable Earth-like planets. Lastly, we compare the results of our fiducial survey strategy to other foreseeable survey versions to quantify which strategy is optimized to reach the SLS-PS science goals. The results of our simulations are made available to the community on github. △ Less

Submitted 18 December, 2017; originally announced December 2017.

Comments: Accepted for publication in AJ. 19 figures including 15 interactive figures when viewed in Adobe Acrobat. Simulation results provided at https://github.com/r-cloutier/SLSPS_Simulations

SPIRou Input Catalogue: Global properties of 440 M dwarfs observed with ESPaDOnS at CFHT

Authors: Pascal Fouqué, Claire Moutou, Lison Malo, Eder Martioli, Olivia Lim, Arvind Rajpurohit, Etienne Artigau, Xavier Delfosse, Jean-François Donati, Thierry Forveille, Julien Morin, France Allard, Raphaël Delage, René Doyon, Elodie Hébrard, Vasco Neves

Abstract: Present and future high-precision radial-velocity spectrometers dedicated to the discovery of low-mass planets orbiting low-mass dwarfs need to focus on the best selected stars to make an efficient use of telescope time. In the framework of the preparation of the SPIRou Input Catalog, the CoolSnap program aims at screening M dwarfs in the solar neighborhood against binarity, rapid rotation, activi… ▽ More Present and future high-precision radial-velocity spectrometers dedicated to the discovery of low-mass planets orbiting low-mass dwarfs need to focus on the best selected stars to make an efficient use of telescope time. In the framework of the preparation of the SPIRou Input Catalog, the CoolSnap program aims at screening M dwarfs in the solar neighborhood against binarity, rapid rotation, activity, ... To optimize the selection, the present paper describes the methods used to compute effective temperature, metallicity, projected rotation velocity of a large sample of 440 M dwarfs observed in the visible with the high-resolution spectro-polarimeter ESPaDOnS at CFHT. It also summarizes known and newly-discovered spectroscopic binaries, and stars known to belong to visual multiple systems. A calibration of the projected rotation velocity versus measured line widths for M dwarfs observed by the ESPaDOnS spectro-polarimeter is derived, and the resulting values are compared to equatorial rotation velocities deduced from rotation periods and radii. A comparison of the derived effective temperatures and metallicities with literature values is also conducted. Finally, the radial velocity uncertainty of each star in the sample is estimated, to narrow down the selection of stars to be included into the SPIRou Input Catalogue (SPIC). △ Less

Submitted 12 December, 2017; originally announced December 2017.

Comments: 34 pages, 12 figures, 13 tables, MNRAS in press December 11, 2017

Utilizing Domain Knowledge in End-to-End Audio Processing

Authors: Tycho Max Sylvester Tax, Jose Luis Diez Antich, Hendrik Purwins, Lars Maaløe

Abstract: End-to-end neural network based approaches to audio modelling are generally outperformed by models trained on high-level data representations. In this paper we present preliminary work that shows the feasibility of training the first layers of a deep convolutional neural network (CNN) model to learn the commonly-used log-scaled mel-spectrogram transformation. Secondly, we demonstrate that upon ini… ▽ More End-to-end neural network based approaches to audio modelling are generally outperformed by models trained on high-level data representations. In this paper we present preliminary work that shows the feasibility of training the first layers of a deep convolutional neural network (CNN) model to learn the commonly-used log-scaled mel-spectrogram transformation. Secondly, we demonstrate that upon initializing the first layers of an end-to-end CNN classifier with the learned transformation, convergence and performance on the ESC-50 environmental sound classification dataset are similar to a CNN-based model trained on the highly pre-processed log-scaled mel-spectrogram features. △ Less

Submitted 1 December, 2017; originally announced December 2017.

Comments: Accepted at the ML4Audio workshop at the NIPS 2017

BANYAN. X. Discovery of a wide, low-gravity L-type companion to a fast-rotating M3 dwarf

Authors: Marie-Eve Desrochers, Etienne Artigau, Jonathan Gagné, René Doyon, Lison Malo, Jacqueline K. Faherty, David Lafrenière

