Showing 1–27 of 27 results for author: Lourie, N

Contextual Counting: A Mechanistic Study of Transformers on a Quantitative Task

Authors: Siavash Golkar, Alberto Bietti, Mariel Pettee, Michael Eickenberg, Miles Cranmer, Keiya Hirashima, Geraud Krawezik, Nicholas Lourie, Michael McCabe, Rudy Morel, Ruben Ohana, Liam Holden Parker, Bruno Régaldo-Saint Blancard, Kyunghyun Cho, Shirley Ho

Abstract: Transformers have revolutionized machine learning across diverse domains, yet understanding their behavior remains crucial, particularly in high-stakes applications. This paper introduces the contextual counting task, a novel toy problem aimed at enhancing our understanding of Transformers in quantitative and scientific contexts. This task requires precise localization and computation within datas… ▽ More Transformers have revolutionized machine learning across diverse domains, yet understanding their behavior remains crucial, particularly in high-stakes applications. This paper introduces the contextual counting task, a novel toy problem aimed at enhancing our understanding of Transformers in quantitative and scientific contexts. This task requires precise localization and computation within datasets, akin to object detection or region-based scientific analysis. We present theoretical and empirical analysis using both causal and non-causal Transformer architectures, investigating the influence of various positional encodings on performance and interpretability. In particular, we find that causal attention is much better suited for the task, and that no positional embeddings lead to the best accuracy, though rotary embeddings are competitive and easier to train. We also show that out of distribution performance is tightly linked to which tokens it uses as a bias term. △ Less

Submitted 30 May, 2024; originally announced June 2024.

Show Your Work with Confidence: Confidence Bands for Tuning Curves

Authors: Nicholas Lourie, Kyunghyun Cho, He He

Abstract: The choice of hyperparameters greatly impacts performance in natural language processing. Often, it is hard to tell if a method is better than another or just better tuned. Tuning curves fix this ambiguity by accounting for tuning effort. Specifically, they plot validation performance as a function of the number of hyperparameter choices tried so far. While several estimators exist for these curve… ▽ More The choice of hyperparameters greatly impacts performance in natural language processing. Often, it is hard to tell if a method is better than another or just better tuned. Tuning curves fix this ambiguity by accounting for tuning effort. Specifically, they plot validation performance as a function of the number of hyperparameter choices tried so far. While several estimators exist for these curves, it is common to use point estimates, which we show fail silently and give contradictory results when given too little data. Beyond point estimates, confidence bands are necessary to rigorously establish the relationship between different approaches. We present the first method to construct valid confidence bands for tuning curves. The bands are exact, simultaneous, and distribution-free, thus they provide a robust basis for comparing methods. Empirical analysis shows that while bootstrap confidence bands, which serve as a baseline, fail to approximate their target confidence, ours achieve it exactly. We validate our design with ablations, analyze the effect of sample size, and provide guidance on comparing models with our method. To promote confident comparisons in future work, we release opda: an easy-to-use library that you can install with pip. https://github.com/nicholaslourie/opda △ Less

Submitted 8 April, 2024; v1 submitted 15 November, 2023; originally announced November 2023.

Comments: Accepted to NAACL 2024. 18 pages, 20 figures

A luminous dust-obscured Tidal Disruption Event candidate in a star forming galaxy at 42 Mpc

Authors: Christos Panagiotou, Kishalay De, Megan Masterson, Erin Kara, Michael Calzadilla, Anna-Christina Eilers, Danielle Frostig, Viraj Karambelkar, Mansi Kasliwal, Nathan Lourie, Aaron M. Meisner, Robert A. Simcoe, Robert Stein, Jeffry Zolkower

Abstract: While the vast majority of Tidal Disruption Events (TDEs) has been identified by wide-field sky surveys in the optical and X-ray bands, recent studies indicate that a considerable fraction of TDEs may be dust obscured, and thus preferentially detected in the infrared (IR) wavebands. In this Letter, we present the discovery of a luminous mid-IR nuclear flare (termed WTP 14adbjsh) identified in a sy… ▽ More While the vast majority of Tidal Disruption Events (TDEs) has been identified by wide-field sky surveys in the optical and X-ray bands, recent studies indicate that a considerable fraction of TDEs may be dust obscured, and thus preferentially detected in the infrared (IR) wavebands. In this Letter, we present the discovery of a luminous mid-IR nuclear flare (termed WTP 14adbjsh) identified in a systematic transient search of archival images from the NEOWISE mid-IR survey. The source reached a peak luminosity of $L \simeq 10^{43} \text{erg s}^{-1}$ at 4.6 $μ$m in 2015, before fading in the IR with a TDE-like $F \propto t^{-5/3}$ decline, radiating a total of more than $ 3\times 10^{51}$ erg in the last 7 years. The transient event took place in the nearby galaxy NGC 7392, at a distance of around 42 Mpc; yet, no optical or X-ray flare is detected. We interpret the transient as the nearest TDE candidate detected in the last decade, which was missed at other wavelengths due to dust obscuration, hinting at the existence of TDEs that have been historically overlooked. Unlike most previously detected TDEs, the transient was discovered in a star forming galaxy, corroborating earlier suggestions that dust obscuration suppresses significantly the detection of TDEs in these environments. Our results demonstrate that the study of IR-detected TDEs is critical in order to obtain a complete understanding of the physics of TDEs, and to conclude whether TDEs occur preferentially in a particular class of galaxies. △ Less

Submitted 5 March, 2023; originally announced March 2023.

Comments: 13 pages, 4 figures, accepted for publication by ApJL

An Infrared Search for Kilonovae with the WINTER Telescope. I. Binary Neutron Star Mergers

Authors: Danielle Frostig, Sylvia Biscoveanu, Geoffrey Mo, Viraj Karambelkar, Tito Dal Canton, Hsin-Yu Chen, Mansi Kasliwal, Erik Katsavounidis, Nathan P. Lourie, Robert A. Simcoe, Salvatore Vitale

