Showing 1–50 of 66 results for author: Hayes, L

Searching for rapid pulsations in solar flare X-ray data

Authors: Andrew R. Inglis, Laura A. Hayes

Abstract: Most studies of quasi-periodic pulsations in solar flares have identified characteristic periods in the 5 - 300s range. Due to observational limitations there have been few attempts to probe the < 5s period regime and understand the prevalence of such short-period quasi-periodic pulsations. However, the Fermi Gamma-ray Burst Monitor (GBM) has observed approximately 1500 solar flares to date in hig… ▽ More Most studies of quasi-periodic pulsations in solar flares have identified characteristic periods in the 5 - 300s range. Due to observational limitations there have been few attempts to probe the < 5s period regime and understand the prevalence of such short-period quasi-periodic pulsations. However, the Fermi Gamma-ray Burst Monitor (GBM) has observed approximately 1500 solar flares to date in high cadence 16 Hz burst mode, providing us with an opportunity to study short-period quasi-periodic pulsations at X-ray energies. We systematically analyse every solar flare observed by Fermi/GBM in burst mode, estimating the prevalence of quasi-periodic pulsations in multiple X-ray energy bands. To better understand these results, we complement this with analysis of synthetic solar flare lightcurves, both with and without oscillatory signals present. Using these synthetic lightcurves, we can understand the likely false alarm and true positive rates in the real solar GBM data. We do not find strong evidence for widespread short-period quasi-periodic pulsations, indicating either a low base occurrence rate of such signatures or that their typical signal-to-noise ratios must be low - less than 1 - in Fermi/GBM data. Finally, we present a selection of the most interesting potential quasi-periodic pulsation events that were identified in the GBM solar X-ray data. △ Less

Submitted 11 June, 2024; originally announced June 2024.

Comments: Accepted for publication in the Astrophysical Journal. 14 pages, 14 figures

SHiNe: Semantic Hierarchy Nexus for Open-vocabulary Object Detection

Authors: Mingxuan Liu, Tyler L. Hayes, Elisa Ricci, Gabriela Csurka, Riccardo Volpi

Abstract: Open-vocabulary object detection (OvOD) has transformed detection into a language-guided task, empowering users to freely define their class vocabularies of interest during inference. However, our initial investigation indicates that existing OvOD detectors exhibit significant variability when dealing with vocabularies across various semantic granularities, posing a concern for real-world deployme… ▽ More Open-vocabulary object detection (OvOD) has transformed detection into a language-guided task, empowering users to freely define their class vocabularies of interest during inference. However, our initial investigation indicates that existing OvOD detectors exhibit significant variability when dealing with vocabularies across various semantic granularities, posing a concern for real-world deployment. To this end, we introduce Semantic Hierarchy Nexus (SHiNe), a novel classifier that uses semantic knowledge from class hierarchies. It runs offline in three steps: i) it retrieves relevant super-/sub-categories from a hierarchy for each target class; ii) it integrates these categories into hierarchy-aware sentences; iii) it fuses these sentence embeddings to generate the nexus classifier vector. Our evaluation on various detection benchmarks demonstrates that SHiNe enhances robustness across diverse vocabulary granularities, achieving up to +31.9% mAP50 with ground truth hierarchies, while retaining improvements using hierarchies generated by large language models. Moreover, when applied to open-vocabulary classification on ImageNet-1k, SHiNe improves the CLIP zero-shot baseline by +2.8% accuracy. SHiNe is training-free and can be seamlessly integrated with any off-the-shelf OvOD detector, without incurring additional computational overhead during inference. The code is open source. △ Less

Submitted 16 May, 2024; originally announced May 2024.

Comments: Accepted as a conference paper (highlight) at CVPR 2024

The solar cycle 25 multi-spacecraft solar energetic particle event catalog of the SERPENTINE project

Authors: N. Dresing, A. Yli-Laurila, S. Valkila, J. Gieseler, D. E. Morosan, G. U. Farwa, Y. Kartavykh, C. Palmroos, I. Jebaraj, S. Jensen, P. Kühl, B. Heber, F. Espinosa, R. Gómez-Herrero, E. Kilpua, V. -V. Linho, P. Oleynik, L. A. Hayes, A. Warmuth, F. Schuller, H. Collier, H. Xiao, E. Asvestari, D. Trotta, J. G. Mitchell , et al. (4 additional authors not shown)

Abstract: The Solar energetic particle analysis platform for the inner heliosphere (SERPENTINE) project presents it's new multi-spacecraft SEP event catalog for events observed in solar cycle 25. Observations from five different viewpoints are utilized, provided by Solar Orbiter, Parker Solar Probe, STEREO A, BepiColombo, and the near-Earth spacecraft Wind and SOHO. The catalog contains key SEP parameters f… ▽ More The Solar energetic particle analysis platform for the inner heliosphere (SERPENTINE) project presents it's new multi-spacecraft SEP event catalog for events observed in solar cycle 25. Observations from five different viewpoints are utilized, provided by Solar Orbiter, Parker Solar Probe, STEREO A, BepiColombo, and the near-Earth spacecraft Wind and SOHO. The catalog contains key SEP parameters for 25-40 MeV protons, 1 MeV electrons, and 100 keV electrons. Furthermore, basic parameters of the associated flare and type-II radio burst are listed, as well as the coordinates of the observer and solar source locations. SEP onset times are determined using the Poisson-CUSUM method. SEP peak times and intensities refer to the global intensity maximum. If different viewing directions are available, we use the one with the earliest onset for the onset determination and the one with the highest peak intensity for the peak identification. Associated flares are identified using observations from near Earth and Solar Orbiter. Associated type II radio bursts are determined from ground-based observations in the metric frequency range and from spacecraft observations in the decametric range. The current version of the catalog contains 45 multi-spacecraft events observed in the period from Nov 2020 until May 2023, of which 13 were widespread events and four were classified as narrow-spread events. Using X-ray observations by GOES/XRS and Solar Orbiter/STIX, we were able to identify the associated flare in all but four events. Using ground-based and space-borne radio observations, we found an associated type-II radio burst for 40 events. In total, the catalog contains 142 single event observations, of which 20 (45) have been observed at radial distances below 0.6 AU (0.8 AU). △ Less

Submitted 1 March, 2024; originally announced March 2024.

Journal ref: A&A 687, A72 (2024)

PANDAS: Prototype-based Novel Class Discovery and Detection

Authors: Tyler L. Hayes, César R. de Souza, Namil Kim, Jiwon Kim, Riccardo Volpi, Diane Larlus

Abstract: Object detectors are typically trained once and for all on a fixed set of classes. However, this closed-world assumption is unrealistic in practice, as new classes will inevitably emerge after the detector is deployed in the wild. In this work, we look at ways to extend a detector trained for a set of base classes so it can i) spot the presence of novel classes, and ii) automatically enrich its re… ▽ More Object detectors are typically trained once and for all on a fixed set of classes. However, this closed-world assumption is unrealistic in practice, as new classes will inevitably emerge after the detector is deployed in the wild. In this work, we look at ways to extend a detector trained for a set of base classes so it can i) spot the presence of novel classes, and ii) automatically enrich its repertoire to be able to detect those newly discovered classes together with the base ones. We propose PANDAS, a method for novel class discovery and detection. It discovers clusters representing novel classes from unlabeled data, and represents old and new classes with prototypes. During inference, a distance-based classifier uses these prototypes to assign a label to each detected object instance. The simplicity of our method makes it widely applicable. We experimentally demonstrate the effectiveness of PANDAS on the VOC 2012 and COCO-to-LVIS benchmarks. It performs favorably against the state of the art for this task while being computationally more affordable. △ Less

Submitted 30 April, 2024; v1 submitted 27 February, 2024; originally announced February 2024.

Comments: Accepted to the Conference on Lifelong Learning Agents (CoLLAs 2024)

Localising pulsations in the hard X-ray and microwave emission of an X-class flare

Authors: Hannah Collier, Laura A. Hayes, Sijie Yu, Andrea F. Battaglia, William Ashfield, Vanessa Polito, Louise K. Harra, Säm Krucker

Abstract: Aims: This work aims to identify the mechanism driving pulsations in hard X-ray (HXR) and microwave emission during solar flares. Here, by using combined HXR and microwave observations from Solar Orbiter/STIX and EOVSA we investigate an X1.3 GOES class flare, 2022-03-30T17:21:00, which displays pulsations on timescales evolving from ~ 7 s in the impulsive phase to ~ 35 s later in the flare. Meth… ▽ More Aims: This work aims to identify the mechanism driving pulsations in hard X-ray (HXR) and microwave emission during solar flares. Here, by using combined HXR and microwave observations from Solar Orbiter/STIX and EOVSA we investigate an X1.3 GOES class flare, 2022-03-30T17:21:00, which displays pulsations on timescales evolving from ~ 7 s in the impulsive phase to ~ 35 s later in the flare. Methods: The temporal, spatial and spectral evolution of the HXR and microwave pulsations during the impulsive phase of the flare are analysed. Images are reconstructed for individual peaks in the impulsive phase and spectral fitting is performed at high cadence throughout the first phase of pulsations. Results: Imaging analysis demonstrates that the HXR and microwave emission originates from multiple sites along the flare ribbons. The brightest sources and the location of the emission changes in time. Through HXR spectral analysis, the electron spectral index is found to be anti-correlated with the HXR flux showing a "soft-hard-soft" spectral index evolution for each pulsation. The timing of the associated filament eruption coincides with the early impulsive phase. Conclusions: Our results indicate that periodic acceleration and/or injection of electrons from multiple sites along the flare arcade is responsible for the pulsations observed in HXR and microwave. The evolution of pulsation timescales is likely a result of changes in the 3D magnetic field configuration in time related to the associated filament eruption. △ Less

Submitted 16 February, 2024; originally announced February 2024.