Abstract: We report the discovery of a substellar-mass co-moving companion to 2MASS J22501512+2325342, an M3 candidate member of the young (130--200 Myr) AB Doradus Moving Group (ABDMG). This L3 beta companion was discovered in a 2MASS search for companions at separations of 3--18" from a list of 2 812 stars suspected to be young (< 500 Myr) in the literature, and was confirmed with follow-up astrometry and… ▽ More We report the discovery of a substellar-mass co-moving companion to 2MASS J22501512+2325342, an M3 candidate member of the young (130--200 Myr) AB Doradus Moving Group (ABDMG). This L3 beta companion was discovered in a 2MASS search for companions at separations of 3--18" from a list of 2 812 stars suspected to be young (< 500 Myr) in the literature, and was confirmed with follow-up astrometry and spectroscopy. Evolutionary models yield a companion mass of 30 (-4,+11) Mj at the age of ABDMG. The 2MASS J22501512+2325342 AB system appears to be a spatial outlier to the bulk of ABDMG members, similarly to the young 2MASS J22362452+4751425 AB system. Future searches for young objects around these two systems would make it possible to determine whether they are part of a spatial extension of the known ABDMG distribution. △ Less

Submitted 29 November, 2017; originally announced November 2017.

Comments: 16 pages, 12 figures

Exploiting Nontrivial Connectivity for Automatic Speech Recognition

Authors: Marius Paraschiv, Lasse Borgholt, Tycho Max Sylvester Tax, Marco Singh, Lars Maaløe

Abstract: Nontrivial connectivity has allowed the training of very deep networks by addressing the problem of vanishing gradients and offering a more efficient method of reusing parameters. In this paper we make a comparison between residual networks, densely-connected networks and highway networks on an image classification task. Next, we show that these methodologies can easily be deployed into automatic… ▽ More Nontrivial connectivity has allowed the training of very deep networks by addressing the problem of vanishing gradients and offering a more efficient method of reusing parameters. In this paper we make a comparison between residual networks, densely-connected networks and highway networks on an image classification task. Next, we show that these methodologies can easily be deployed into automatic speech recognition and provide significant improvements to existing models. △ Less

Submitted 28 November, 2017; originally announced November 2017.

Comments: Accepted at the ML4Audio workshop at the NIPS 2017

PSYM-WIDE: a survey for large-separation planetary-mass companions to late spectral type members of young moving groups

Authors: Marie-Eve Naud, Étienne Artigau, René Doyon, Lison Malo, Jonathan Gagné, David Lafrenière, Christian Wolf, Eugene A. Magnier

Abstract: We present the results of a direct-imaging survey for very large separation ($>$100 au), companions around 95 nearby young K5-L5 stars and brown dwarfs. They are high-likelihood candidates or confirmed members of the young ($\lessapprox$150 Myr) $β$ Pictoris and AB Doradus moving groups (ABDMG) and the TW Hya, Tucana-Horologium, Columba, Carina, and Argus associations. Images in $i'$ and $z'$ filt… ▽ More We present the results of a direct-imaging survey for very large separation ($>$100 au), companions around 95 nearby young K5-L5 stars and brown dwarfs. They are high-likelihood candidates or confirmed members of the young ($\lessapprox$150 Myr) $β$ Pictoris and AB Doradus moving groups (ABDMG) and the TW Hya, Tucana-Horologium, Columba, Carina, and Argus associations. Images in $i'$ and $z'$ filters were obtained with the Gemini Multi-Object Spectrograph (GMOS) on Gemini South to search for companions down to an apparent magnitude of $z'\sim$22-24 at separations $\gtrapprox$20" from the targets and in the remainder of the wide 5.5' $\times$ 5.5' GMOS field of view. This allowed us to probe the most distant region where planetary-mass companions could be gravitationally bound to the targets. This region was left largely unstudied by past high-contrast imaging surveys, which probed much closer-in separations. This survey led to the discovery of a planetary-mass (9-13 $\,M_{\rm{Jup}}$) companion at 2000 au from the M3V star GU Psc, a highly probable member of ABDMG. No other substellar companions were identified. These results allowed us to constrain the frequency of distant planetary-mass companions (5-13 $\,M_{\rm{Jup}}$) to 0.84$_{-0.66}^{+6.73}$% (95% confidence) at semimajor axes between 500 and 5000 au around young K5-L5 stars and brown dwarfs. This is consistent with other studies suggesting that gravitationally bound planetary-mass companions at wide separations from low-mass stars are relatively rare. △ Less

Submitted 4 October, 2017; originally announced October 2017.