Abstract: The Wide-Field Infrared Transient Explorer (WINTER) is a new 1 $\text{deg}^2$ seeing-limited time-domain survey instrument designed for dedicated near-infrared follow-up of kilonovae from binary neutron star (BNS) and neutron star-black hole mergers. WINTER will observe in the near-infrared Y, J, and short-H bands (0.9-1.7 microns, to $\text{J}_{AB}=21$ magnitudes) on a dedicated 1-meter telescope… ▽ More The Wide-Field Infrared Transient Explorer (WINTER) is a new 1 $\text{deg}^2$ seeing-limited time-domain survey instrument designed for dedicated near-infrared follow-up of kilonovae from binary neutron star (BNS) and neutron star-black hole mergers. WINTER will observe in the near-infrared Y, J, and short-H bands (0.9-1.7 microns, to $\text{J}_{AB}=21$ magnitudes) on a dedicated 1-meter telescope at Palomar Observatory. To date, most prompt kilonova follow-up has been in optical wavelengths; however, near-infrared emission fades more slowly and depends less on geometry and viewing angle than optical emission. We present an end-to-end simulation of a follow-up campaign during the fourth observing run (O4) of the LIGO, Virgo, and KAGRA interferometers, including simulating 625 BNS mergers, their detection in gravitational waves, low-latency and full parameter estimation skymaps, and a suite of kilonova lightcurves from two different model grids. We predict up to five new kilonovae independently discovered by WINTER during O4, given a realistic BNS merger rate. Using a larger grid of kilonova parameters, we find that kilonova emission is $\approx$2 times longer-lived and red kilonovae are detected $\approx$1.5 times further in the infrared than in the optical. For 90% localization areas smaller than 150 (450) $\rm{deg}^{2}$, WINTER will be sensitive to more than 10% of the kilonova model grid out to 350 (200) Mpc. We develop a generalized toolkit to create an optimal BNS follow-up strategy with any electromagnetic telescope and present WINTER's observing strategy with this framework. This toolkit, all simulated gravitational-wave events, and skymaps are made available for use by the community. △ Less

Submitted 28 February, 2022; v1 submitted 4 October, 2021; originally announced October 2021.

Comments: 19 pages, 11 figures, published in ApJ

Report number: LIGO Document P-2100340

Journal ref: The Astrophysical Journal, Volume 926, Number 2, Page 152, 2022

Design Requirements for the Wide-field Infrared TransientExplorer (WINTER)

Authors: Danielle Frostig, John W. Baker, Joshua Brown, Richard S. Burruss, Kristin Clark, Gábor Fűrész, Nicolae Ganciu, Erik Hinrichsen, Viraj R. Karambelkar, Mansi M. Kasliwal, Nathan P. Lourie, Andrew Malonis, Robert A. Simcoe, Jeffry Zolkower

Abstract: The Wide-field Infrared Transient Explorer (WINTER) is a 1x1 degree infrared survey telescope under development at MIT and Caltech, and slated for commissioning at Palomar Observatory in 2021. WINTER is a seeing-limited infrared time-domain survey and has two main science goals: (1) the discovery of IR kilonovae and r-process materials from binary neutron star mergers and (2) the study of general… ▽ More The Wide-field Infrared Transient Explorer (WINTER) is a 1x1 degree infrared survey telescope under development at MIT and Caltech, and slated for commissioning at Palomar Observatory in 2021. WINTER is a seeing-limited infrared time-domain survey and has two main science goals: (1) the discovery of IR kilonovae and r-process materials from binary neutron star mergers and (2) the study of general IR transients, including supernovae, tidal disruption events, and transiting exoplanets around low mass stars. We plan to meet these science goals with technologies that are relatively new to astrophysical research: hybridized InGaAs sensors as an alternative to traditional, but expensive, HgCdTe arrays and an IR-optimized 1-meter COTS telescope. To mitigate risk, optimize development efforts, and ensure that WINTER meets its science objectives, we use model-based systems engineering (MBSE) techniques commonly featured in aerospace engineering projects. Even as ground-based instrumentation projects grow in complexity, they do not often have the budget for a full-time systems engineer. We present one example of systems engineering for the ground-based WINTER project, featuring software tools that allow students or staff to learn the fundamentals of MBSE and capture the results in a formalized software interface. We focus on the top-level science requirements with a detailed example of how the goal of detecting kilonovae flows down to WINTER's optical design. In particular, we discuss new methods for tolerance simulations, eliminating stray light, and maximizing image quality of a fly's-eye design that slices the telescope's focus onto 6 non-buttable, IR detectors. We also include a discussion of safety constraints for a robotic telescope. △ Less

Submitted 3 May, 2021; originally announced May 2021.

Comments: Published in SPIE Astronomical Telescopes + Instrumentation 2020, Ground-based and Airborne Instrumentation for Astronomy VIII. 12 pages, 9 figures

Journal ref: Proc. SPIE 11447, Ground-based and Airborne Instrumentation for Astronomy VIII, 2020

UNICORN on RAINBOW: A Universal Commonsense Reasoning Model on a New Multitask Benchmark

Authors: Nicholas Lourie, Ronan Le Bras, Chandra Bhagavatula, Ye** Choi

Abstract: Commonsense AI has long been seen as a near impossible goal -- until recently. Now, research interest has sharply increased with an influx of new benchmarks and models. We propose two new ways to evaluate commonsense models, emphasizing their generality on new tasks and building on diverse, recently introduced benchmarks. First, we propose a new multitask benchmark, RAINBOW, to promote research… ▽ More Commonsense AI has long been seen as a near impossible goal -- until recently. Now, research interest has sharply increased with an influx of new benchmarks and models. We propose two new ways to evaluate commonsense models, emphasizing their generality on new tasks and building on diverse, recently introduced benchmarks. First, we propose a new multitask benchmark, RAINBOW, to promote research on commonsense models that generalize well over multiple tasks and datasets. Second, we propose a novel evaluation, the cost equivalent curve, that sheds new insight on how the choice of source datasets, pretrained language models, and transfer learning methods impacts performance and data efficiency. We perform extensive experiments -- over 200 experiments encompassing 4800 models -- and report multiple valuable and sometimes surprising findings, e.g., that transfer almost always leads to better or equivalent performance if following a particular recipe, that QA-based commonsense datasets transfer well with each other, while commonsense knowledge graphs do not, and that perhaps counter-intuitively, larger models benefit more from transfer than smaller ones. Last but not least, we introduce a new universal commonsense reasoning model, UNICORN, that establishes new state-of-the-art performance across 8 popular commonsense benchmarks, aNLI (87.3%), CosmosQA (91.8%), HellaSWAG (93.9%), PIQA (90.1%), SocialIQa (83.2%), WinoGrande (86.6%), CycIC (94.0%) and CommonsenseQA (79.3%). △ Less

Submitted 24 March, 2021; originally announced March 2021.

Comments: 27 pages, 19 figures, 34 tables. Accepted to AAAI 2021. For associated code and data see https://github.com/allenai/rainbow

The wide-field infrared transient explorer (WINTER)

Authors: Nathan P. Lourie, John W. Baker, Richard S. Burruss, Mark Egan, Gábor Fűrész, Danielle Frostig, Allan A. Garcia-Zych, Nicolae Ganciu, Kari Haworth, Erik Hinrichsen, Mansi M. Kasliwal, Viraj R. Karambelkar, Andrew Malonis, Robert A. Simcoe, Jeffry Zolkower