Comments: 15 pages, 8 figures, accepted by A&A

Journal ref: A&A 684, A215 (2024)

A Modelling Investigation for Solar Flare X-ray Stereoscopy with Solar Orbiter/STIX and Earth Orbiting Missions

Authors: Natasha L. S. Jeffrey, Säm Krucker, Morgan Stores, Eduard P. Kontar, Pascal Saint-Hilaire, Andrea F. Battaglia, Laura Hayes, Hannah Collier, Astrid Veronig, Yang Su, Srikar Paavan Tadepalli, Fanxiaoyu Xia

Abstract: The Spectrometer/Telescope for Imaging X-rays (STIX) on board Solar Orbiter (SolO) provides a unique opportunity to systematically perform stereoscopic X-ray observations of solar flares with current and upcoming X-ray missions at Earth. These observations will produce the first reliable measurements of hard X-ray (HXR) directivity in decades, providing a new diagnostic of the flare-accelerated el… ▽ More The Spectrometer/Telescope for Imaging X-rays (STIX) on board Solar Orbiter (SolO) provides a unique opportunity to systematically perform stereoscopic X-ray observations of solar flares with current and upcoming X-ray missions at Earth. These observations will produce the first reliable measurements of hard X-ray (HXR) directivity in decades, providing a new diagnostic of the flare-accelerated electron angular distribution and hel** to constrain the processes that accelerate electrons in flares. However, such observations must be compared to modelling, taking into account electron and X-ray transport effects and realistic plasma conditions, all of which can change the properties of the measured HXR directivity. Here, we show how HXR directivity, defined as the ratio of X-ray spectra at different spacecraft viewing angles, varies with different electron and flare properties (e.g., electron angular distribution, highest energy electrons, and magnetic configuration), and how modelling can be used to extract these typically unknown properties from the data. Lastly, we present a preliminary HXR directivity analysis of two flares, observed by the Fermi Gamma-ray Burst Monitor (GBM) and SolO/STIX, demonstrating the feasibility and challenges associated with such observations, and how HXR directivity can be extracted by comparison with the modelling presented here. △ Less

Submitted 29 January, 2024; originally announced January 2024.

Comments: Accepted for publication in ApJ (January 2024)

Continual Learning: Applications and the Road Forward

Authors: Eli Verwimp, Rahaf Aljundi, Shai Ben-David, Matthias Bethge, Andrea Cossu, Alexander Gepperth, Tyler L. Hayes, Eyke Hüllermeier, Christopher Kanan, Dhireesha Kudithipudi, Christoph H. Lampert, Martin Mundt, Razvan Pascanu, Adrian Popescu, Andreas S. Tolias, Joost van de Weijer, Bing Liu, Vincenzo Lomonaco, Tinne Tuytelaars, Gido M. van de Ven

Abstract: Continual learning is a subfield of machine learning, which aims to allow machine learning models to continuously learn on new data, by accumulating knowledge without forgetting what was learned in the past. In this work, we take a step back, and ask: "Why should one care about continual learning in the first place?". We set the stage by examining recent continual learning papers published at four… ▽ More Continual learning is a subfield of machine learning, which aims to allow machine learning models to continuously learn on new data, by accumulating knowledge without forgetting what was learned in the past. In this work, we take a step back, and ask: "Why should one care about continual learning in the first place?". We set the stage by examining recent continual learning papers published at four major machine learning conferences, and show that memory-constrained settings dominate the field. Then, we discuss five open problems in machine learning, and even though they might seem unrelated to continual learning at first sight, we show that continual learning will inevitably be part of their solution. These problems are model editing, personalization and specialization, on-device learning, faster (re-)training and reinforcement learning. Finally, by comparing the desiderata from these unsolved problems and the current assumptions in continual learning, we highlight and discuss four future directions for continual learning research. We hope that this work offers an interesting perspective on the future of continual learning, while displaying its potential value and the paths we have to pursue in order to make it successful. This work is the result of the many discussions the authors had at the Dagstuhl seminar on Deep Continual Learning, in March 2023. △ Less

Submitted 28 March, 2024; v1 submitted 20 November, 2023; originally announced November 2023.

Journal ref: Transactions on Machine Learning Research (TMLR), 2024

The eruption of a magnetic flux rope observed by \textit{Solar Orbiter} and \textit{Parker Solar Probe}

Authors: David M. Long, Lucie M. Green, Francesco Pecora, David H. Brooks, Hanna Strecker, David Orozco-Suárez, Laura A. Hayes, Emma E. Davies, Ute V. Amerstorfer, Marilena Mierla, David Lario, David Berghmans, Andrei N. Zhukov, Hannah T. Rüdisser

Abstract: Magnetic flux ropes are a key component of coronal mass ejections, forming the core of these eruptive phenomena. However, determining whether a flux rope is present prior to eruption onset and, if so, the rope's handedness and the number of turns that any helical field lines make is difficult without magnetic field modelling or in-situ detection of the flux rope. We present two distinct observatio… ▽ More Magnetic flux ropes are a key component of coronal mass ejections, forming the core of these eruptive phenomena. However, determining whether a flux rope is present prior to eruption onset and, if so, the rope's handedness and the number of turns that any helical field lines make is difficult without magnetic field modelling or in-situ detection of the flux rope. We present two distinct observations of plasma flows along a filament channel on 4 and 5 September 2022 made using the \textit{Solar Orbiter} spacecraft. Each plasma flow exhibited helical motions in a right-handed sense as the plasma moved from the source active region across the solar disk to the quiet Sun, suggesting that the magnetic configuration of the filament channel contains a flux rope with positive chirality and at least one turn. The length and velocity of the plasma flow increased from the first to the second observation, suggesting evolution of the flux rope, with the flux rope subsequently erupting within $\sim$5~hours of the second plasma flow. The erupting flux rope then passed over the \textit{Parker Solar Probe} spacecraft during its Encounter 13, enabling \textit{in-situ} diagnostics of the structure. Although complex and consistent with the flux rope erupting from underneath the heliospheric current sheet, the \textit{in-situ} measurements support the inference of a right-handed flux rope from remote-sensing observations. These observations provide a unique insight into the eruption and evolution of a magnetic flux rope near the Sun. △ Less

Submitted 28 August, 2023; originally announced August 2023.

Comments: 19 pages, 11 figures, accepted for publication in The Astrophysical Journal

Statistical analysis of the onset temperature of solar flares in 2010-2011

Authors: Douglas Félix da Silva, Li Hui, Paulo J. A. Simões, Adriana Valio, Joaquim C. E. R., Hugh S. Hudson, Paulo J. A. Simoes, Lyndsay Fletcher, Laura A. Hayes, Iain G. Hannah

Abstract: Understanding the physical processes that trigger solar flares is paramount to help with forecasting space weather and mitigating the effects on our technological infrastructure. A previously unknown phenomenon was recently identified in solar flares: the plasma temperature, derived from soft X-ray (SXR) data, at the onset of four flares, was revealed to be in the range 10-15 MK, without evidence… ▽ More Understanding the physical processes that trigger solar flares is paramount to help with forecasting space weather and mitigating the effects on our technological infrastructure. A previously unknown phenomenon was recently identified in solar flares: the plasma temperature, derived from soft X-ray (SXR) data, at the onset of four flares, was revealed to be in the range 10-15 MK, without evidence of gradual heating. To investigate how common the hot-onset phenomenon may be, we extend this investigation to solar flares of B1.2- X6.9 classes recorded by the X-ray Sensor (XRS) on-board the GOES-14 and GOES-15 satellites between 2010 and 2011. For this statistical study, we employed the same methodology as in recent work, where the pre-flare SXR flux of each flare is obtained manually, and the temperature and emission measure values are obtained by the flux ratio of the two GOES/XRS channels using the standard software. From 3224 events listed in the GOES flare catalog for 2010-2011, we have selected and analyzed 745 events for which the flare heliographic location was provided in the list, to investigate center-to-limb effects of the hot-onset phenomenon. Our results show that 559 out of 745 flares (75%) exhibit an onset temperature above 8.6 MK (the first quartile), with respective log10 of the emission measure values between 46.0 - 47.25 cm-3, indicating that small amounts of plasma are quickly heated to high temperatures. These results suggest that the hot-onset phenomenon is very common in solar flares. △ Less

Submitted 21 August, 2023; originally announced August 2023.

Comments: 6 pages,7 figures

A multiple spacecraft detection of the 2 April 2022 M-class flare and filament eruption during the first close Solar Orbiter perihelion

Authors: M. Janvier, S. Mzerguat, P. R. Young, É. Buchlin, A. Manou, G. Pelouze, D. M. Long, L. Green, A. Warmuth, F. Schuller, P. Démoulin, D. Calchetti, F. Kahil, L. Bellot Rubio, S. Parenti, S. Baccar, K. Barczynski, L. K. Harra, L. A. Hayes, W. T. Thompson, D. Müller, D. Baker, S. Yardley, D. Berghmans, C. Verbeeck , et al. (34 additional authors not shown)

Abstract: The Solar Orbiter mission completed its first remote-sensing observation windows in the spring of 2022. On 2/4/2022, an M-class flare followed by a filament eruption was seen both by the instruments on board the mission and from several observatories in Earth's orbit. The complexity of the observed features is compared with the predictions given by the standard flare model in 3D. We use the observ… ▽ More The Solar Orbiter mission completed its first remote-sensing observation windows in the spring of 2022. On 2/4/2022, an M-class flare followed by a filament eruption was seen both by the instruments on board the mission and from several observatories in Earth's orbit. The complexity of the observed features is compared with the predictions given by the standard flare model in 3D. We use the observations from a multi-view dataset, which includes EUV imaging to spectroscopy and magnetic field measurements. These data come from IRIS, SDO, Hinode, as well as several instruments on Solar Orbiter. Information given by SDO/HMI and Solar Orbiter PHI/HRT shows that a parasitic polarity emerging underneath the filament is responsible for bringing the flux rope to an unstable state. As the flux rope erupts, Hinode/EIS captures blue-shifted emission in the transition region and coronal lines in the northern leg of the flux rope prior to the flare peak. Solar Orbiter SPICE captures the whole region, complementing the Doppler diagnostics of the filament eruption. Analyses of the formation and evolution of a complex set of flare ribbons and loops show that the parasitic emerging bipole plays an important role in the evolution of the flaring region. While the analysed data are overall consistent with the standard flare model, the present particular magnetic configuration shows that surrounding magnetic activity such as nearby emergence needs to be taken into account to fully understand the processes at work. This filament eruption is the first to be covered from different angles by spectroscopic instruments, and provides an unprecedented diagnostic of the multi-thermal structures present before and during the flare. This dataset of an eruptive event showcases the capabilities of coordinated observations with the Solar Orbiter mission. △ Less

Submitted 5 July, 2023; originally announced July 2023.