Comments: 29 pages, 9 figures, 5 tables

Journal ref: The Astronomical Journal, Volume 154, Issue 3, article id. 129, 20 pp. (2017 September 4)

A search for photometric variability in the young T3.5 planetary-mass companion GU Psc b

Authors: Marie-Eve Naud, Étienne Artigau, Jason F. Rowe, René Doyon, Lison Malo, Loïc Albert, Jonathan Gagné, Sandie Bouchard

Abstract: We present a photometric $J$-band variability study of GU Psc b, a T3.5 co-moving planetary-mass companion (9-13$M_{\rm{Jup}}$) to a young ($\sim$150 Myr) M3 member of the AB Doradus Moving Group. The large separation between GU Psc b and its host star (42") provides a rare opportunity to study the photometric variability of a planetary-mass companion. The study presented here is based on observat… ▽ More We present a photometric $J$-band variability study of GU Psc b, a T3.5 co-moving planetary-mass companion (9-13$M_{\rm{Jup}}$) to a young ($\sim$150 Myr) M3 member of the AB Doradus Moving Group. The large separation between GU Psc b and its host star (42") provides a rare opportunity to study the photometric variability of a planetary-mass companion. The study presented here is based on observations obtained from 2013 to 2014 over three nights with durations of 5-6 hr each with the WIRCam imager at Canada-France-Hawaii Telescope. Photometric variability with a peak-to-peak amplitude of $4\pm1$% at a timescale of $\sim$6 hr was marginally detected on 2014 October 11. No high-significance variability was detected on 2013 December 22 and 2014 October 10. The amplitude and timescale of the variability seen here, as well as its evolving nature, is comparable to what was observed for a variety of field T dwarfs and suggests that mechanisms invoked to explain brown dwarf variability may be applicable to low-gravity objects such as GU Psc b. Rotation-induced photometric variability due to the formation and dissipation of atmospheric features such as clouds is a plausible hypothesis for the tentative variation detected here. Additional photometric measurements, particularly on longer timescales, will be required to confirm and characterize the variability of GU Psc b, determine its periodicity and to potentially measure its rotation period. △ Less

Submitted 4 October, 2017; originally announced October 2017.

Comments: 17 pages, 12 figures

Journal ref: The Astronomical Journal, Volume 154, Issue 4, article id. 138, 13 pp. (2017 September 8)

SPIRou Input Catalog: Activity, Rotation and Magnetic Field of Cool Dwarfs

Authors: C. Moutou, E. M. Hebrard, J. Morin, L. Malo, P. Fouque, A. Torres-Rivas, E. Martioli, X. Delfosse, E. Artigau, R. Doyon

Abstract: Based on optical high-resolution spectra obtained with CFHT/ESPaDOnS, we present new measurements of activity and magnetic field proxies of 442 low-mass K5-M7 dwarfs. The objects were analysed as potential targets to search for planetary-mass companions with the new spectropolarimeter and high-precision velocimeter, SPIRou. We have analysed their high-resolution spectra in an homogeneous way: circ… ▽ More Based on optical high-resolution spectra obtained with CFHT/ESPaDOnS, we present new measurements of activity and magnetic field proxies of 442 low-mass K5-M7 dwarfs. The objects were analysed as potential targets to search for planetary-mass companions with the new spectropolarimeter and high-precision velocimeter, SPIRou. We have analysed their high-resolution spectra in an homogeneous way: circular polarisation, chromospheric features, and Zeeman broadening of the FeH infrared line. The complex relationship between these activity indicators is analysed: while no strong connection is found between the large-scale and small-scale magnetic fields, the latter relates with the non-thermal flux originating in the chromosphere. We then examine the relationship between various activity diagnostics and the optical radial-velocity jitter available in the literature, especially for planet host stars. We use this to derive for all stars an activity merit function (higher for quieter stars) with the goal of identifying the most favorable stars where the radial-velocity jitter is low enough for planet searches. We find that the main contributors to the RV jitter are the large-scale magnetic field and the chromospheric non-thermal emission. In addition, three stars (GJ 1289, GJ 793, and GJ 251) have been followed along their rotation using the spectropolarimetric mode, and we derive their magnetic topology. These very slow rotators are good representatives of future SPIRou targets. They are compared to other stars where the magnetic topology is also known. The poloidal component of the magnetic field is predominent in all three stars. △ Less

Submitted 5 September, 2017; originally announced September 2017.