Abstract: The Wide-Field Infrared Transient Explorer (WINTER) is a new infrared time-domain survey instrument which will be deployed on a dedicated 1 meter robotic telescope at Palomar Observatory. WINTER will perform a seeing-limited time domain survey of the infrared (IR) sky, with a particular emphasis on identifying r-process material in binary neutron star (BNS) merger remnants detected by LIGO. We des… ▽ More The Wide-Field Infrared Transient Explorer (WINTER) is a new infrared time-domain survey instrument which will be deployed on a dedicated 1 meter robotic telescope at Palomar Observatory. WINTER will perform a seeing-limited time domain survey of the infrared (IR) sky, with a particular emphasis on identifying r-process material in binary neutron star (BNS) merger remnants detected by LIGO. We describe the scientific goals and survey design of the WINTER instrument. With a dedicated trigger and the ability to map the full LIGO O4 positional error contour in the IR to a distance of 190 Mpc within four hours, WINTER will be a powerful kilonova discovery engine and tool for multi-messenger astrophysics investigations. In addition to follow-up observations of merging binaries, WINTER will facilitate a wide range of time-domain astronomical observations, all the while building up a deep coadded image of the static infrared sky suitable for survey science. WINTER's custom camera features six commercial large-format Indium Gallium Arsenide (InGaAs) sensors and a tiled optical system which covers a $>$1-square-degree field of view with 90% fill factor. The instrument observes in Y, J and a short-H (Hs) band tuned to the long-wave cutoff of the InGaAs sensors, covering a wavelength range from 0.9 - 1.7 microns. We present the design of the WINTER instrument and current progress towards final integration at Palomar Observatory and commissioning planned for mid-2021. △ Less

Submitted 1 February, 2021; originally announced February 2021.

Comments: Published in SPIE Astronomical Telescopes + Instrumentation 2020, Ground-based and Airborne Instrumentation for Astronomy VIII 14 pages, 7 figures

Journal ref: Proc. SPIE 11447, Ground-based and Airborne Instrumentation for Astronomy VIII, 2020

GENIE: Toward Reproducible and Standardized Human Evaluation for Text Generation

Authors: Daniel Khashabi, Gabriel Stanovsky, Jonathan Bragg, Nicholas Lourie, Jungo Kasai, Ye** Choi, Noah A. Smith, Daniel S. Weld

Abstract: While often assumed a gold standard, effective human evaluation of text generation remains an important, open area for research. We revisit this problem with a focus on producing consistent evaluations that are reproducible -- over time and across different populations. We study this goal in different stages of the human evaluation pipeline. In particular, we consider design choices for the annota… ▽ More While often assumed a gold standard, effective human evaluation of text generation remains an important, open area for research. We revisit this problem with a focus on producing consistent evaluations that are reproducible -- over time and across different populations. We study this goal in different stages of the human evaluation pipeline. In particular, we consider design choices for the annotation interface used to elicit human judgments and their impact on reproducibility. Furthermore, we develop an automated mechanism for maintaining annotator quality via a probabilistic model that detects and excludes noisy annotators. Putting these lessons together, we introduce GENIE: a system for running standardized human evaluations across different generation tasks. We instantiate GENIE with datasets representing four core challenges in text generation: machine translation, summarization, commonsense reasoning, and machine comprehension. For each task, GENIE offers a leaderboard that automatically crowdsources annotations for submissions, evaluating them along axes such as correctness, conciseness, and fluency. We have made the GENIE leaderboards publicly available, and have already ranked 50 submissions from 10 different research groups. We hope GENIE encourages further progress toward effective, standardized evaluations for text generation. △ Less

Submitted 31 October, 2022; v1 submitted 16 January, 2021; originally announced January 2021.

Comments: Accepted to EMNLP 2022 main conference, visit our project page at: https://genie.apps.allenai.org

The Balloon-Borne Large Aperture Submillimeter Telescope Observatory

Authors: Ian Lowe, Gabriele Coppi, Peter A. R. Ade, Peter C. Ashton, Jason E. Austermann, James Beall, Susan Clark, Erin G. Cox, Mark J. Devlin, Simon Dicker, Bradley J. Dober, Valentina Fanfani, Laura M. Fissel, Nicholas Galitzki, Jiansong Gao, Brandon Hensley, Johannes Hubmayr, Steven Li, Zhi-Yun Li, Nathan P. Lourie, Peter G. Martin, Philip Mauskopf, Federico Nati, Giles Novak, Giampaolo Pisano , et al. (8 additional authors not shown)

Abstract: The BLAST Observatory is a proposed superpressure balloon-borne polarimeter designed for a future ultra-long duration balloon campaign from Wanaka, New Zealand. To maximize scientific output while staying within the stringent superpressure weight envelope, BLAST will feature new 1.8m off-axis optical system contained within a lightweight monocoque structure gondola. The payload will incorporate a… ▽ More The BLAST Observatory is a proposed superpressure balloon-borne polarimeter designed for a future ultra-long duration balloon campaign from Wanaka, New Zealand. To maximize scientific output while staying within the stringent superpressure weight envelope, BLAST will feature new 1.8m off-axis optical system contained within a lightweight monocoque structure gondola. The payload will incorporate a 300L $^4$He cryogenic receiver which will cool 8,274 microwave kinetic inductance detectors (MKIDs) to 100mK through the use of an adiabatic demagnetization refrigerator (ADR) in combination with a $^3$He sorption refrigerator all backed by a liquid helium pumped pot operating at 2K. The detector readout utilizes a new Xilinx RFSOC-based system which will run the next-generation of the BLAST-TNG KIDPy software. With this instrument we aim to answer outstanding questions about dust dynamics as well as provide community access to the polarized submillimeter sky made possible by high-altitude observing unrestricted by atmospheric transmission. The BLAST Observatory is designed for a minimum 31-day flight of which 70$\%$ will be dedicated to observations for BLAST scientific goals and the remaining 30$\%$ will be open to proposals from the wider astronomical community through a shared-risk proposals program. △ Less

Submitted 2 December, 2020; originally announced December 2020.

Comments: Presented at SPIE Ground-based and Airborne Telescopes VIII, December 13-18, 2020

Characterization, deployment, and in-flight performance of the BLAST-TNG cryogenic receiver

Authors: Ian Lowe, Peter A. R. Ade, Peter C. Ashton, Jason E. Austermann, Gabriele Coppi, Erin G. Cox, Mark J. Devlin, Bradley J. Dober, Valentina Fanfani, Laura M. Fissel, Nicholas Galitzki, Jiansong Gao, Samuel Gordon, Christopher E. Groppi, Gene C. Hilton, Johannes Hubmayr, Jeffrey Klein, Dale Li, Nathan P. Lourie, Hamdi Mani, Philip Mauskopf, Christopher McKenney, Federico Nati, Giles Novak, Giampaolo Pisano , et al. (8 additional authors not shown)

Abstract: The Next Generation Balloon-borne Large Aperture Submillimeter Telescope (BLAST-TNG) is a submillimeter polarimeter designed to map interstellar dust and galactic foregrounds at 250, 350, and 500 microns during a 24-day Antarctic flight. The BLAST-TNG detector arrays are comprised of 918, 469, and 272 MKID pixels, respectively. The pixels are formed from two orthogonally oriented, crossed, linear-… ▽ More The Next Generation Balloon-borne Large Aperture Submillimeter Telescope (BLAST-TNG) is a submillimeter polarimeter designed to map interstellar dust and galactic foregrounds at 250, 350, and 500 microns during a 24-day Antarctic flight. The BLAST-TNG detector arrays are comprised of 918, 469, and 272 MKID pixels, respectively. The pixels are formed from two orthogonally oriented, crossed, linear-polarization sensitive MKID antennae. The arrays are cooled to sub 300mK temperatures and stabilized via a closed cycle $^3$He sorption fridge in combination with a $^4$He vacuum pot. The detectors are read out through a combination of the second-generation Reconfigurable Open Architecture Computing Hardware (ROACH2) and custom RF electronics designed for BLAST-TNG. The firmware and software designed to readout and characterize these detectors was built from scratch by the BLAST team around these detectors, and has been adapted for use by other MKID instruments such as TolTEC and OLIMPO. We present an overview of these systems as well as in-depth methodology of the ground-based characterization and the measured in-flight performance. △ Less

Submitted 12 January, 2021; v1 submitted 2 December, 2020; originally announced December 2020.