Comments: Accepted for publication in the Astronomy & Astrophysics special edition "Solar Orbiter First Results (Nominal Mission Phase)" (23/05/2023)

Journal ref: A&A 677, A130 (2023)

The Focusing Optics X-ray Solar Imager (FOXSI)

Authors: Steven Christe, Meriem Alaoui, Joel Allred, Marina Battaglia, Wayne Baumgartner, Juan Camilo Buitrago-Casas, Amir Caspi, Bin Chen, Thomas Chen, Brian Dennis, James Drake, Lindsay Glesener, Iain Hannah, Laura A. Hayes, Hugh Hudson, Andrew Inglis, Jack Ireland, James Klimchuk, Adam Kowalski, Säm Krucker, Anna Maria Massone, Sophie Musset, Michele Piana, Daniel Ryan, Albert Y. Shih , et al. (4 additional authors not shown)

Abstract: FOXSI is a direct-imaging, hard X-ray (HXR) telescope optimized for solar flare observations. It detects hot plasma and energetic electrons in and near energy release sites in the solar corona via bremsstrahlung emission, measuring both spatial structure and particle energy distributions. It provides two orders of magnitude faster imaging spectroscopy than previously available, probing physically… ▽ More FOXSI is a direct-imaging, hard X-ray (HXR) telescope optimized for solar flare observations. It detects hot plasma and energetic electrons in and near energy release sites in the solar corona via bremsstrahlung emission, measuring both spatial structure and particle energy distributions. It provides two orders of magnitude faster imaging spectroscopy than previously available, probing physically relevant timescales (<1s) never before accessible to address fundamental questions of energy release and efficient particle acceleration that have importance far beyond their solar application (e.g., planetary magnetospheres, flaring stars, accretion disks). FOXSI measures not only the bright chromospheric X-ray emission where electrons lose most of their energy, but also simultaneous emission from electrons as they are accelerated in the corona and propagate along magnetic field lines. FOXSI detects emission from high in the tenuous corona, where previous instruments have been blinded by nearby bright features and will fully characterizes the accelerated electrons and hottest plasmas as they evolve in energy, space, and time to solve the mystery of how impulsive energy release leads to solar eruptions, the primary drivers of space weather at Earth, and how those eruptions are energized and evolve. △ Less

Submitted 26 June, 2023; originally announced June 2023.

Comments: White paper submitted to the Decadal Survey for Solar and Space Physics (Heliophysics) 2024-2033; 14 pages, 4 figures, 1 table

Journal ref: Bulletin of the AAS, Vol. 55, Issue 3, Whitepaper #065 (13pp); 2023 July 31

Fundamentals of impulsive energy release in the corona

Authors: Albert Y. Shih, Lindsay Glesener, Säm Krucker, Silvina Guidoni, Steven Christe, Katharine K. Reeves, Szymon Gburek, Amir Caspi, Meriem Alaoui, Joel Allred, Marina Battaglia, Wayne Baumgartner, Brian Dennis, James Drake, Keith Goetz, Leon Golub, Iain Hannah, Laura Hayes, Gordon Holman, Andrew Inglis, Jack Ireland, Graham Kerr, James Klimchuk, David McKenzie, Christopher S. Moore , et al. (8 additional authors not shown)

Abstract: It is essential that there be coordinated and co-optimized observations in X-rays, gamma-rays, and EUV during the peak of solar cycle 26 (~2036) to significantly advance our understanding of impulsive energy release in the corona. The open questions include: What are the physical origins of space-weather events? How are particles accelerated at the Sun? How is impulsively released energy transport… ▽ More It is essential that there be coordinated and co-optimized observations in X-rays, gamma-rays, and EUV during the peak of solar cycle 26 (~2036) to significantly advance our understanding of impulsive energy release in the corona. The open questions include: What are the physical origins of space-weather events? How are particles accelerated at the Sun? How is impulsively released energy transported throughout the solar atmosphere? How is the solar corona heated? Many of the processes involved in triggering, driving, and sustaining solar eruptive events -- including magnetic reconnection, particle acceleration, plasma heating, and energy transport in magnetized plasmas -- also play important roles in phenomena throughout the Universe. This set of observations can be achieved through a single flagship mission or, with foreplanning, through a combination of major missions (e.g., the previously proposed FIERCE mission concept). △ Less

Submitted 20 June, 2023; originally announced June 2023.

Comments: White paper submitted to the Decadal Survey for Solar and Space Physics (Heliophysics) 2024-2033; 5 pages, 1 figure

Journal ref: Bulletin of the AAS, Vol. 55, Issue 3, Whitepaper #364 (5pp); 2023 July 31

The need for focused, hard X-ray investigations of the Sun

Authors: Lindsay Glesener, Albert Y. Shih, Amir Caspi, Ryan Milligan, Hugh Hudson, Mitsuo Oka, Juan Camilo Buitrago-Casas, Fan Guo, Dan Ryan, Eduard Kontar, Astrid Veronig, Laura A. Hayes, Andrew Inglis, Leon Golub, Nicole Vilmer, Dale Gary, Hamish Reid, Iain Hannah, Graham S. Kerr, Katharine K. Reeves, Joel Allred, Silvina Guidoni, Sijie Yu, Steven Christe, Sophie Musset , et al. (24 additional authors not shown)

Abstract: Understanding the nature of energetic particles in the solar atmosphere is one of the most important outstanding problems in heliophysics. Flare-accelerated particles compose a huge fraction of the flare energy budget; they have large influences on how events develop; they are an important source of high-energy particles found in the heliosphere; and they are the single most important corollary to… ▽ More Understanding the nature of energetic particles in the solar atmosphere is one of the most important outstanding problems in heliophysics. Flare-accelerated particles compose a huge fraction of the flare energy budget; they have large influences on how events develop; they are an important source of high-energy particles found in the heliosphere; and they are the single most important corollary to other areas of high-energy astrophysics. Despite the importance of this area of study, this topic has in the past decade received only a small fraction of the resources necessary for a full investigation. For example, NASA has selected no new Explorer-class instrument in the past two decades that is capable of examining this topic. The advances that are currently being made in understanding flare-accelerated electrons are largely undertaken with data from EOVSA (NSF), STIX (ESA), and NuSTAR (NASA Astrophysics). This is despite the inclusion in the previous Heliophysics decadal survey of the FOXSI concept as part of the SEE2020 mission, and also despite NASA's having invested heavily in readying the technology for such an instrument via four flights of the FOXSI sounding rocket experiment. Due to that investment, the instrumentation stands ready to implement a hard X-ray mission to investigate flare-accelerated electrons. This white paper describes the scientific motivation for why this venture should be undertaken soon. △ Less

Submitted 8 June, 2023; originally announced June 2023.

Comments: White paper submitted to the Decadal Survey for Solar and Space Physics (Heliophysics) 2024-2033; 15 pages, 5 figures

Journal ref: Bulletin of the AAS, Vol. 55, Issue 3, Whitepaper #129 (14pp); 2023 July 31

Prevalence of non-stationarity in quasi-periodic pulsations (QPPs) associated with M- and X-class solar flares

Authors: Tishtrya Mehta, Anne-Marie Broomhall, Laura Hayes

Abstract: Quasi-periodic pulsations (QPPs) are frequently observed in solar and stellar flare emission, with recent studies suggesting that an increasing instantaneous period is a common characteristic of QPPs. Determining the prevalence of non-stationarity in QPPs contributes to a better understanding of which mechanisms are responsible in QPP generation. We obtain the rate of period evolution from QPPs in… ▽ More Quasi-periodic pulsations (QPPs) are frequently observed in solar and stellar flare emission, with recent studies suggesting that an increasing instantaneous period is a common characteristic of QPPs. Determining the prevalence of non-stationarity in QPPs contributes to a better understanding of which mechanisms are responsible in QPP generation. We obtain the rate of period evolution from QPPs in 98 M- and X-class flares from Solar Cycle 24 with average periods between 8-130s and investigate the prevalence of QPP non-stationarity. We also investigate whether the presence of a Coronal Mass Ejection (CME) impacts the period evolution of QPPs. We analyse soft X-ray lightcurves obtained from GOES' X-Ray Sensor (XRS) and assess the dominant periods in the impulsive and decay phases of the flares using the Fast Fourier Transform. We relate the rate of period evolution to flare duration, peak flare energy, and average QPP period. We find evidence of non-stationarity in 81% of the flares assessed, with most QPPs exhibiting a period evolution of less than 10s between the impulsive and decay phases, of which 66% exhibited an apparent period growth and 14% showed an apparent period shrinkage. We find a positive correlation between the absolute magnitude of period evolution and the duration of the flare and no correlation between the period evolution of the QPPs and flare energy or CME presence. Furthermore, we conclude that non-stationarity is common in solar QPPs and must be accounted for in flare analysis. △ Less

Submitted 31 May, 2023; originally announced May 2023.

Comments: 10 pages, 11 figures, 1 table

The SunPy Project: An Interoperable Ecosystem for Solar Data Analysis

Authors: The SunPy Community, Will Barnes, Steven Christe, Nabil Freij, Laura Hayes, David Stansby, Jack Ireland, Stuart Mumford, Daniel Ryan, Albert Shih

Abstract: The SunPy Project is a community of scientists and software developers creating an ecosystem of Python packages for solar physics. The project includes the sunpy core package as well as a set of affiliated packages. The sunpy core package provides general purpose tools to access data from different providers, read image and time series data, and transform between commonly used coordinate systems.… ▽ More The SunPy Project is a community of scientists and software developers creating an ecosystem of Python packages for solar physics. The project includes the sunpy core package as well as a set of affiliated packages. The sunpy core package provides general purpose tools to access data from different providers, read image and time series data, and transform between commonly used coordinate systems. Affiliated packages perform more specialized tasks that do not fall within the more general scope of the sunpy core package. In this article, we give a high-level overview of the SunPy Project, how it is broader than the sunpy core package, and how the project curates and fosters the affiliated package system. We demonstrate how components of the SunPy ecosystem, including sunpy and several affiliated packages, work together to enable multi-instrument data analysis workflows. We also describe members of the SunPy Project and how the project interacts with the wider solar physics and scientific Python communities. Finally, we discuss the future direction and priorities of the SunPy Project. △ Less

Submitted 19 April, 2023; originally announced April 2023.

Comments: 15 pages, 1 figure, published in Frontiers

Journal ref: Frontiers in Astronomy and Space Sciences 10 (2023): 29

How Efficient Are Today's Continual Learning Algorithms?

Authors: Md Yousuf Harun, Jhair Gallardo, Tyler L. Hayes, Christopher Kanan

Abstract: Supervised Continual learning involves updating a deep neural network (DNN) from an ever-growing stream of labeled data. While most work has focused on overcoming catastrophic forgetting, one of the major motivations behind continual learning is being able to efficiently update a network with new information, rather than retraining from scratch on the training dataset as it grows over time. Despit… ▽ More Supervised Continual learning involves updating a deep neural network (DNN) from an ever-growing stream of labeled data. While most work has focused on overcoming catastrophic forgetting, one of the major motivations behind continual learning is being able to efficiently update a network with new information, rather than retraining from scratch on the training dataset as it grows over time. Despite recent continual learning methods largely solving the catastrophic forgetting problem, there has been little attention paid to the efficiency of these algorithms. Here, we study recent methods for incremental class learning and illustrate that many are highly inefficient in terms of compute, memory, and storage. Some methods even require more compute than training from scratch! We argue that for continual learning to have real-world applicability, the research community cannot ignore the resources used by these algorithms. There is more to continual learning than mitigating catastrophic forgetting. △ Less

Submitted 3 April, 2023; v1 submitted 29 March, 2023; originally announced March 2023.