Comments: to be published in MNRAS

The beta Pictoris association low-mass members: membership assessment, rotation period distribution, and dependence on multiplicity

Authors: S. Messina, A. C. Lanzafame, L. Malo, S. Desidera, A. Buccino, L. Zhang, S. Artemenko, M. Millward, F. -J. Hambsch

Abstract: Low-mass members of young stellar associations exhibit a wide spread of rotation periods. Such a spread originates from distributions of masses and initial rotation periods. However, multiplicity can also play a significant role. We investigate the role played by physical companions in shortening the primordial disc lifetime. We have compiled the most extensive list of low-mass members of the youn… ▽ More Low-mass members of young stellar associations exhibit a wide spread of rotation periods. Such a spread originates from distributions of masses and initial rotation periods. However, multiplicity can also play a significant role. We investigate the role played by physical companions in shortening the primordial disc lifetime. We have compiled the most extensive list of low-mass members of the young 25-Myr beta Pictoris association. We have measured the rotation periods of about all members and used updated UVWXYZ components to assess their membership. We built the rotation period distribution distinguishing between bona fide members and candidate members and according to their multiplicity status. We found that single stars and components of multiple systems in wide orbits (>80 AU) have rotation periods that exhibit a well defined sequence arising from mass distribution. All components of multiple systems in close orbits (<80 AU) have rotation periods significantly shorter than their equal-mass single counterparts. A comparison with the younger 13 Myr h Per cluster and with the older 40-Myr open clusters/stellar associations NGC2547, IC2391, Argus, and IC2602 and the 130-Myr Pleiades shows that whereas the evolution of F-G stars is well reproduced by angular momentum evolution models, this is not the case for the slow K and early-M stars. Finally, we found that the amplitude of their light curves is correlated neither with rotation nor with mass. Once single stars and wide components of multiple systems are separated from close components of multiple systems, the rotation period distributions exhibit a well defined dependence on mass that allows to make a meaningful comparison with similar distributions of either younger or older associations/clusters. Such cleaned distributions allow to use the stellar rotation period as age indicator, meaningfully for F and G type stars. △ Less

Submitted 6 July, 2017; originally announced July 2017.

Comments: 20 pages, 9 figures, 4 tables; accepted by Astronomy and Astrophysics

Journal ref: A&A 607, A3 (2017)

Semi-Supervised Generation with Cluster-aware Generative Models

Authors: Lars Maaløe, Marco Fraccaro, Ole Winther

Abstract: Deep generative models trained with large amounts of unlabelled data have proven to be powerful within the domain of unsupervised learning. Many real life data sets contain a small amount of labelled data points, that are typically disregarded when training generative models. We propose the Cluster-aware Generative Model, that uses unlabelled information to infer a latent representation that model… ▽ More Deep generative models trained with large amounts of unlabelled data have proven to be powerful within the domain of unsupervised learning. Many real life data sets contain a small amount of labelled data points, that are typically disregarded when training generative models. We propose the Cluster-aware Generative Model, that uses unlabelled information to infer a latent representation that models the natural clustering of the data, and additional labelled data points to refine this clustering. The generative performances of the model significantly improve when labelled information is exploited, obtaining a log-likelihood of -79.38 nats on permutation invariant MNIST, while also achieving competitive semi-supervised classification accuracies. The model can also be trained fully unsupervised, and still improve the log-likelihood performance with respect to related methods. △ Less

Submitted 3 April, 2017; originally announced April 2017.

CFBDSIR 2149-0403: young isolated planetary-mass object or high-metallicity low-mass brown dwarf??

Authors: P. Delorme, T. Dupuy, J. Gagné, C. Reylé, T. Forveille, Michael C. Liu, E. Artigau, L. Albert, X. Delfosse, F. Allard, D. Homeier, L. Malo, C. Morley, M. E. Naud, M. Bonnefoy

Abstract: We conducted a multi-wavelength, multi-instrument observational characterisation of the candidate free-floating planet CFBDSIR~J214947.2-040308.9, a late T-dwarf with possible low-gravity features, in order to constrain its physical properties. We analyzed 9 hours of X-Shooter spectroscopy with signal detectable from 0.8--2.3$μ$m, as well as additional photometry in the mid-infrared using the Spit… ▽ More We conducted a multi-wavelength, multi-instrument observational characterisation of the candidate free-floating planet CFBDSIR~J214947.2-040308.9, a late T-dwarf with possible low-gravity features, in order to constrain its physical properties. We analyzed 9 hours of X-Shooter spectroscopy with signal detectable from 0.8--2.3$μ$m, as well as additional photometry in the mid-infrared using the Spitzer Space Telescope. Combined with a VLT/HAWK-I astrometric parallax, this enabled a full characterisation of the absolute flux from the visible to 5$μ$m, encompassing more than 90\% of the expected energy emitted by such a cool late T-type object. Our analysis of the spectrum also provided the radial velocity and therefore the determination of its full 3-D kinematics. While our new spectrum confirms the low gravity and/or high metallicity of CFBDSIR2149, the parallax and kinematics safely rule out membership to any known young moving group, including AB~Doradus. We use the equivalent width of the KI doublet at 1.25$μ$m as a promising tool to discriminate the effects of low-gravity from the effects of high-metallicity on the emission spectra of cool atmospheres. In the case of CFBDSIR2149, the observed KI doublet clearly favours the low-gravity solution. CFBDSIR2149 is therefore a peculiar late-T dwarf that is probably a young, planetary-mass object (2--13Mjup, $<$500Myr) possibly similar to the exoplanet 51Erib, or perhaps a 2--40Mjup brown dwarf with super-solar metallicity. △ Less