Comments: Presented at SPIE Millimeter, Submillimeter, and Far-Infrared Detectors and Instrumentation for Astronomy X, December 13-18, 2020

Journal ref: Proc. SPIE 11453, Millimeter, Submillimeter, and Far-Infrared Detectors and Instrumentation for Astronomy X, 1145304 (13 December 2020)

In-flight performance of the BLAST-TNG telescope platform

Authors: Gabriele Coppi, Peter A. R. Ade, Peter C. Ashton, Jason E. Austermann, Erin G. Cox, Mark J. Devlin, Bradley J. Dober, Valentina Fanfani, Laura M. Fissel, Nicholas Galitzki, Jiansong Gao, Samuel Gordon, Christopher E. Groppi, Gene C. Hilton, Johannes Hubmayr, Jeffrey Klein, Dale Li, Nathan P. Lourie, Ian Lowe, Hamdi Mani, Philip Mauskopf, Christopher McKenney, Federico Nati, Giles Novak, Giampaolo Pisano , et al. (8 additional authors not shown)

Abstract: The Next Generation Balloon-Borne Large Aperture Submillimeter Telescope (BLAST-TNG) was a unique instrument for characterizing the polarized submillimeter sky at high-angular resolution. BLAST-TNG flew from the Long Duration Balloon Facility in Antarctica in January 2020. Despite the short flight duration, the instrument worked very well and is providing significant information about each subsyst… ▽ More The Next Generation Balloon-Borne Large Aperture Submillimeter Telescope (BLAST-TNG) was a unique instrument for characterizing the polarized submillimeter sky at high-angular resolution. BLAST-TNG flew from the Long Duration Balloon Facility in Antarctica in January 2020. Despite the short flight duration, the instrument worked very well and is providing significant information about each subsystem that will be invaluable for future balloon missions. In this contribution, we discuss the performance of telescope and gondola. △ Less

Submitted 2 January, 2021; v1 submitted 2 December, 2020; originally announced December 2020.

Comments: Submitted to SPIE Astronomical Telescopes + Instrumentation, Ground-based and Airborne Telescopes VIII

Journal ref: Proc. SPIE 11445, Ground-based and Airborne Telescopes VIII, 2020

Learning from Task Descriptions

Authors: Orion Weller, Nicholas Lourie, Matt Gardner, Matthew E. Peters

Abstract: Typically, machine learning systems solve new tasks by training on thousands of examples. In contrast, humans can solve new tasks by reading some instructions, with perhaps an example or two. To take a step toward closing this gap, we introduce a framework for develo** NLP systems that solve new tasks after reading their descriptions, synthesizing prior work in this area. We instantiate this fra… ▽ More Typically, machine learning systems solve new tasks by training on thousands of examples. In contrast, humans can solve new tasks by reading some instructions, with perhaps an example or two. To take a step toward closing this gap, we introduce a framework for develo** NLP systems that solve new tasks after reading their descriptions, synthesizing prior work in this area. We instantiate this framework with a new English language dataset, ZEST, structured for task-oriented evaluation on unseen tasks. Formulating task descriptions as questions, we ensure each is general enough to apply to many possible inputs, thus comprehensively evaluating a model's ability to solve each task. Moreover, the dataset's structure tests specific types of systematic generalization. We find that the state-of-the-art T5 model achieves a score of 12% on ZEST, leaving a significant challenge for NLP researchers. △ Less

Submitted 16 November, 2020; originally announced November 2020.

Comments: EMNLP 2020

Dataset Cartography: Map** and Diagnosing Datasets with Training Dynamics

Authors: Swabha Swayamdipta, Roy Schwartz, Nicholas Lourie, Yizhong Wang, Hannaneh Hajishirzi, Noah A. Smith, Ye** Choi

Abstract: Large datasets have become commonplace in NLP research. However, the increased emphasis on data quantity has made it challenging to assess the quality of data. We introduce Data Maps---a model-based tool to characterize and diagnose datasets. We leverage a largely ignored source of information: the behavior of the model on individual instances during training (training dynamics) for building data… ▽ More Large datasets have become commonplace in NLP research. However, the increased emphasis on data quantity has made it challenging to assess the quality of data. We introduce Data Maps---a model-based tool to characterize and diagnose datasets. We leverage a largely ignored source of information: the behavior of the model on individual instances during training (training dynamics) for building data maps. This yields two intuitive measures for each example---the model's confidence in the true class, and the variability of this confidence across epochs---obtained in a single run of training. Experiments across four datasets show that these model-dependent measures reveal three distinct regions in the data map, each with pronounced characteristics. First, our data maps show the presence of "ambiguous" regions with respect to the model, which contribute the most towards out-of-distribution generalization. Second, the most populous regions in the data are "easy to learn" for the model, and play an important role in model optimization. Finally, data maps uncover a region with instances that the model finds "hard to learn"; these often correspond to labeling errors. Our results indicate that a shift in focus from quantity to quality of data could lead to robust models and improved out-of-distribution generalization. △ Less

Submitted 15 October, 2020; v1 submitted 22 September, 2020; originally announced September 2020.

Comments: Proceedings of EMNLP 2020

Scruples: A Corpus of Community Ethical Judgments on 32,000 Real-Life Anecdotes

Authors: Nicholas Lourie, Ronan Le Bras, Ye** Choi

Abstract: As AI systems become an increasing part of people's everyday lives, it becomes ever more important that they understand people's ethical norms. Motivated by descriptive ethics, a field of study that focuses on people's descriptive judgments rather than theoretical prescriptions on morality, we investigate a novel, data-driven approach to machine ethics. We introduce Scruples, the first large-sca… ▽ More As AI systems become an increasing part of people's everyday lives, it becomes ever more important that they understand people's ethical norms. Motivated by descriptive ethics, a field of study that focuses on people's descriptive judgments rather than theoretical prescriptions on morality, we investigate a novel, data-driven approach to machine ethics. We introduce Scruples, the first large-scale dataset with 625,000 ethical judgments over 32,000 real-life anecdotes. Each anecdote recounts a complex ethical situation, often posing moral dilemmas, paired with a distribution of judgments contributed by the community members. Our dataset presents a major challenge to state-of-the-art neural language models, leaving significant room for improvement. However, when presented with simplified moral situations, the results are considerably more promising, suggesting that neural models can effectively learn simpler ethical building blocks. A key take-away of our empirical analysis is that norms are not always clean-cut; many situations are naturally divisive. We present a new method to estimate the best possible performance on such tasks with inherently diverse label distributions, and explore likelihood functions that separate intrinsic from model uncertainty. △ Less

Submitted 24 March, 2021; v1 submitted 20 August, 2020; originally announced August 2020.