Comments: To appear in the IEEE Conference on Computer Vision and Pattern Recognition Workshop (CVPR-W) on Continual Learning in Computer Vision (CLVision) 2023

SIESTA: Efficient Online Continual Learning with Sleep

Authors: Md Yousuf Harun, Jhair Gallardo, Tyler L. Hayes, Ronald Kemker, Christopher Kanan

Abstract: In supervised continual learning, a deep neural network (DNN) is updated with an ever-growing data stream. Unlike the offline setting where data is shuffled, we cannot make any distributional assumptions about the data stream. Ideally, only one pass through the dataset is needed for computational efficiency. However, existing methods are inadequate and make many assumptions that cannot be made for… ▽ More In supervised continual learning, a deep neural network (DNN) is updated with an ever-growing data stream. Unlike the offline setting where data is shuffled, we cannot make any distributional assumptions about the data stream. Ideally, only one pass through the dataset is needed for computational efficiency. However, existing methods are inadequate and make many assumptions that cannot be made for real-world applications, while simultaneously failing to improve computational efficiency. In this paper, we propose a novel continual learning method, SIESTA based on wake/sleep framework for training, which is well aligned to the needs of on-device learning. The major goal of SIESTA is to advance compute efficient continual learning so that DNNs can be updated efficiently using far less time and energy. The principal innovations of SIESTA are: 1) rapid online updates using a rehearsal-free, backpropagation-free, and data-driven network update rule during its wake phase, and 2) expedited memory consolidation using a compute-restricted rehearsal policy during its sleep phase. For memory efficiency, SIESTA adapts latent rehearsal using memory indexing from REMIND. Compared to REMIND and prior arts, SIESTA is far more computationally efficient, enabling continual learning on ImageNet-1K in under 2 hours on a single GPU; moreover, in the augmentation-free setting it matches the performance of the offline learner, a milestone critical to driving adoption of continual learning in real-world applications. △ Less

Submitted 2 November, 2023; v1 submitted 19 March, 2023; originally announced March 2023.

Comments: Accepted to TMLR 2023

Quasi-periodic pulsations in solar flares: a key diagnostic of energy release on the Sun

Authors: Andrew Inglis, Laura Hayes, Silvina Guidoni, James McLaughlin, Valery M. Nakariakov, Tom Van Doorsselaere, Ernesto Zurbriggen, Mariana Cécere, Marie Dominique, Jeff Reep, Ivan Zimovets, Elena Kupriyanova, Dmitrii Kolotkov, Bo Li, Marina Battaglia, Christopher Moore, Hannah Collier, Crisel Suarez, Tishtrya Mehta, Trevor Knuth, Thomas Y. Chen

Abstract: Solar flares are among the most powerful and disruptive events in our solar system, however the physical mechanisms driving and transporting this energetic release are not fully understood. An important signature associated with flare energy release is highly variable emission on timescales of sub-seconds to minutes which often exhibit oscillatory behaviour, features collectively known as quasi-pe… ▽ More Solar flares are among the most powerful and disruptive events in our solar system, however the physical mechanisms driving and transporting this energetic release are not fully understood. An important signature associated with flare energy release is highly variable emission on timescales of sub-seconds to minutes which often exhibit oscillatory behaviour, features collectively known as quasi-periodic pulsations (QPPs). To fully identify the driving mechanism of QPPs, exploit their potential as a diagnostic tool, and incorporate them into our understanding of solar and stellar flares, new observational capabilities and initiatives are required. There is a clear community need for flare-focused, rapid cadence, high resolution, multi-wavelength imaging of the Sun, with high enough sensitivity and dynamic range to observe small fluctuations in intensity in the presence of a large overall intensity. Furthermore, multidisciplinary funding and initiatives are required to narrow the gap between numerical models and observations. QPPs are direct signatures of the physics occurring in flare magnetic reconnection and energy release sites and hence are critical to include in a unified flare model. Despite significant modelling and theoretical work, no single mechanism or model can fully explain the presence of QPPs in flares. Moreover, it is also likely that QPPs fall into different categories that are produced by different mechanisms. At present we have insufficient information to observationally distinguish between mechanisms. The motivation to understand QPPs is strengthened by the geo-effectiveness of flares on the Earth's ionosphere, and by the fact that stellar flares exhibit similar QPP signatures. QPPs present a golden opportunity to better understand flare physics and exploit the solar-stellary analogy, benefiting both astrophysics, heliophysics, and the solar-terrestrial connection. △ Less

Submitted 14 March, 2023; v1 submitted 22 February, 2023; originally announced February 2023.

Comments: White paper submitted to the Decadal Survey for Solar and Space Physics (Heliophysics) 2024-2033. 8 pages, 3 figures, 1 table

Characterising fast-time variations in the hard X-ray time profiles of solar flares using Solar Orbiter's STIX

Authors: Hannah Collier, Laura A. Hayes, Andrea F. Battaglia, Louise K. Harra, Säm Krucker

Abstract: Aims: The aim of this work is to develop a method to systematically detect and characterise fast-time variations ($\gtrsim 1$s) in the non-thermal hard X-ray (HXR) time profiles of solar flares using high-resolution data from Solar Orbiter's Spectrometer/Telescope for Imaging X-rays (STIX). Methods: The HXR time profiles were smoothed using Gaussian Process (GP) regression. The time profiles wer… ▽ More Aims: The aim of this work is to develop a method to systematically detect and characterise fast-time variations ($\gtrsim 1$s) in the non-thermal hard X-ray (HXR) time profiles of solar flares using high-resolution data from Solar Orbiter's Spectrometer/Telescope for Imaging X-rays (STIX). Methods: The HXR time profiles were smoothed using Gaussian Process (GP) regression. The time profiles were then fitted with a linear combination of Gaussians to decompose the time profile. From the Gaussian decomposition, key characteristics such as the periodicity, full width at half maximum (FWHM), time evolution, and amplitude can be derived. Results: We present the outcome of applying this method to four M and X GOES-class flares from the first year of Solar Orbiter science operations. The HXR time profiles of these flares were decomposed into individual Gaussians and their periods were derived. The quality of fit is quantified by the standard deviation of the residuals (difference between observed and fitted curve, normalised by the error on the observed data), for which we obtain $\leq 1.8$ for all flares presented. In this work, the first detection of fast-time variations with Solar Orbiter's STIX instrument has been made on timescales across the range of 4-128s. Conclusions: A new method for identifying and characterising fast-time variations in the non-thermal HXR profiles of solar flares has been developed, in which the time profiles are fit with a linear combination of Gaussian bursts. The opportunity to study time variations in flares has greatly improved with the new observations from STIX on Solar Orbiter. △ Less

Submitted 19 January, 2023; originally announced January 2023.

Comments: 12 pages, 10 figures, A&A journal

Journal ref: A&A 671, A79 (2023)

Science Platforms for Heliophysics Data Analysis

Authors: Monica G. Bobra, Will T. Barnes, Thomas Y. Chen, Mark C. M. Cheung, Laura A. Hayes, Jack Ireland, Miho Janvier, Michael S. F. Kirk, James P. Mason, Stuart J. Mumford, Paul J. Wright

Abstract: We recommend that NASA maintain and fund science platforms that enable interactive and scalable data analysis in order to maximize the scientific return of data collected from space-based instruments. We recommend that NASA maintain and fund science platforms that enable interactive and scalable data analysis in order to maximize the scientific return of data collected from space-based instruments. △ Less

Submitted 2 January, 2023; originally announced January 2023.

Comments: Heliophysics 2050 White Paper

System Design for an Integrated Lifelong Reinforcement Learning Agent for Real-Time Strategy Games

Authors: Indranil Sur, Zachary Daniels, Abrar Rahman, Kamil Faber, Gianmarco J. Gallardo, Tyler L. Hayes, Cameron E. Taylor, Mustafa Burak Gurbuz, James Smith, Sahana Joshi, Nathalie Japkowicz, Michael Baron, Zsolt Kira, Christopher Kanan, Roberto Corizzo, Ajay Divakaran, Michael Piacentino, Jesse Hostetler, Aswin Raghavan

Abstract: As Artificial and Robotic Systems are increasingly deployed and relied upon for real-world applications, it is important that they exhibit the ability to continually learn and adapt in dynamically-changing environments, becoming Lifelong Learning Machines. Continual/lifelong learning (LL) involves minimizing catastrophic forgetting of old tasks while maximizing a model's capability to learn new ta… ▽ More As Artificial and Robotic Systems are increasingly deployed and relied upon for real-world applications, it is important that they exhibit the ability to continually learn and adapt in dynamically-changing environments, becoming Lifelong Learning Machines. Continual/lifelong learning (LL) involves minimizing catastrophic forgetting of old tasks while maximizing a model's capability to learn new tasks. This paper addresses the challenging lifelong reinforcement learning (L2RL) setting. Pushing the state-of-the-art forward in L2RL and making L2RL useful for practical applications requires more than develo** individual L2RL algorithms; it requires making progress at the systems-level, especially research into the non-trivial problem of how to integrate multiple L2RL algorithms into a common framework. In this paper, we introduce the Lifelong Reinforcement Learning Components Framework (L2RLCF), which standardizes L2RL systems and assimilates different continual learning components (each addressing different aspects of the lifelong learning problem) into a unified system. As an instantiation of L2RLCF, we develop a standard API allowing easy integration of novel lifelong learning components. We describe a case study that demonstrates how multiple independently-developed LL components can be integrated into a single realized system. We also introduce an evaluation environment in order to measure the effect of combining various system components. Our evaluation environment employs different LL scenarios (sequences of tasks) consisting of Starcraft-2 minigames and allows for the fair, comprehensive, and quantitative comparison of different combinations of components within a challenging common evaluation environment. △ Less

Submitted 8 December, 2022; originally announced December 2022.

Comments: The Second International Conference on AIML Systems, October 12--15, 2022, Bangalore, India

A Significant Sudden Ionospheric Disturbance associated with Gamma-Ray Burst GRB 221009A

Authors: Laura A. Hayes, Peter T. Gallagher

Abstract: We report the detection of a significant ionospheric disturbance in the D-region of Earth's ionosphere which was associated with the massive gamma-ray burst GRB 221009A that occurred on October 9 2022. We identified the disturbance over northern Europe - a result of the increased ionisation by X- and gamma-ray emission from the GRB - using very low frequency (VLF) radio waves as a probe of the D-r… ▽ More We report the detection of a significant ionospheric disturbance in the D-region of Earth's ionosphere which was associated with the massive gamma-ray burst GRB 221009A that occurred on October 9 2022. We identified the disturbance over northern Europe - a result of the increased ionisation by X- and gamma-ray emission from the GRB - using very low frequency (VLF) radio waves as a probe of the D-region. These observations demonstrate that an extra-galactic GRB can have a significant impact on the terrestrial ionosphere and illustrates that the Earth's ionosphere can be used as a giant X- and gamma-ray detector. Indeed, these observations may provide insights into the impacts of GRBs on the ionospheres of planets in our solar system and beyond. △ Less

Submitted 27 October, 2022; originally announced October 2022.