Submitted 2 March, 2017; originally announced March 2017.

Comments: Accepted in A&A

Journal ref: A&A 602, A82 (2017)

Follow-up and characterization of the TESS exoplanets with SOPHIE, SPIRou, and JWST

Authors: N. Crouzet, X. Bonfils, X. Delfosse, I. Boisse, G. Hébrard, T. Forveille, J. -F. Donati, F. Bouchy, C. Moutou, R. Doyon, E. Artigau, L. Albert, L. Malo, A. Lecavelier des Etangs, A. Santerne

Abstract: The NASA TESS mission will deliver hundreds of transiting exoplanet candidates orbiting bright stars. The spectrometers SOPHIE at OHP and SPIRou at CFHT will be ideal to obtain radial velocities of these candidates, confirm their nature, and derive the planets' masses. These measurements will be crucial to deliver the best targets for atmospheric characterization with JWST. Here, we calculate the… ▽ More The NASA TESS mission will deliver hundreds of transiting exoplanet candidates orbiting bright stars. The spectrometers SOPHIE at OHP and SPIRou at CFHT will be ideal to obtain radial velocities of these candidates, confirm their nature, and derive the planets' masses. These measurements will be crucial to deliver the best targets for atmospheric characterization with JWST. Here, we calculate the required observing time with SOPHIE, SPIRou, and JWST for each of the TESS targets in order to prepare follow-up observations. To infer their potential for JWST, we restrict the calculations to the case of transmission spectroscopy with NIRISS. The radial velocity follow-up of the giant planets ($R_p > 4 \rm \, R_E$) could be achieved with SOPHIE, with a median observing time of 3.47 hours per target, and a total observing time of 305 hours that includes the 80% most favorable cases. Several small planets ($R_p < 4 \rm \, R_E$) could also be confirmed, but most of them would require an unrealistic time investment. On the other hand, SPIRou is ideally suited to the follow-up of the small planets, with a median observing time of 2.65 hours per target, and a median observing time of 4.70 hours for the terrestrial planets in the habitable zone ($R_p < 2 \rm \, R_E$, $S < 2 \rm \, S_E$). With JWST, the 10% most favorable small planets (184 planets) have a median observing time of 16.2 hours, and the 10% most favorable habitable zone terrestrial planets (7 planets) have a median observing time of 59.7 hours. Overall, this study will help define a follow-up strategy and prepare observation programs with SOPHIE and SPIRou before the first planet candidates are delivered by TESS. △ Less

Submitted 12 January, 2017; originally announced January 2017.

Comments: 5 pages, 3 figures, Proceedings of the Société Française d'Astronomie et d'Astrophysique (SF2A) 2016

A hot Jupiter around the very active weak-line T Tauri star TAP 26

Authors: L. Yu, J. -F. Donati, E. M. Hébrard, C. Moutou, L. Malo, K. Grankin, G. Hussain, A. Collier Cameron, A. A. Vidotto, C. Baruteau, S. H. P. Alencar, J. Bouvier, P. Petit, M. Takami, G. Herczeg, S. G. Gregory, M. Jardine, J. Morin, F. Ménard, the MaTYSSE collaboration