Comments: 18 pages, 14 tables, 18 figures. Accepted to AAAI 2021. For associated code and data, see https://github.com/allenai/scruples

CommonsenseQA: A Question Answering Challenge Targeting Commonsense Knowledge

Authors: Alon Talmor, Jonathan Herzig, Nicholas Lourie, Jonathan Berant

Abstract: When answering a question, people often draw upon their rich world knowledge in addition to the particular context. Recent work has focused primarily on answering questions given some relevant document or context, and required very little general background. To investigate question answering with prior knowledge, we present CommonsenseQA: a challenging new dataset for commonsense question answerin… ▽ More When answering a question, people often draw upon their rich world knowledge in addition to the particular context. Recent work has focused primarily on answering questions given some relevant document or context, and required very little general background. To investigate question answering with prior knowledge, we present CommonsenseQA: a challenging new dataset for commonsense question answering. To capture common sense beyond associations, we extract from ConceptNet (Speer et al., 2017) multiple target concepts that have the same semantic relation to a single source concept. Crowd-workers are asked to author multiple-choice questions that mention the source concept and discriminate in turn between each of the target concepts. This encourages workers to create questions with complex semantics that often require prior knowledge. We create 12,247 questions through this procedure and demonstrate the difficulty of our task with a large number of strong baselines. Our best baseline is based on BERT-large (Devlin et al., 2018) and obtains 56% accuracy, well below human performance, which is 89%. △ Less

Submitted 15 March, 2019; v1 submitted 2 November, 2018; originally announced November 2018.

Comments: accepted as a long paper at NAACL 2019

ATOMIC: An Atlas of Machine Commonsense for If-Then Reasoning

Authors: Maarten Sap, Ronan LeBras, Emily Allaway, Chandra Bhagavatula, Nicholas Lourie, Hannah Rashkin, Brendan Roof, Noah A. Smith, Ye** Choi

Abstract: We present ATOMIC, an atlas of everyday commonsense reasoning, organized through 877k textual descriptions of inferential knowledge. Compared to existing resources that center around taxonomic knowledge, ATOMIC focuses on inferential knowledge organized as typed if-then relations with variables (e.g., "if X pays Y a compliment, then Y will likely return the compliment"). We propose nine if-then re… ▽ More We present ATOMIC, an atlas of everyday commonsense reasoning, organized through 877k textual descriptions of inferential knowledge. Compared to existing resources that center around taxonomic knowledge, ATOMIC focuses on inferential knowledge organized as typed if-then relations with variables (e.g., "if X pays Y a compliment, then Y will likely return the compliment"). We propose nine if-then relation types to distinguish causes vs. effects, agents vs. themes, voluntary vs. involuntary events, and actions vs. mental states. By generatively training on the rich inferential knowledge described in ATOMIC, we show that neural models can acquire simple commonsense capabilities and reason about previously unseen events. Experimental results demonstrate that multitask models that incorporate the hierarchical structure of if-then relation types lead to more accurate inference compared to models trained in isolation, as measured by both automatic and human evaluation. △ Less

Submitted 7 February, 2019; v1 submitted 31 October, 2018; originally announced November 2018.

Comments: AAAI 2019 CR

Design and Characterization of a Balloon-Borne Diffraction-Limited Submillimeter Telescope Platform for BLAST-TNG

Authors: Nathan P. Lourie, Francisco E. Angile, Peter C. Ashton, Brian Catanzaro, Mark J. Devlin, Simon Dicker, Joy Didier, Bradley Dober, Laura M. Fissel, Nicholas Galitzki, Samuel Gordon, Jeffrey Klein, Ian Lowe, Philip Mauskopf, Federico Nati, Giles Novak, L. Javier Romualdez, Juan D. Soler, Paul A. Williams

Abstract: The Next Generation Balloon-borne Large Aperture Submillimeter Telescope (BLAST-TNG) is a submillimeter map** experiment planned for a 28 day long-duration balloon (LDB) flight from McMurdo Station, Antarctica during the 2018-2019 season. BLAST-TNG will detect submillimeter polarized interstellar dust emission, tracing magnetic fields in galactic molecular clouds. BLAST-TNG will be the first pol… ▽ More The Next Generation Balloon-borne Large Aperture Submillimeter Telescope (BLAST-TNG) is a submillimeter map** experiment planned for a 28 day long-duration balloon (LDB) flight from McMurdo Station, Antarctica during the 2018-2019 season. BLAST-TNG will detect submillimeter polarized interstellar dust emission, tracing magnetic fields in galactic molecular clouds. BLAST-TNG will be the first polarimeter with the sensitivity and resolution to probe the $\sim$0.1 parsec-scale features that are critical to understanding the origin of structures in the interstellar medium. With three detector arrays operating at 250, 350, and 500 $μ$m (1200, 857, and 600 GHz), BLAST-TNG will obtain diffraction-limited resolution at each waveband of 30, 41, and 59 arcseconds respectively. To achieve the submillimeter resolution necessary for its science goals, the BLAST-TNG telescope features a 2.5 m aperture carbon fiber composite primary mirror, one of the largest mirrors flown on a balloon platform. Successful performance of such a large telescope on a balloon-borne platform requires stiff, lightweight optical components and mounting structures. Through a combination of optical metrology and finite element modeling of thermal and mechanical stresses on both the telescope optics and mounting structures, we expect diffraction-limited resolution at all our wavebands. We expect pointing errors due to deformation of the telescope mount to be negligible. We have developed a detailed thermal model of the sun shielding, gondola, and optical components to optimize our observing strategy and increase the stability of the telescope over the flight. We present preflight characterization of the telescope and its platform. △ Less

Submitted 26 August, 2018; originally announced August 2018.