Comments: 3 pages, 1 figure

Journal ref: 2022 Res. Notes AAS 6 222

The Spectrometer Telescope for Imaging X-rays (STIX) on Solar Orbiter

Authors: Laura A. Hayes, Sophie Musset, Daniel M üller, S äm Krucker

Abstract: The Spectrometer/Telescope for Imaging X-rays (STIX) is one of the 10 instruments on-board the scientific payload of ESA's Solar Orbiter mission. STIX provides hard X-ray imaging spectroscopy in the 4-150~keV energy range, observing hard X-ray bremsstrahlung emission from the Sun. These observations provide diagnostics of the hottest thermal plasmas ($>$10~MK) and information on the non-thermal en… ▽ More The Spectrometer/Telescope for Imaging X-rays (STIX) is one of the 10 instruments on-board the scientific payload of ESA's Solar Orbiter mission. STIX provides hard X-ray imaging spectroscopy in the 4-150~keV energy range, observing hard X-ray bremsstrahlung emission from the Sun. These observations provide diagnostics of the hottest thermal plasmas ($>$10~MK) and information on the non-thermal energetic electrons accelerated above 10~keV during solar flares. STIX has a spectral resolution of 1~keV, and employs the use of in-direct bi-grid Fourier imaging to spatially locate hard X-ray emission. Given that STIX provides critical information about accelerated electrons at the Sun through hard X-ray diagnostics, it is a powerful contribution to the Solar Orbiter suite and has a significant role to explore the dynamics of solar inputs to the heliosphere. This chapter describes the STIX instrument, its design, objectives, first observations and outlines the new perspectives STIX provides over the mission lifetime of Solar Orbiter. △ Less

Submitted 5 July, 2022; originally announced July 2022.

Comments: 18 pages, 11 figures, Book Chapter for Handbook of X-ray and Gamma-ray Astrophysics

Prominence eruption observed in He II 304 Å up to $>6 R_\sun$ by EUI/FSI aboard Solar Orbiter

Authors: M. Mierla, A. N. Zhukov, D. Berghmans, S. Parenti, F. Auchere, P. Heinzel, D. B. Seaton, E. Palmerio, S. Jejcic, J. Janssens, E. Kraaikamp, B. Nicula, D. M. Long, L. A. Hayes, I. C. Jebaraj, D. -C. Talpeanu, E. D'Huys, L. Dolla, S. Gissot, J. Magdalenic, L. Rodriguez, S. Shestov, K. Stegen, C. Verbeeck, C. Sasso , et al. (2 additional authors not shown)

Abstract: We report observations of a unique, large prominence eruption that was observed in the He II 304 Å passband of the the Extreme Ultraviolet Imager/Full Sun Imager telescope aboard Solar Orbiter on 15-16 February 2022. Observations from several vantage points (Solar Orbiter, the Solar-Terrestrial Relations Observatory, the Solar and Heliospheric Observatory, and Earth-orbiting satellites) were used… ▽ More We report observations of a unique, large prominence eruption that was observed in the He II 304 Å passband of the the Extreme Ultraviolet Imager/Full Sun Imager telescope aboard Solar Orbiter on 15-16 February 2022. Observations from several vantage points (Solar Orbiter, the Solar-Terrestrial Relations Observatory, the Solar and Heliospheric Observatory, and Earth-orbiting satellites) were used to measure the kinematics of the erupting prominence and the associated coronal mass ejection. Three-dimensional reconstruction was used to calculate the deprojected positions and speeds of different parts of the prominence. Observations in several passbands allowed us to analyse the radiative properties of the erupting prominence. The leading parts of the erupting prominence and the leading edge of the corresponding coronal mass ejection propagate at speeds of around 1700 km/s and 2200 km/s, respectively, while the trailing parts of the prominence are significantly slower (around 500 km/s). Parts of the prominence are tracked up to heights of over $6 R_\sun$. The He II emission is probably produced via collisional excitation rather than scattering. Surprisingly, the brightness of a trailing feature increases with height. The reported prominence is the first observed in He II 304 Å emission at such a great height (above 6 $R_\sun$). △ Less

Submitted 30 May, 2022; originally announced May 2022.

Journal ref: A&A 662, L5 (2022)

Online Continual Learning for Embedded Devices

Authors: Tyler L. Hayes, Christopher Kanan

Abstract: Real-time on-device continual learning is needed for new applications such as home robots, user personalization on smartphones, and augmented/virtual reality headsets. However, this setting poses unique challenges: embedded devices have limited memory and compute capacity and conventional machine learning models suffer from catastrophic forgetting when updated on non-stationary data streams. While… ▽ More Real-time on-device continual learning is needed for new applications such as home robots, user personalization on smartphones, and augmented/virtual reality headsets. However, this setting poses unique challenges: embedded devices have limited memory and compute capacity and conventional machine learning models suffer from catastrophic forgetting when updated on non-stationary data streams. While several online continual learning models have been developed, their effectiveness for embedded applications has not been rigorously studied. In this paper, we first identify criteria that online continual learners must meet to effectively perform real-time, on-device learning. We then study the efficacy of several online continual learning methods when used with mobile neural networks. We measure their performance, memory usage, compute requirements, and ability to generalize to out-of-domain inputs. △ Less

Submitted 15 July, 2022; v1 submitted 20 March, 2022; originally announced March 2022.

Comments: To appear in the Conference on Lifelong Learning Agents (CoLLAs-2022)

Can I see an Example? Active Learning the Long Tail of Attributes and Relations

Authors: Tyler L. Hayes, Maximilian Nickel, Christopher Kanan, Ludovic Denoyer, Arthur Szlam

Abstract: There has been significant progress in creating machine learning models that identify objects in scenes along with their associated attributes and relationships; however, there is a large gap between the best models and human capabilities. One of the major reasons for this gap is the difficulty in collecting sufficient amounts of annotated relations and attributes for training these systems. While… ▽ More There has been significant progress in creating machine learning models that identify objects in scenes along with their associated attributes and relationships; however, there is a large gap between the best models and human capabilities. One of the major reasons for this gap is the difficulty in collecting sufficient amounts of annotated relations and attributes for training these systems. While some attributes and relations are abundant, the distribution in the natural world and existing datasets is long tailed. In this paper, we address this problem by introducing a novel incremental active learning framework that asks for attributes and relations in visual scenes. While conventional active learning methods ask for labels of specific examples, we flip this framing to allow agents to ask for examples from specific categories. Using this framing, we introduce an active sampling method that asks for examples from the tail of the data distribution and show that it outperforms classical active learning methods on Visual Genome. △ Less

Submitted 7 October, 2022; v1 submitted 11 March, 2022; originally announced March 2022.

Comments: To appear in the British Machine Vision Conference (BMVC-2022)

Solar Flare Effects on the Earth's Lower Ionosphere

Authors: Laura A. Hayes, Oscar S. D. O'Hara, Sophie A. Murray, Peter T. Gallagher

Abstract: Solar flares significantly impact the conditions of the Earth's ionosphere. In particular, the sudden increase in X-ray flux during a flare penetrates down to the lowest-lying D-region and dominates ionization at these altitudes (60-100 km). Measurements of very low frequency (VLF: 3-30kHz) radio waves that reflect at D-region altitudes provide a unique remote-sensing probe to investigate the D-re… ▽ More Solar flares significantly impact the conditions of the Earth's ionosphere. In particular, the sudden increase in X-ray flux during a flare penetrates down to the lowest-lying D-region and dominates ionization at these altitudes (60-100 km). Measurements of very low frequency (VLF: 3-30kHz) radio waves that reflect at D-region altitudes provide a unique remote-sensing probe to investigate the D-region response to solar flare emissions. Here, using a combination of VLF amplitude measurements at 24kHz together with X-ray observations from the Geostationary Operational Environment Satellite (GOES) X-ray sensor, we present a large-scale statistical study of 334 solar flare events and their impacts on the D-region over the past solar cycle. Focusing on both GOES broadband X-ray channels, we investigate how the flare peak fluxes and position on the solar disk dictate an ionospheric response and extend this to investigate the characteristic time delay between incident X-ray flux and the D-region response. We show that the VLF amplitude linearly correlates with both the 1-8 A and 0.5-4 A channels, with correlation coefficients of 0.80 and 0.79, respectively. Unlike higher altitude ionospheric regions for which the location of the flare on the solar disk affects the ionospheric response, we find that the D-region response to solar flares does not depend on the flare location. By comparing the time delays between the peak X-ray fluxes in both GOES channels and VLF amplitudes, we find that there is an important difference between the D-region response and the X-ray spectral band. We also demonstrate for several flare events that show a negative time delay, the peak VLF amplitude matches with the impulsive 25-50 keV hard X-ray fluxes measured by the Ramaty High Energy Solar Spectroscopic Imager (RHESSI). △ Less

Submitted 14 September, 2021; originally announced September 2021.

Comments: 19 pages, 8 figures, accepted in Solar Physics

Disentangling Transfer and Interference in Multi-Domain Learning

Authors: Yipeng Zhang, Tyler L. Hayes, Christopher Kanan

Abstract: Humans are incredibly good at transferring knowledge from one domain to another, enabling rapid learning of new tasks. Likewise, transfer learning has enabled enormous success in many computer vision problems using pretraining. However, the benefits of transfer in multi-domain learning, where a network learns multiple tasks defined by different datasets, has not been adequately studied. Learning m… ▽ More Humans are incredibly good at transferring knowledge from one domain to another, enabling rapid learning of new tasks. Likewise, transfer learning has enabled enormous success in many computer vision problems using pretraining. However, the benefits of transfer in multi-domain learning, where a network learns multiple tasks defined by different datasets, has not been adequately studied. Learning multiple domains could be beneficial, or these domains could interfere with each other given limited network capacity. Understanding how deep neural networks of varied capacity facilitate transfer across inputs from different distributions is a critical step towards open world learning. In this work, we decipher the conditions where interference and knowledge transfer occur in multi-domain learning. We propose new metrics disentangling interference and transfer, set up experimental protocols, and examine the roles of network capacity, task grou**, and dynamic loss weighting in reducing interference and facilitating transfer. △ Less

Submitted 14 January, 2022; v1 submitted 1 July, 2021; originally announced July 2021.