Abstract: We report the results of an extended spectropolarimetric and photometric monitoring of the weak-line T Tauri star TAP 26, carried out within the MaTYSSE programme with the ESPaDOnS spectropolarimeter at the 3.6 m Canada-France-Hawaii Telescope. Applying Zeeman-Doppler Imaging to our observations, concentrating in 2015 November and 2016 January and spanning 72 d in total, 16 d in 2015 November and… ▽ More We report the results of an extended spectropolarimetric and photometric monitoring of the weak-line T Tauri star TAP 26, carried out within the MaTYSSE programme with the ESPaDOnS spectropolarimeter at the 3.6 m Canada-France-Hawaii Telescope. Applying Zeeman-Doppler Imaging to our observations, concentrating in 2015 November and 2016 January and spanning 72 d in total, 16 d in 2015 November and 13 d in 2016 January, we reconstruct surface brightness and magnetic field maps for both epochs and demonstrate that both distributions exhibit temporal evolution not explained by differential rotation alone. We report the detection of a hot Jupiter (hJ) around TAP 26 using three different methods, two using Zeeman-Doppler Imaging (ZDI) and one Gaussian-Process Regression (GPR), with a false-alarm probability smaller than 6.10^-4. However, as a result of the aliasing related to the observing window, the orbital period cannot be uniquely determined; the orbital period with highest likelihood is 10.79 +/- 0.14 d followed by 8.99 +/- 0.09 d. Assuming the most likely period, and that the planet orbits in the stellar equatorial plane, we obtain that the planet has a minimum mass M.sin(i) of 1.66 +/- 0.31 M_Jup and orbits at 0.0968 +/- 0.0032 au from its host star. This new detection suggests that disc type II migration is efficient at generating newborn hJs, and that hJs may be more frequent around young T Tauri stars than around mature stars (or that the MaTYSSE sample is biased towards hJ-hosting stars). △ Less

Submitted 5 January, 2017; originally announced January 2017.

Comments: 25 pages, 3 appendices, 27 figures, 10 tables

Gemini/GRACES Spectroscopy of Stars in Triangulum II

Authors: K. A. Venn, E. Starkenburg, L. Malo, N. Martin, B. P. M. Laevens

Abstract: The chemical abundance ratios and radial velocities for two stars in the recently discovered Triangulum II faint dwarf galaxy have been determined from high resolution, medium signal-to-noise ratio spectra from the Gemini-GRACES facility. These stars have stellar parameters and metallicities similar to those derived from their photometry and medium-resolution Ca II triplet spectra, and supports th… ▽ More The chemical abundance ratios and radial velocities for two stars in the recently discovered Triangulum II faint dwarf galaxy have been determined from high resolution, medium signal-to-noise ratio spectra from the Gemini-GRACES facility. These stars have stellar parameters and metallicities similar to those derived from their photometry and medium-resolution Ca II triplet spectra, and supports that Triangulum II has a metallicity spread consistent with chemical evolution in a dwarf galaxy. The elemental abundances show that both stars have typical calcium abundances and barium upper limits for their metallicities, but low magnesium and sodium. This chemical composition resembles some stars in dwarf galaxies, attributed to inhomogeneous mixing in a low star formation environment, and/or yields from only a few supernova events. One of our targets (Star40) has an enhancement in potassium, and resembles some stars in the unusual outer halo star cluster, NGC 2419. Our other target (Star46) appears to be a binary based on a change in its radial velocity (Delta v(rad) = 24.5 +/- 2.1 km/s). This is consistent with variations found in binary stars in other dwarf galaxies. While this serves as a reminder of the high binary fraction in these ultra faint dwarf galaxies, this particular object has had little impact on the previous determination of the velocity dispersion in Triangulum II. △ Less

Submitted 14 December, 2016; originally announced December 2016.

Comments: 18 pages, 5 tables, 7 figures: Accepted, MNRAS, 2016 December 6

The beta Pictoris association: Catalog of photometric rotational periods of low-mass members and candidate members

Authors: S. Messina, M. Millward, A. Buccino, L. Zhang, B. J. Medhi, E. Jofre', R. Petrucci, Q. Pi, F. -J. Hambsch, P. Kehusmaa, C. Harlingten, S. Artemenko, I. Curtis, V. -P. Hentunen, L. Malo, P. Mauas, B. Monard, M. Muro Serrano, R. Naves, R. Santallo, A. Savuskin, T. G. Tan

Abstract: We intended to compile the most complete catalog of bona fide members and candidate members of the beta Pictoris association, and to measure their rotation periods and basic properties from our own observations, public archives, and exploring the literature. We carried out a multi-observatories campaign to get our own photometric time series and collected all archived public photometric data time… ▽ More We intended to compile the most complete catalog of bona fide members and candidate members of the beta Pictoris association, and to measure their rotation periods and basic properties from our own observations, public archives, and exploring the literature. We carried out a multi-observatories campaign to get our own photometric time series and collected all archived public photometric data time series for the stars in our catalog. Each time series was analyzed with the Lomb-Scargle and CLEAN periodograms to search for the stellar rotation periods. We complemented the measured rotational properties with detailed information on multiplicity, membership, and projected rotational velocity available in the literature and discussed star by star. We measured the rotation periods of 112 out of 117 among bona fide members and candidate members of the beta Pictoris association and, whenever possible, we also measured the luminosity, radius, and inclination of the stellar rotation axis. This represents to date the largest catalog of rotation periods of any young loose stellar association. We provided an extensive catalog of rotation periods together with other relevant basic properties useful to explore a number of open issues, such as the causes of spread of rotation periods among coeval stars, evolution of angular momentum, and lithium-rotation connection. △ Less

Submitted 14 December, 2016; originally announced December 2016.