Comments: 19 pages, 11 figures, 2018 SPIE Astronomical Telescopes & Instrumentation Conference

Journal ref: SPIE Proceedings Volume 10700, Ground-based and Airborne Telescopes VII; 1070022 (2018)

Preflight Characterization of the BLAST-TNG Receiver and Detector Arrays

Authors: Nathan P. Lourie, Peter A. R. Ade, Francisco E. Angile, Peter C. Ashton, Jason E. Austermann, Mark J. Devlin, Bradley Dober, Nicholas Galitzki, Jiansong Gao, Sam Gordon, Christopher E. Groppi, Jeffrey Klein, Gene C. Hilton, Johannes Hubmayr, Dale Li, Ian Lowe, Hamdi Mani, Philip Mauskopf, Christopher M. McKenney, Federico Nati, Giles Novak, Enzo Pascale, Giampaolo Pisano, Adrian Sinclair, Juan D. Soler , et al. (4 additional authors not shown)

Abstract: The Next Generation Balloon-borne Large Aperture Submillimeter Telescope (BLAST-TNG) is a submillimeter map** experiment planned for a 28 day long-duration balloon (LDB) flight from McMurdo Station, Antarctica during the 2018-2019 season. BLAST-TNG will detect submillimeter polarized interstellar dust emission, tracing magnetic fields in galactic molecular clouds. BLAST-TNG will be the first pol… ▽ More The Next Generation Balloon-borne Large Aperture Submillimeter Telescope (BLAST-TNG) is a submillimeter map** experiment planned for a 28 day long-duration balloon (LDB) flight from McMurdo Station, Antarctica during the 2018-2019 season. BLAST-TNG will detect submillimeter polarized interstellar dust emission, tracing magnetic fields in galactic molecular clouds. BLAST-TNG will be the first polarimeter with the sensitivity and resolution to probe the $\sim$0.1 parsec-scale features that are critical to understanding the origin of structures in the interstellar medium. BLAST-TNG features three detector arrays operating at wavelengths of 250, 350, and 500 $μ$m (1200, 857, and 600 GHz) comprised of 918, 469, and 272 dual-polarization pixels, respectively. Each pixel is made up of two crossed microwave kinetic inductance detectors (MKIDs). These arrays are cooled to 275 mK in a cryogenic receiver. Each MKID has a different resonant frequency, allowing hundreds of resonators to be read out on a single transmission line. This inherent ability to be frequency-domain multiplexed simplifies the cryogenic readout hardware, but requires careful optical testing to map out the physical location of each resonator on the focal plane. Receiver-level optical testing was carried out using both a cryogenic source mounted to a movable xy-stage with a shutter, and a beam-filling, heated blackbody source able to provide a 10-50 $^\circ$C temperature chop. The focal plane array noise properties, responsivity, polarization efficiency, instrumental polarization were measured. We present the preflight characterization of the BLAST-TNG cryogenic system and array-level optical testing of the MKID detector arrays in the flight receiver. △ Less

Submitted 25 August, 2018; originally announced August 2018.

Comments: 15 pages, 7 figures, SPIE Astronomical Telescopes and Instrumentation Conference

Journal ref: Proceedings Volume 10708, Millimeter, Submillimeter, and Far-Infrared Detectors and Instrumentation for Astronomy IX; 107080L (2018)

Extracting Scientific Figures with Distantly Supervised Neural Networks

Authors: Noah Siegel, Nicholas Lourie, Russell Power, Waleed Ammar

Abstract: Non-textual components such as charts, diagrams and tables provide key information in many scientific documents, but the lack of large labeled datasets has impeded the development of data-driven methods for scientific figure extraction. In this paper, we induce high-quality training labels for the task of figure extraction in a large number of scientific documents, with no human intervention. To a… ▽ More Non-textual components such as charts, diagrams and tables provide key information in many scientific documents, but the lack of large labeled datasets has impeded the development of data-driven methods for scientific figure extraction. In this paper, we induce high-quality training labels for the task of figure extraction in a large number of scientific documents, with no human intervention. To accomplish this we leverage the auxiliary data provided in two large web collections of scientific documents (arXiv and PubMed) to locate figures and their associated captions in the rasterized PDF. We share the resulting dataset of over 5.5 million induced labels---4,000 times larger than the previous largest figure extraction dataset---with an average precision of 96.8%, to enable the development of modern data-driven methods for this task. We use this dataset to train a deep neural network for end-to-end figure detection, yielding a model that can be more easily extended to new domains compared to previous work. The model was successfully deployed in Semantic Scholar, a large-scale academic search engine, and used to extract figures in 13 million scientific documents. △ Less

Submitted 30 May, 2018; v1 submitted 6 April, 2018; originally announced April 2018.

Comments: 10 pages, 5 figures, paper accepted at JCDL 2018

An Open Source, FPGA-based LeKID readout for BLAST-TNG: Pre-flight Results

Authors: Samuel Gordon, Bradley Dober, Adrian Sinclair, Samuel Rowe, Sean Bryan, Philip Mauskopf, Jason Austermann, Mark Devlin, Simon Dicker, Jiansong Gao, Gene C. Hilton, Johannes Hubmayr, Glenn Jones, Jeffrey Klein, Nathan P. Lourie, Christopher McKenney, Federico Nati, Juan D. Soler, Matthew Strader, Michael Vissers

Abstract: We present a highly frequency multiplexed readout for large-format superconducting detector arrays intended for use in the next generation of balloon-borne and space-based sub-millimeter and far-infrared missions. We will demonstrate this technology on the upcoming NASA Next Generation Balloon-borne Large Aperture Sub-millimeter Telescope (BLAST-TNG) to measure the polarized emission of Galactic d… ▽ More We present a highly frequency multiplexed readout for large-format superconducting detector arrays intended for use in the next generation of balloon-borne and space-based sub-millimeter and far-infrared missions. We will demonstrate this technology on the upcoming NASA Next Generation Balloon-borne Large Aperture Sub-millimeter Telescope (BLAST-TNG) to measure the polarized emission of Galactic dust at wavelengths of 250, 350 and 500 microns. The BLAST-TNG receiver incorporates the first arrays of Lumped Element Kinetic Inductance Detectors (LeKID) along with the first microwave multiplexing readout electronics to fly in a space-like environment and will significantly advance the TRL for these technologies. After the flight of BLAST-TNG, we will continue to improve the performance of the detectors and readout electronics for the next generation of balloon-borne instruments and for use in a future FIR Surveyor. △ Less

Submitted 16 November, 2016; originally announced November 2016.