Comments: AAAI 2022 PracticalDL Workshop

Replay in Deep Learning: Current Approaches and Missing Biological Elements

Authors: Tyler L. Hayes, Giri P. Krishnan, Maxim Bazhenov, Hava T. Siegelmann, Terrence J. Sejnowski, Christopher Kanan

Abstract: Replay is the reactivation of one or more neural patterns, which are similar to the activation patterns experienced during past waking experiences. Replay was first observed in biological neural networks during sleep, and it is now thought to play a critical role in memory formation, retrieval, and consolidation. Replay-like mechanisms have been incorporated into deep artificial neural networks th… ▽ More Replay is the reactivation of one or more neural patterns, which are similar to the activation patterns experienced during past waking experiences. Replay was first observed in biological neural networks during sleep, and it is now thought to play a critical role in memory formation, retrieval, and consolidation. Replay-like mechanisms have been incorporated into deep artificial neural networks that learn over time to avoid catastrophic forgetting of previous knowledge. Replay algorithms have been successfully used in a wide range of deep learning methods within supervised, unsupervised, and reinforcement learning paradigms. In this paper, we provide the first comprehensive comparison between replay in the mammalian brain and replay in artificial neural networks. We identify multiple aspects of biological replay that are missing in deep learning systems and hypothesize how they could be utilized to improve artificial neural networks. △ Less

Submitted 28 May, 2021; v1 submitted 1 April, 2021; originally announced April 2021.

Comments: Accepted for publication in the MIT Press journal of Neural Computation

Avalanche: an End-to-End Library for Continual Learning

Authors: Vincenzo Lomonaco, Lorenzo Pellegrini, Andrea Cossu, Antonio Carta, Gabriele Graffieti, Tyler L. Hayes, Matthias De Lange, Marc Masana, Jary Pomponi, Gido van de Ven, Martin Mundt, Qi She, Keiland Cooper, Jeremy Forest, Eden Belouadah, Simone Calderara, German I. Parisi, Fabio Cuzzolin, Andreas Tolias, Simone Scardapane, Luca Antiga, Subutai Amhad, Adrian Popescu, Christopher Kanan, Joost van de Weijer , et al. (3 additional authors not shown)

Abstract: Learning continually from non-stationary data streams is a long-standing goal and a challenging problem in machine learning. Recently, we have witnessed a renewed and fast-growing interest in continual learning, especially within the deep learning community. However, algorithmic solutions are often difficult to re-implement, evaluate and port across different settings, where even results on standa… ▽ More Learning continually from non-stationary data streams is a long-standing goal and a challenging problem in machine learning. Recently, we have witnessed a renewed and fast-growing interest in continual learning, especially within the deep learning community. However, algorithmic solutions are often difficult to re-implement, evaluate and port across different settings, where even results on standard benchmarks are hard to reproduce. In this work, we propose Avalanche, an open-source end-to-end library for continual learning research based on PyTorch. Avalanche is designed to provide a shared and collaborative codebase for fast prototy**, training, and reproducible evaluation of continual learning algorithms. △ Less

Submitted 1 April, 2021; originally announced April 2021.

Comments: Official Website: https://avalanche.continualai.org

Self-Supervised Training Enhances Online Continual Learning

Authors: Jhair Gallardo, Tyler L. Hayes, Christopher Kanan

Abstract: In continual learning, a system must incrementally learn from a non-stationary data stream without catastrophic forgetting. Recently, multiple methods have been devised for incrementally learning classes on large-scale image classification tasks, such as ImageNet. State-of-the-art continual learning methods use an initial supervised pre-training phase, in which the first 10% - 50% of the classes i… ▽ More In continual learning, a system must incrementally learn from a non-stationary data stream without catastrophic forgetting. Recently, multiple methods have been devised for incrementally learning classes on large-scale image classification tasks, such as ImageNet. State-of-the-art continual learning methods use an initial supervised pre-training phase, in which the first 10% - 50% of the classes in a dataset are used to learn representations in an offline manner before continual learning of new classes begins. We hypothesize that self-supervised pre-training could yield features that generalize better than supervised learning, especially when the number of samples used for pre-training is small. We test this hypothesis using the self-supervised MoCo-V2, Barlow Twins, and SwAV algorithms. On ImageNet, we find that these methods outperform supervised pre-training considerably for online continual learning, and the gains are larger when fewer samples are available. Our findings are consistent across three online continual learning algorithms. Our best system achieves a 14.95% relative increase in top-1 accuracy on class incremental ImageNet over the prior state of the art for online continual learning. △ Less

Submitted 22 October, 2021; v1 submitted 25 March, 2021; originally announced March 2021.

Comments: Accepted to BMVC-2021

Selective Replay Enhances Learning in Online Continual Analogical Reasoning

Authors: Tyler L. Hayes, Christopher Kanan

Abstract: In continual learning, a system learns from non-stationary data streams or batches without catastrophic forgetting. While this problem has been heavily studied in supervised image classification and reinforcement learning, continual learning in neural networks designed for abstract reasoning has not yet been studied. Here, we study continual learning of analogical reasoning. Analogical reasoning t… ▽ More In continual learning, a system learns from non-stationary data streams or batches without catastrophic forgetting. While this problem has been heavily studied in supervised image classification and reinforcement learning, continual learning in neural networks designed for abstract reasoning has not yet been studied. Here, we study continual learning of analogical reasoning. Analogical reasoning tests such as Raven's Progressive Matrices (RPMs) are commonly used to measure non-verbal abstract reasoning in humans, and recently offline neural networks for the RPM problem have been proposed. In this paper, we establish experimental baselines, protocols, and forward and backward transfer metrics to evaluate continual learners on RPMs. We employ experience replay to mitigate catastrophic forgetting. Prior work using replay for image classification tasks has found that selectively choosing the samples to replay offers little, if any, benefit over random selection. In contrast, we find that selective replay can significantly outperform random selection for the RPM task. △ Less

Submitted 19 April, 2021; v1 submitted 5 March, 2021; originally announced March 2021.

Comments: To appear in the IEEE Conference on Computer Vision and Pattern Recognition Workshop (CVPR-W) on Continual Learning in Computer Vision (CLVision) 2021

Quasi-Periodic Particle Acceleration in a Solar Flare

Authors: Brendan P. Clarke, Laura A. Hayes, Peter T. Gallagher, Shane A. Maloney, Eoin P. Carley

Abstract: A common feature of electromagnetic emission from solar flares is the presence of intensity pulsations that vary as a function of time. Known as quasi-periodic pulsations (QPPs), these variations in flux appear to include periodic components and characteristic time-scales. Here, we analyse a GOES M3.7 class flare exhibiting pronounced QPPs across a broad band of wavelengths using imaging and time-… ▽ More A common feature of electromagnetic emission from solar flares is the presence of intensity pulsations that vary as a function of time. Known as quasi-periodic pulsations (QPPs), these variations in flux appear to include periodic components and characteristic time-scales. Here, we analyse a GOES M3.7 class flare exhibiting pronounced QPPs across a broad band of wavelengths using imaging and time-series analysis. We identify QPPs in the timeseries of X-ray, low frequency radio and EUV wavelengths using wavelet analysis, and localise the region of the flare site from which the QPPs originate via X-ray and EUV imaging. It was found that the pulsations within the 171 Ȧ, 1600 Ȧ, soft X-ray (SXR), and hard X-ray (HXR) light curves yielded similar periods of $\sim$122 s, $\sim$131s, $\sim$123 s, and $\sim$137 s, respectively, indicating a common progenitor. The low frequency radio emission at 2.5 MHz contained a longer period of $\sim$231 s. Imaging analysis indicates that the location of the X-ray and EUV pulsations originates from a HXR footpoint linked to a system of nearby open magnetic field lines. Our results suggest that intermittent particle acceleration, likely due to 'bursty' magnetic reconnection, is responsible for the QPPs. The precipitating electrons accelerated towards the chromosphere produce the X-ray and EUV pulsations, while the esca** electrons result in low frequency radio pulses in the form of type III radio bursts. The modulation of the reconnection process, resulting in episodic particle acceleration, explains the presence of these QPPs across the entire spatial range of flaring emission. △ Less

Submitted 8 February, 2021; originally announced February 2021.

Enacting Musical Worlds: Common Approaches to using NIMEs within Performance and Person-Centred Arts Practices

Authors: Lauren Hayes

Abstract: Live music making can be understood as an enactive process, whereby musical experiences are created through human action. This suggests that musical worlds coevolve with their agents through repeated sensorimotor interactions with the environment (where the music is being created), and at the same time cannot be separated from their sociocultural contexts. This paper investigates this claim by exp… ▽ More Live music making can be understood as an enactive process, whereby musical experiences are created through human action. This suggests that musical worlds coevolve with their agents through repeated sensorimotor interactions with the environment (where the music is being created), and at the same time cannot be separated from their sociocultural contexts. This paper investigates this claim by exploring ways in which technology, physiology, and context are bound up within two different musical scenarios: live electronic musical performance; and person-centred arts applications of NIMEs. In this paper I outline an ethnographic and phenomenological enquiry into my experiences as both a performer of live electronic and electro-instrumental music, as well as my extensive background in working with new technologies in various therapeutic and person-centred artistic situations. This is in order to explore the sociocultural and technological contexts in which these activities take place. I propose that by understanding creative musical participation as a highly contextualised practice, we may discover that the greatest impact of rapidly develo** technological resources is their ability to afford richly diverse, personalised, and embodied forms of music making. I argue that this is applicable over a wide range of musical communities. △ Less

Submitted 1 December, 2020; originally announced December 2020.

ACM Class: H.5.5

Journal ref: Proceedings of the International Conference on New Interfaces for Musical Expression, 2015

Nuanced and Interrelated Mediations and Exigencies (NIME): Addressing the Prevailing Political and Epistemological Crises

Authors: Lauren Hayes, Adnan Marquez-Borbon

Abstract: Nearly two decades after its inception as a workshop at the ACM Conference on Human Factors in Computing Systems, NIME exists as an established international conference significantly distinct from its precursor. While this origin story is often noted, the implications of NIME's history as emerging from a field predominantly dealing with human-computer interaction have rarely been discussed. In thi… ▽ More Nearly two decades after its inception as a workshop at the ACM Conference on Human Factors in Computing Systems, NIME exists as an established international conference significantly distinct from its precursor. While this origin story is often noted, the implications of NIME's history as emerging from a field predominantly dealing with human-computer interaction have rarely been discussed. In this paper we highlight many of the recent -- and some not so recent -- challenges that have been brought upon the NIME community as it attempts to maintain and expand its identity as a platform for multidisciplinary research into HCI, interface design, and electronic and computer music. We discuss the relationship between the market demands of the neoliberal university -- which have underpinned academia's drive for innovation -- and the quantification and economisation of research performance which have facilitated certain disciplinary and social frictions to emerge within NIME-related research and practice. Drawing on work that engages with feminist theory and cultural studies, we suggest that critical reflection and moreover mediation is necessary in order to address burgeoning concerns which have been raised within the NIME discourse in relation to methodological approaches, `diversity and inclusion', `accessibility', and the fostering of rigorous interdisciplinary research. △ Less

Submitted 1 December, 2020; originally announced December 2020.