Comments: Forthcoming article, Received: 20 June 2016 / Accepted: 09 September 2016; 40 pages, 2 figures. The online figures A1-A73 are available at CDS

Journal ref: Astronomy and Astrophysics, 2016

BANYAN. IX. The Initial Mass Function and Planetary-Mass Object Space Density of the TW Hya Association

Authors: Jonathan Gagné, Jacqueline K. Faherty, Eric E. Mamajek, Lison Malo, René Doyon, Joseph C. Filippazzo, Alycia J. Weinberger, Jessica K. Donaldson, Sébastien Lépine, David Lafrenière, Étienne Artigau, Adam J. Burgasser, Dagny Looper, Anne Boucher, Yuri Beletsky, Sara Camnasio, Charles Brunette, Geneviève Arboit

Abstract: A determination of the initial mass function (IMF) of the current, incomplete census of the 10 Myr-old TW Hya association (TWA) is presented. This census is built from a literature compilation supplemented with new spectra and 17 new radial velocities, as well as a re-analysis of Hipparcos data that confirmed HR 4334 (A2Vn) as a member. Though the dominant uncertainty in the IMF remains census inc… ▽ More A determination of the initial mass function (IMF) of the current, incomplete census of the 10 Myr-old TW Hya association (TWA) is presented. This census is built from a literature compilation supplemented with new spectra and 17 new radial velocities, as well as a re-analysis of Hipparcos data that confirmed HR 4334 (A2Vn) as a member. Though the dominant uncertainty in the IMF remains census incompleteness, a detailed statistical treatment is carried out to make the IMF determination independent of binning, while accounting for small number statistics. The currently known high-likelihood members are fitted by a log-normal distribution with a central mass of $0.21^{+0.11}_{-0.06}$ $M_{\odot}$ and a characteristic width of $0.8^{+0.2}_{-0.1}$ dex in the 12 $M_{\rm Jup}$-2 $M_{\odot}$ range, whereas a Salpeter power law with $α= 2.2^{+1.1}_{-0.5}$ best describes the IMF slope in the $0.1-2$ $M_{\odot}$ range. This characteristic width is higher than other young associations, which may be due to incompleteness in the current census of low-mass TWA stars. A tentative overpopulation of isolated planetary-mass members similar to 2MASS J11472421-2040204 and 2MASS J11193254-1137466 is identified: this indicates that there might be as many as $10^{+13}_{-5}$ similar members of TWA with hot-start model-dependent masses estimated at $\sim$ 5-7 $M_{\rm Jup}$, most of which would be too faint to be detected in 2MASS. Our new radial velocity measurements corroborate the membership of 2MASS J11472421-2040204, and secure TWA 28 (M8.5$γ$), TWA 29 (M9.5$γ$) and TWA 33 (M4.5e) as members. The discovery of 2MASS J09553336-0208403, a young L7-type interloper unrelated to TWA, is also presented. △ Less

Submitted 16 December, 2016; v1 submitted 8 December, 2016; originally announced December 2016.

Comments: 62 pages, 27 figures, 14 tables. Accepted for publication in Astrophysical Journal Supplement Series. Added Marsh et al. (2010) data in Figure 15, and the data behind Figure 15 in article material

The hot Jupiter of the magnetically-active weak-line T Tauri star V830 Tau

Authors: J. -F. Donati, L. Yu, C. Moutou, A. C. Cameron, L. Malo, K. Grankin, E. Hébrard, G. A. J. Hussain, A. A. Vidotto, S. H. P. Alencar, R. D. Haywood, J. Bouvier, P. Petit, M. Takami, G. J. Herczeg, S. G. Gregory, M. M. Jardine, J. Morin, the MaTYSSE collaboration