Instrumental performance and results from testing of the BLAST-TNG receiver, submillimeter optics, and MKID arrays

Authors: Nicholas Galitzki, Peter Ade, Francesco E. Angile, Peter Ashton, Jason Austermann, Tashalee Billings, George Che, Hsiao-Mei Cho, Kristina Davis, Mark Devlin, Simon Dicker, Bradley J. Dober, Laura M. Fissel, Yasuo Fukui, Jiansong Gao, Samuel Gordon, Christopher E. Groppi, Seth Hillbrand, Gene C. Hilton, Johannes Hubmayr, Kent D. Irwin, Jeffrey Klein, Dale Li, Zhi-Yun Li, Nathan P. Lourie , et al. (17 additional authors not shown)

Abstract: Polarized thermal emission from interstellar dust grains can be used to map magnetic fields in star forming molecular clouds and the diffuse interstellar medium (ISM). The Balloon-borne Large Aperture Submillimeter Telescope for Polarimetry (BLASTPol) flew from Antarctica in 2010 and 2012 and produced degree-scale polarization maps of several nearby molecular clouds with arcminute resolution. The… ▽ More Polarized thermal emission from interstellar dust grains can be used to map magnetic fields in star forming molecular clouds and the diffuse interstellar medium (ISM). The Balloon-borne Large Aperture Submillimeter Telescope for Polarimetry (BLASTPol) flew from Antarctica in 2010 and 2012 and produced degree-scale polarization maps of several nearby molecular clouds with arcminute resolution. The success of BLASTPol has motivated a next-generation instrument, BLAST-TNG, which will use more than 3000 linear polarization sensitive microwave kinetic inductance detectors (MKIDs) combined with a 2.5m diameter carbon fiber primary mirror to make diffraction-limited observations at 250, 350, and 500 $μ$m. With 16 times the map** speed of BLASTPol, sub-arcminute resolution, and a longer flight time, BLAST-TNG will be able to examine nearby molecular clouds and the diffuse galactic dust polarization spectrum in unprecedented detail. The 250 $μ$m detector array has been integrated into the new cryogenic receiver, and is undergoing testing to establish the optical and polarization characteristics of the instrument. BLAST-TNG will demonstrate the effectiveness of kilo-pixel MKID arrays for applications in submillimeter astronomy. BLAST-TNG is scheduled to fly from Antarctica in December 2017 for 28 days and will be the first balloon-borne telescope to offer a quarter of the flight for "shared risk" observing by the community. △ Less

Submitted 18 August, 2016; originally announced August 2016.

Comments: Presented at SPIE Millimeter, Submillimeter, and Far-Infrared Detectors and Instrumentation for Astronomy VIII, June 29th, 2016

Journal ref: Proc. SPIE 9914, Millimeter, Submillimeter, and Far-Infrared Detectors and Instrumentation for Astronomy VIII, 99140J (July 19, 2016)

Optical Demonstration of THz, Dual-Polarization Sensitive Microwave Kinetic Inductance Detectors

Authors: B. Dober, J. A. Austermann, J. A. Beall, D. Becker, G. Che, H. M. Cho, M. Devlin, S. M. Duff, N. Galitzki, J. Gao, C. Groppi, G. C. Hilton, J. Hubmayr, K. D. Irwin, C. M. McKenney, D. Li, N. Lourie, P. Mauskopf, M. R. Vissers, Y. Wang

Abstract: The next generation BLAST experiment (BLAST-TNG) is a suborbital balloon payload that seeks to map polarized dust emission in the 250 $μ$m, 350 $μ$m and 500 $μ$m wavebands. The instrument utilizes a stepped half-wave plate to reduce systematics. The general requirement of the detectors is that they are photon-noise-limited and dual-polarization sensitive. To achieve this goal, we are develo** th… ▽ More The next generation BLAST experiment (BLAST-TNG) is a suborbital balloon payload that seeks to map polarized dust emission in the 250 $μ$m, 350 $μ$m and 500 $μ$m wavebands. The instrument utilizes a stepped half-wave plate to reduce systematics. The general requirement of the detectors is that they are photon-noise-limited and dual-polarization sensitive. To achieve this goal, we are develo** three monolithic arrays of cryogenic sensors, one for each waveband. Each array is feedhorn-coupled and each spatial pixel consists of two orthogonally spaced polarization-sensitive microwave kinetic inductance detectors (MKIDs) fabricated from a Ti/TiN multilayer film. In previous work, we demonstrated photon-noise-limited sensitivity in 250 $μ$m waveband single polarization devices. In this work, we present the first results of dual-polarization sensitive MKIDs at 250 $μ$m. △ Less

Submitted 9 March, 2016; originally announced March 2016.

Comments: 7 pages, 4 figures, published by JLTP

Journal ref: Journal of Low Temperature Physics (2015): 1-7

The Next Generation BLAST Experiment

Authors: Nicholas Galitzki, Peter A. R. Ade, Francesco E. Angilè, Peter Ashton, James A. Beall, Dan Becker, Kristi J. Bradford, George Che, Hsiao-Mei Cho, Mark J. Devlin, Bradley J. Dober, Laura M. Fissel, Yasuo Fukui, Jiansong Gao, Christopher E. Groppi, Seth Hillbrand, Gene C. Hilton, Johannes Hubmayr, Kent D. Irwin, Jeffrey Klein, Jeff Van Lanen, Dale Li, Zhi-Yun Li, Nathan P. Lourie, Hamdi Mani , et al. (16 additional authors not shown)

Abstract: The Balloon-borne Large Aperture Submillimeter Telescope for Polarimetry (BLASTPol) was a suborbital experiment designed to map magnetic fields in order to study their role in star formation processes. BLASTPol made detailed polarization maps of a number of molecular clouds during its successful flights from Antarctica in 2010 and 2012. We present the next-generation BLASTPol instrument (BLAST-TNG… ▽ More The Balloon-borne Large Aperture Submillimeter Telescope for Polarimetry (BLASTPol) was a suborbital experiment designed to map magnetic fields in order to study their role in star formation processes. BLASTPol made detailed polarization maps of a number of molecular clouds during its successful flights from Antarctica in 2010 and 2012. We present the next-generation BLASTPol instrument (BLAST-TNG) that will build off the success of the previous experiment and continue its role as a unique instrument and a test bed for new technologies. With a 16-fold increase in map** speed, BLAST-TNG will make larger and deeper maps. Major improvements include a 2.5 m carbon fiber mirror that is 40% wider than the BLASTPol mirror and ~3000 polarization sensitive detectors. BLAST-TNG will observe in three bands at 250, 350, and 500 microns. The telescope will serve as a pathfinder project for microwave kinetic inductance detector (MKID) technology, as applied to feedhorn coupled submillimeter detector arrays. The liquid helium cooled cryostat will have a 28-day hold time and will utilize a closed-cycle $^3$He refrigerator to cool the detector arrays to 270 mK. This will enable a detailed map** of more targets with higher polarization resolution than any other submillimeter experiment to date. BLAST-TNG will also be the first balloon-borne telescope to offer shared risk observing time to the community. This paper outlines the motivation for the project and the instrumental design. △ Less

Submitted 12 November, 2014; v1 submitted 24 September, 2014; originally announced September 2014.