Comments: Proceedings of the International Conference on New Interfaces for Musical Expression, 2020

ACM Class: H.5.5

Cross-Modal Terrains: Navigating Sonic Space through Haptic Feedback

Authors: Gabriella Isaac, Lauren Hayes, Todd Ingalls

Abstract: This paper explores the idea of using virtual textural terrains as a means of generating haptic profiles for force-feedback controllers. This approach breaks from the para-digm established within audio-haptic research over the last few decades where physical models within virtual environments are designed to transduce gesture into sonic output. We outline a method for generating multimodal terrain… ▽ More This paper explores the idea of using virtual textural terrains as a means of generating haptic profiles for force-feedback controllers. This approach breaks from the para-digm established within audio-haptic research over the last few decades where physical models within virtual environments are designed to transduce gesture into sonic output. We outline a method for generating multimodal terrains using basis functions, which are rendered into monochromatic visual representations for inspection. This visual terrain is traversed using a haptic controller, the NovInt Falcon, which in turn receives force information based on the grayscale value of its location in this virtual space. As the image is traversed by a performer the levels of resistance vary, and the image is realized as a physical terrain. We discuss the potential of this approach to afford engaging musical experiences for both the performer and the audience as iterated through numerous performances. △ Less

Submitted 1 December, 2020; originally announced December 2020.

ACM Class: H.5.5

Journal ref: Proceedings of the International Conference on New Interfaces for Musical Expression, 2017

Solar Flare Energy Partitioning and Transport -- the Gradual Phase (a Heliophysics 2050 White Paper)

Authors: Graham S. Kerr, Meriem Alaoui, Joel C. Allred, Nicholas H. Bian, Brian R. Dennis, A. Gordon Emslie, Lyndsay Fletcher, Silvina Guidoni, Laura A. Hayes, Gordon D. Holman, Hugh S. Hudson, Judith T. Karpen, Adam F. Kowalski, Ryan O. Milligan, Vanessa Polito, Jiong Qiu, Daniel F. Ryan

Abstract: Solar flares are a fundamental component of solar eruptive events (SEEs; along with solar energetic particles, SEPs, and coronal mass ejections, CMEs). Flares are the first component of the SEE to impact our atmosphere, which can set the stage for the arrival of the associated SEPs and CME. Magnetic reconnection drives SEEs by restructuring the solar coronal magnetic field, liberating a tremendous… ▽ More Solar flares are a fundamental component of solar eruptive events (SEEs; along with solar energetic particles, SEPs, and coronal mass ejections, CMEs). Flares are the first component of the SEE to impact our atmosphere, which can set the stage for the arrival of the associated SEPs and CME. Magnetic reconnection drives SEEs by restructuring the solar coronal magnetic field, liberating a tremendous amount of energy which is partitioned into various physical manifestations: particle acceleration, mass and magnetic-field eruption, atmospheric heating, and the subsequent emission of radiation as solar flares. To explain and ultimately predict these geoeffective events, the heliophysics community requires a comprehensive understanding of the processes that transform and distribute stored magnetic energy into other forms, including the broadband radiative enhancement that characterises flares. This white paper, submitted to the Heliophysics 2050 Workshop, discusses the flare gradual phase part of SEEs, setting out the questions that need addressing via a combination of theoretical, modelling, and observational research. In short, the flare gradual phase persists much longer than predicted so, by 2050, we must identify the characteristics of the significant energy deposition sustaining the gradual phase, and address the fundamental processes of turbulence and non-local heat flux. △ Less

Submitted 17 September, 2020; originally announced September 2020.

Comments: White paper submitted to the Heliophysics 2050 Workshop

Solar Flare Energy Partitioning and Transport -- the Impulsive Phase (a Heliophysics 2050 White Paper)

Authors: Graham S. Kerr, Meriem Alaoui, Joel C. Allred, Nicholas H. Bian, Brian R. Dennis, A. Gordon Emslie, Lyndsay Fletcher, Silvina Guidoni, Laura A. Hayes, Gordon D. Holman, Hugh S. Hudson, Judith T. Karpen, Adam F. Kowalski, Ryan O. Milligan, Vanessa Polito, Jiong Qiu, Daniel F. Ryan

Abstract: Solar flares are a fundamental component of solar eruptive events (SEEs; along with solar energetic particles, SEPs, and coronal mass ejections, CMEs). Flares are the first component of the SEE to impact our atmosphere, which can set the stage for the arrival of the associated SEPs and CME. Magnetic reconnection drives SEEs by restructuring the solar coronal magnetic field, liberating a tremendous… ▽ More Solar flares are a fundamental component of solar eruptive events (SEEs; along with solar energetic particles, SEPs, and coronal mass ejections, CMEs). Flares are the first component of the SEE to impact our atmosphere, which can set the stage for the arrival of the associated SEPs and CME. Magnetic reconnection drives SEEs by restructuring the solar coronal magnetic field, liberating a tremendous amount of energy which is partitioned into various physical manifestations: particle acceleration, mass and magnetic-field eruption, atmospheric heating, and the subsequent emission of radiation as solar flares. To explain and ultimately predict these geoeffective events, the heliophysics community requires a comprehensive understanding of the processes that transform and distribute stored magnetic energy into other forms, including the broadband radiative enhancement that characterises flares. This white paper, submitted to the Heliophysics 2050 Workshop, discusses the flare impulsive phase part of SEEs, setting out the questions that need addressing via a combination of theoretical, modelling, and observational research. In short, by 2050 we must determine the mechanisms of particle acceleration and propagation, and must push beyond the paradigm of energy transport via nonthermal electron beams, to also account for accelerated protons & ions and downward directed Alfven waves. △ Less

Submitted 17 September, 2020; originally announced September 2020.

Comments: White paper submitted to the Heliophysics 2050 Workshop

Improved Robustness to Open Set Inputs via Tempered Mixup

Authors: Ryne Roady, Tyler L. Hayes, Christopher Kanan

Abstract: Supervised classification methods often assume that evaluation data is drawn from the same distribution as training data and that all classes are present for training. However, real-world classifiers must handle inputs that are far from the training distribution including samples from unknown classes. Open set robustness refers to the ability to properly label samples from previously unseen catego… ▽ More Supervised classification methods often assume that evaluation data is drawn from the same distribution as training data and that all classes are present for training. However, real-world classifiers must handle inputs that are far from the training distribution including samples from unknown classes. Open set robustness refers to the ability to properly label samples from previously unseen categories as novel and avoid high-confidence, incorrect predictions. Existing approaches have focused on either novel inference methods, unique training architectures, or supplementing the training data with additional background samples. Here, we propose a simple regularization technique easily applied to existing convolutional neural network architectures that improves open set robustness without a background dataset. Our method achieves state-of-the-art results on open set classification baselines and easily scales to large-scale open set classification problems. △ Less

Submitted 10 September, 2020; originally announced September 2020.

Comments: Proceedings of the ECCV 2020 Workshop on Adversarial Robustness in the Real World

RODEO: Replay for Online Object Detection

Authors: Manoj Acharya, Tyler L. Hayes, Christopher Kanan

Abstract: Humans can incrementally learn to do new visual detection tasks, which is a huge challenge for today's computer vision systems. Incrementally trained deep learning models lack backwards transfer to previously seen classes and suffer from a phenomenon known as $"catastrophic forgetting."$ In this paper, we pioneer online streaming learning for object detection, where an agent must learn examples on… ▽ More Humans can incrementally learn to do new visual detection tasks, which is a huge challenge for today's computer vision systems. Incrementally trained deep learning models lack backwards transfer to previously seen classes and suffer from a phenomenon known as $"catastrophic forgetting."$ In this paper, we pioneer online streaming learning for object detection, where an agent must learn examples one at a time with severe memory and computational constraints. In object detection, a system must output all bounding boxes for an image with the correct label. Unlike earlier work, the system described in this paper can learn this task in an online manner with new classes being introduced over time. We achieve this capability by using a novel memory replay mechanism that efficiently replays entire scenes. We achieve state-of-the-art results on both the PASCAL VOC 2007 and MS COCO datasets. △ Less

Submitted 14 August, 2020; originally announced August 2020.

Comments: Accepted for poster presentation at BMVC2020

Hot X-ray Onsets of Solar Flares

Authors: Hugh S. Hudson, Paulo J. A. Simoes, Lyndsay Fletcher, Laura A. Hayes, Iain G. Hannah

Abstract: The study of the localized plasma conditions before the impulsive phase of a solar flare can help us understand the physical processes that occur leading up to the main flare energy release. Here, we present evidence of a hot X-ray onset interval of enhanced isothermal plasma temperatures in the range of 10-15~MK up to tens of seconds prior to the flare's impulsive phase. This `hot onset' interval… ▽ More The study of the localized plasma conditions before the impulsive phase of a solar flare can help us understand the physical processes that occur leading up to the main flare energy release. Here, we present evidence of a hot X-ray onset interval of enhanced isothermal plasma temperatures in the range of 10-15~MK up to tens of seconds prior to the flare's impulsive phase. This `hot onset' interval occurs during the initial soft X-ray increase and prior to the detectable hard X-ray emission. The isothermal temperatures, estimated by the Geostationary Operational Environmental Satellite (GOES) X-ray sensor, and confirmed with data from the Reuven Ramaty High Energy Solar Spectroscopic Imager (RHESSI), show no signs of gradual increase, and the `hot onset' phenomenon occurs regardless of flare classification or configuration. In a small sample of four representative flare events we identify this early hot onset soft X-ray emission mainly within footpoint and low-lying loops, rather than with coronal structures, based on images from the Atmospheric Imaging Assembly (AIA). We confirm this via limb occultation of a flaring region. These hot X-ray onsets appear before there is evidence of collisional heating by non-thermal electrons, and hence they challenge the standard flare heating modeling techniques. △ Less

Submitted 10 July, 2020; originally announced July 2020.