Abstract: We report results of an extended spectropolarimetric and photometric monitoring of the weak-line T Tauri star V830 Tau and its recently-detected newborn close-in giant planet. Our observations, carried out within the MaTYSSE programme, were spread over 91d, and involved the ESPaDOnS and Narval spectropolarimeters linked to the 3.6m Canada-France-Hawaii, the 2m Bernard Lyot and the 8-m Gemini-North… ▽ More We report results of an extended spectropolarimetric and photometric monitoring of the weak-line T Tauri star V830 Tau and its recently-detected newborn close-in giant planet. Our observations, carried out within the MaTYSSE programme, were spread over 91d, and involved the ESPaDOnS and Narval spectropolarimeters linked to the 3.6m Canada-France-Hawaii, the 2m Bernard Lyot and the 8-m Gemini-North Telescopes. Using Zeeman-Doppler Imaging, we characterize the surface brightness distributions, magnetic topologies and surface differential rotation of V830 Tau at the time of our observations, and demonstrate that both distributions evolve with time beyond what is expected from differential rotation. We also report that near the end of our observations, V830 Tau triggered one major flare and two weaker precursors, showing up as enhanced red-shifted emission in multiple spectral activity proxies. With 3 different filtering techniques, we model the radial velocity (RV) activity jitter (of semi-amplitude 1.2km/s) that V830 Tau generates, successfully retrieve the 68m/s RV planet signal hiding behind the jitter, further confirm the existence of V830 Tau b and better characterize its orbital parameters. We find that the method based on Gaussian-process regression performs best thanks to its higher ability at modelling not only the activity jitter, but also its temporal evolution over the course of our observations, and succeeds at reproducing our RV data down to a rms precision of 35m/s. Our result provides new observational constraints on scenarios of star / planet formation and demonstrates the scientific potential of large-scale searches for close-in giant planets around T Tauri stars. △ Less

Submitted 7 November, 2016; originally announced November 2016.

Comments: MNRAS, in press (17 pages, 15 figures, 5 tables)

Planets Around Low-Mass Stars (PALMS). VI. Discovery of a Remarkably Red Planetary-Mass Companion to the AB Dor Moving Group Candidate 2MASS J22362452+4751425

Authors: Brendan Bowler, Michael Liu, Dimitri Mawet, Henry Ngo, Lison Malo, Gregory Mace, Jacob McLane, Jessica Lu, Isaiah Tristan, Sasha Hinkley, Lynne Hillenbrand, Evgenya Shkolnik, Bjorn Benneke, William Best

Abstract: We report the discovery of an extremely red planetary-mass companion to 2MASS J22362452+4751425, a $\approx$0.6 $M_{\odot}$ late-K dwarf likely belonging to the $\sim$120 Myr AB Doradus moving group. 2M2236+4751 b was identified in multi-epoch NIRC2 adaptive optics imaging at Keck Observatory at a separation of 3.7$"$, or 230 $\pm$ 20 AU in projection at the kinematic distance of 63 $\pm$ 5 pc to… ▽ More We report the discovery of an extremely red planetary-mass companion to 2MASS J22362452+4751425, a $\approx$0.6 $M_{\odot}$ late-K dwarf likely belonging to the $\sim$120 Myr AB Doradus moving group. 2M2236+4751 b was identified in multi-epoch NIRC2 adaptive optics imaging at Keck Observatory at a separation of 3.7$"$, or 230 $\pm$ 20 AU in projection at the kinematic distance of 63 $\pm$ 5 pc to its host star. Assuming membership in the AB Dor group, as suggested from its kinematics, the inferred mass of 2M2236+4751 b is 11-14 $M_\mathrm{Jup}$. Follow-up Keck/OSIRIS $K$-band spectroscopy of the companion reveals strong CO absorption similar to other faint red L dwarfs and lacks signs of methane absorption despite having an effective temperature of $\approx$900-1200 K. With a ($J$-$K$)$_\mathrm{MKO}$ color of 2.69 $\pm$ 0.12 mag, the near-infrared slope of 2M2236+4751 b is redder than all of the HR 8799 planets and instead resembles the $\approx$23 Myr isolated planetary-mass object PSO J318.5-22, implying that similarly thick photospheric clouds can persist in the atmospheres of giant planets at ages beyond 100 Myr. In near-infrared color-magnitude diagrams, 2M2236+4751 b is located at the tip of the red L dwarf sequence and appears to define the "elbow" of the AB Dor substellar isochrone separating low-gravity L dwarfs from the cooler young T dwarf track. 2M2236+4751 b is the reddest substellar companion to a star and will be a valuable benchmark to study the shared atmospheric properties of young low-mass brown dwarfs and extrasolar giant planets. △ Less

Submitted 1 November, 2016; originally announced November 2016.

Comments: Accepted for publication in The Astronomical Journal