Journal ref: Journal of Astronomical Instrumentation, Vol. 3, No. 2 (2014) 1440001 (16 pages)

Tripod configurations of curves

Authors: Eric Chen, Nick Lourie

Abstract: Tripod configurations of plane curves, formed by certain triples of normal lines coinciding at a point, were introduced by Tabachnikov, who showed that $C^2$ closed convex curves possess at least two tripod configurations. Later, Kao and Wang established the existence of tripod configurations for $C^2$ closed locally convex curves. In this paper we generalize these two results, answering a conject… ▽ More Tripod configurations of plane curves, formed by certain triples of normal lines coinciding at a point, were introduced by Tabachnikov, who showed that $C^2$ closed convex curves possess at least two tripod configurations. Later, Kao and Wang established the existence of tripod configurations for $C^2$ closed locally convex curves. In this paper we generalize these two results, answering a conjecture of Tabachnikov on the existence of tripod configurations for all $C^2$ closed plane curves by proving existence for a generalized notion of tripod configuration. We then demonstrate the existence of the natural extensions of these tripod configurations to the spherical and hyperbolic geometries for a certain class of convex curves, and discuss an analogue of the problem for regular plane polygons. △ Less

Submitted 20 August, 2014; originally announced August 2014.

Journal ref: J.Geom.Phys. 89 (2015) 1-16

The next-generation BLASTPol experiment

Authors: Bradley Dober, Peter A. R. Ade, Peter Ashton, Francesco E. Angilè, James A. Beall, Dan Becker, Kristi J. Bradford, George Che, Hsiao-Mei Cho, Mark J. Devlin, Laura M. Fissel, Yasuo Fukui, Nicholas Galitzki, Jiansong Gao, Christopher E. Groppi, Seth Hillbrand, Gene C. Hilton, Johannes Hubmayr, Kent D. Irwin, Jeffrey Klein, Jeff Van Lanen, Dale Li, Zhi-Yun Li, Nathan P. Lourie, Hamdi Mani , et al. (14 additional authors not shown)

Abstract: The Balloon-borne Large Aperture Submillimeter Telescope for Polarimetry (BLASTPol) is a suborbital map** experiment designed to study the role magnetic fields play in star formation. BLASTPol has had two science flights from McMurdo Station, Antarctica in 2010 and 2012. These flights have produced thousands of polarization vectors at 250, 350 and 500 microns in several molecular cloud targets.… ▽ More The Balloon-borne Large Aperture Submillimeter Telescope for Polarimetry (BLASTPol) is a suborbital map** experiment designed to study the role magnetic fields play in star formation. BLASTPol has had two science flights from McMurdo Station, Antarctica in 2010 and 2012. These flights have produced thousands of polarization vectors at 250, 350 and 500 microns in several molecular cloud targets. We present the design, specifications, and progress towards the next-generation BLASTPol experiment (BLAST-TNG). BLAST-TNG will fly a 40% larger diameter primary mirror, with almost 8 times the number of polarization-sensitive detectors resulting in a factor of 16 increase in map** speed. With a spatial resolution of 22 arcseconds and four times the field of view of BLASTPol, BLAST-TNG will bridge the angular scales between Planck's low resolution all-sky maps and ALMA's ultra-high resolution narrow fields. The new receiver has a larger cryogenics volume, allowing for a 28 day hold time. BLAST-TNG employs three arrays of Microwave Kinetic Inductance Detectors (MKIDs) with 30% fractional bandwidth at 250, 350 and 500 microns. In this paper, we will present the new BLAST-TNG instrument and science objectives. △ Less

Submitted 14 July, 2014; originally announced July 2014.

Comments: 12 pages, 6 figures. Presented at SPIE Ground-based and Airborne Telescopes V, June 23, 2014. To be published in Proceedings of SPIE Volume 9145

Precision control of thermal transport in cryogenic single-crystal silicon devices

Authors: Karwan Rostem, David T. Chuss, Felipe A. Colazo, Erik J. Crowe, Kevin L. Denis, Nathan P. Lourie, Samuel H. Moseley, Thomas R. Stevenson, Edward J. Wollack

Abstract: We report on the diffusive-ballistic thermal conductance of multi-moded single-crystal silicon beams measured below 1 K. It is shown that the phonon mean-free-path $\ell$ is a strong function of the surface roughness characteristics of the beams. This effect is enhanced in diffuse beams with lengths much larger than $\ell$, even when the surface is fairly smooth, 5-10 nm rms, and the peak thermal… ▽ More We report on the diffusive-ballistic thermal conductance of multi-moded single-crystal silicon beams measured below 1 K. It is shown that the phonon mean-free-path $\ell$ is a strong function of the surface roughness characteristics of the beams. This effect is enhanced in diffuse beams with lengths much larger than $\ell$, even when the surface is fairly smooth, 5-10 nm rms, and the peak thermal wavelength is 0.6 $μ$m. Resonant phonon scattering has been observed in beams with a pitted surface morphology and characteristic pit depth of 30 nm. Hence, if the surface roughness is not adequately controlled, the thermal conductance can vary significantly for diffuse beams fabricated across a wafer. In contrast, when the beam length is of order $\ell$, the conductance is dominated by ballistic transport and is effectively set by the beam area. We have demonstrated a uniformity of $\pm$8% in fractional deviation for ballistic beams, and this deviation is largely set by the thermal conductance of diffuse beams that support the micro-electro-mechanical device and electrical leads. In addition, we have found no evidence for excess specific heat in single-crystal silicon membranes. This allows for the precise control of the device heat capacity with normal metal films. We discuss the results in the context of the design and fabrication of large-format arrays of far-infrared and millimeter wavelength cryogenic detectors. △ Less

Submitted 5 March, 2014; originally announced March 2014.

A waveguide-coupled thermally-isolated radiometric source

Authors: Karwan Rostem, David T. Chuss, Nathan P. Lourie, George M. Voellmer, Edward J. Wollack

Abstract: The design and validation of a dual polarization source for waveguide-coupled millimeter and sub-millimeter wave cryogenic sensors is presented. The thermal source is a waveguide mounted absorbing conical dielectric taper. The absorber is thermally isolated with a kinematic suspension that allows the guide to be heat sunk to the lowest bath temperature of the cryogenic system. This approach enable… ▽ More The design and validation of a dual polarization source for waveguide-coupled millimeter and sub-millimeter wave cryogenic sensors is presented. The thermal source is a waveguide mounted absorbing conical dielectric taper. The absorber is thermally isolated with a kinematic suspension that allows the guide to be heat sunk to the lowest bath temperature of the cryogenic system. This approach enables the thermal emission from the metallic waveguide walls to be subdominant to that from the source. The use of low thermal conductivity Kevlar threads for the kinematic mount effectively decouples the absorber from the sensor cold stage. Hence, the absorber can be heated to significantly higher temperatures than the sensor with negligible conductive loading. The kinematic suspension provides high mechanical repeatability and reliability with thermal cycling. A 33-50 GHz blackbody source demonstrates an emissivity of 0.999 over the full waveguide band where the dominant deviation from unity arrises from the waveguide ohmic loss. The observed thermal time constant of the source is 40 s when the absorber temperature is 15 K. The specific heat of the lossy dielectric MF-117 is well approximated by $C_v(T)=0.12\,T\,^{2.06}$ mJ g$^{-1}$ K$^{-1}$ between 3.5 K and 15 K. △ Less

Submitted 21 March, 2013; originally announced March 2013.

Journal ref: Rev. Sci. Instrum. 84, 044701 (2013)