Comments: Submitted to MNRAS 6 July 2020

Recurrent Sum-Product-Max Networks for Decision Making in Perfectly-Observed Environments

Authors: Hari Teja Tatavarti, Prashant Doshi, Layton Hayes

Abstract: Recent investigations into sum-product-max networks (SPMN) that generalize sum-product networks (SPN) offer a data-driven alternative for decision making, which has predominantly relied on handcrafted models. SPMNs computationally represent a probabilistic decision-making problem whose solution scales linearly in the size of the network. However, SPMNs are not well suited for sequential decision m… ▽ More Recent investigations into sum-product-max networks (SPMN) that generalize sum-product networks (SPN) offer a data-driven alternative for decision making, which has predominantly relied on handcrafted models. SPMNs computationally represent a probabilistic decision-making problem whose solution scales linearly in the size of the network. However, SPMNs are not well suited for sequential decision making over multiple time steps. In this paper, we present recurrent SPMNs (RSPMN) that learn from and model decision-making data over time. RSPMNs utilize a template network that is unfolded as needed depending on the length of the data sequence. This is significant as RSPMNs not only inherit the benefits of SPMNs in being data driven and mostly tractable, they are also well suited for sequential problems. We establish conditions on the template network, which guarantee that the resulting SPMN is valid, and present a structure learning algorithm to learn a sound template network. We demonstrate that the RSPMNs learned on a testbed of sequential decision-making data sets generate MEUs and policies that are close to the optimal on perfectly-observed domains. They easily improve on a recent batch-constrained reinforcement learning method, which is important because RSPMNs offer a new model-based approach to offline reinforcement learning. △ Less

Submitted 12 June, 2020; originally announced June 2020.

Do We Need Fully Connected Output Layers in Convolutional Networks?

Authors: Zhongchao Qian, Tyler L. Hayes, Kushal Kafle, Christopher Kanan

Abstract: Traditionally, deep convolutional neural networks consist of a series of convolutional and pooling layers followed by one or more fully connected (FC) layers to perform the final classification. While this design has been successful, for datasets with a large number of categories, the fully connected layers often account for a large percentage of the network's parameters. For applications with mem… ▽ More Traditionally, deep convolutional neural networks consist of a series of convolutional and pooling layers followed by one or more fully connected (FC) layers to perform the final classification. While this design has been successful, for datasets with a large number of categories, the fully connected layers often account for a large percentage of the network's parameters. For applications with memory constraints, such as mobile devices and embedded platforms, this is not ideal. Recently, a family of architectures that involve replacing the learned fully connected output layer with a fixed layer has been proposed as a way to achieve better efficiency. In this paper we examine this idea further and demonstrate that fixed classifiers offer no additional benefit compared to simply removing the output layer along with its parameters. We further demonstrate that the typical approach of having a fully connected final output layer is inefficient in terms of parameter count. We are able to achieve comparable performance to a traditionally learned fully connected classification output layer on the ImageNet-1K, CIFAR-100, Stanford Cars-196, and Oxford Flowers-102 datasets, while not having a fully connected output layer at all. △ Less

Submitted 28 April, 2020; v1 submitted 28 April, 2020; originally announced April 2020.

Statistical Study of GOES X-ray Quasi-Periodic Pulsations in Solar Flares

Authors: Laura A. Hayes, Andrew R. Inglis, Steven Christe, Brian Dennis, Peter T. Gallagher

Abstract: Small amplitude quasi-periodic pulsations (QPPs) detected in soft X-ray emission are commonplace in many flares. To date, the underpinning processes resulting in the QPPs are unknown. In this paper, we attempt to constrain the prevalence of \textit{stationary} QPPs in the largest statistical study to date, including a study of the relationship of QPP periods to the properties of the flaring active… ▽ More Small amplitude quasi-periodic pulsations (QPPs) detected in soft X-ray emission are commonplace in many flares. To date, the underpinning processes resulting in the QPPs are unknown. In this paper, we attempt to constrain the prevalence of \textit{stationary} QPPs in the largest statistical study to date, including a study of the relationship of QPP periods to the properties of the flaring active region, flare ribbons, and CME affiliation. We build upon the work of \cite{inglis2016} and use a model comparison test to search for significant power in the Fourier spectra of lightcurves of the GOES 1--8~Å channel. We analyze all X-, M- and C- class flares of the past solar cycle, a total of 5519 flares, and search for periodicity in the 6-300~s timescale range. Approximately 46\% of X-class, 29\% of M-class and 7\% of C-class flares show evidence of stationary QPPs, with periods that follow a log-normal distribution peaked at 20~s. The QPP periods were found to be independent of flare magnitude, however a positive correlation was found between QPP period and flare duration. No dependence of the QPP periods to the global active region properties was identified. A positive correlation was found between QPPs and ribbon properties including unsigned magnetic flux, ribbon area and ribbon separation distance. We found that both flares with and without an associated CME can host QPPs. Furthermore, we demonstrate that for X- and M- class flares, decay phase QPPs have statistically longer periods than impulsive phase QPPs. △ Less

Submitted 24 April, 2020; originally announced April 2020.

Comments: 17 pages, 13 figures, accepted in The Astrophysical Journal

Simulating Solar Flare Irradiance with Multithreaded Models of Flare Arcades

Authors: Jeffrey W. Reep, Harry P. Warren, Christopher S. Moore, Crisel Suarez, Laura A. Hayes

Abstract: Understanding how energy is released in flares is one of the central problems of solar and stellar astrophysics. Observations of high temperature flare plasma hold many potential clues as to the nature of this energy release. It is clear, however, that flares are not composed of a few impulsively heated loops, but are the result of heating on many small-scale threads that are energized over time,… ▽ More Understanding how energy is released in flares is one of the central problems of solar and stellar astrophysics. Observations of high temperature flare plasma hold many potential clues as to the nature of this energy release. It is clear, however, that flares are not composed of a few impulsively heated loops, but are the result of heating on many small-scale threads that are energized over time, making it difficult to compare observations and numerical simulations in detail. Several previous studies have shown that it is possible to reproduce some aspects of the observed emission by considering the flare as a sequence of independently heated loops, but these studies generally focus on small-scale features while ignoring the global features of the flare. In this paper, we develop a multithreaded model that encompasses the time-varying geometry and heating rate for a series of successively-heated loops comprising an arcade. To validate, we compare with spectral observations of five flares made with the MinXSS CubeSat as well as light curves measured with GOES/XRS and SDO/AIA. We show that this model can successfully reproduce the light curves and quasi-periodic pulsations in GOES/XRS, the soft X-ray spectra seen with MinXSS, and the light curves in various AIA passbands. The AIA light curves are most consistent with long duration heating, but elemental abundances cannot be constrained with the model. Finally, we show how this model can be used to extrapolate to spectra of extreme events that can predict irradiance across a wide wavelength range including unobserved wavelengths. △ Less

Submitted 7 May, 2020; v1 submitted 23 March, 2020; originally announced March 2020.

Comments: Accepted to ApJ. v2: Mistake in temperature plots in Figure 1 corrected

srlearn: A Python Library for Gradient-Boosted Statistical Relational Models

Authors: Alexander L. Hayes

Abstract: We present srlearn, a Python library for boosted statistical relational models. We adapt the scikit-learn interface to this setting and provide examples for how this can be used to express learning and inference problems. We present srlearn, a Python library for boosted statistical relational models. We adapt the scikit-learn interface to this setting and provide examples for how this can be used to express learning and inference problems. △ Less

Submitted 17 December, 2019; originally announced December 2019.

Comments: Ninth International Workshop on Statistical Relational AI (StarAI 2020). Software online at https://github.com/hayesall/srlearn

User Friendly Automatic Construction of Background Knowledge: Mode Construction from ER Diagrams

Authors: Alexander L. Hayes, Mayukh Das, Phillip Odom, Sriraam Natarajan

Abstract: One of the key advantages of Inductive Logic Programming systems is the ability of the domain experts to provide background knowledge as modes that allow for efficient search through the space of hypotheses. However, there is an inherent assumption that this expert should also be an ILP expert to provide effective modes. We relax this assumption by designing a graphical user interface that allows… ▽ More One of the key advantages of Inductive Logic Programming systems is the ability of the domain experts to provide background knowledge as modes that allow for efficient search through the space of hypotheses. However, there is an inherent assumption that this expert should also be an ILP expert to provide effective modes. We relax this assumption by designing a graphical user interface that allows the domain expert to interact with the system using Entity Relationship diagrams. These interactions are used to construct modes for the learning system. We evaluate our algorithm on a probabilistic logic learning system where we demonstrate that the user is able to construct effective background knowledge on par with the expert-encoded knowledge on five data sets. △ Less

Submitted 16 December, 2019; originally announced December 2019.

Comments: 8 pages. Published in Proceedings of the Knowledge Capture Conference, 2017

Journal ref: Proceedings of the Knowledge Capture Conference (2017) 30:1-30:8

Are Out-of-Distribution Detection Methods Effective on Large-Scale Datasets?

Authors: Ryne Roady, Tyler L. Hayes, Ronald Kemker, Ayesha Gonzales, Christopher Kanan

Abstract: Supervised classification methods often assume the train and test data distributions are the same and that all classes in the test set are present in the training set. However, deployed classifiers often require the ability to recognize inputs from outside the training set as unknowns. This problem has been studied under multiple paradigms including out-of-distribution detection and open set recog… ▽ More Supervised classification methods often assume the train and test data distributions are the same and that all classes in the test set are present in the training set. However, deployed classifiers often require the ability to recognize inputs from outside the training set as unknowns. This problem has been studied under multiple paradigms including out-of-distribution detection and open set recognition. For convolutional neural networks, there have been two major approaches: 1) inference methods to separate knowns from unknowns and 2) feature space regularization strategies to improve model robustness to outlier inputs. There has been little effort to explore the relationship between the two approaches and directly compare performance on anything other than small-scale datasets that have at most 100 categories. Using ImageNet-1K and Places-434, we identify novel combinations of regularization and specialized inference methods that perform best across multiple outlier detection problems of increasing difficulty level. We found that input perturbation and temperature scaling yield the best performance on large scale datasets regardless of the feature space regularization strategy. Improving the feature space by regularizing against a background class can be helpful if an appropriate background class can be found, but this is impractical for large scale image classification datasets. △ Less

Submitted 30 October, 2019; originally announced October 2019.

A blueprint of state-of-the-art techniques for detecting quasi-periodic pulsations in solar and stellar flares

Authors: Anne-Marie Broomhall, James R. A. Davenport, Laura A. Hayes, Andrew R. Inglis, Dmitrii Y. Kolotkov, James A. McLaughlin, Tishtrya Mehta, Valery M. Nakariakov, Yuta Notsu, David J. Pascoe, Chloe E. Pugh, Tom Van Doorsselaere

Abstract: Quasi-periodic pulsations (QPPs) appear to be a common feature observed in the light curves of both solar and stellar flares. However, their quasi-periodic nature, along with the fact that they can be small in amplitude and short-lived, makes QPPs difficult to unequivocally detect. In this paper, we test the strengths and limitations of state-of-the-art methods for detecting QPPs using a series of… ▽ More Quasi-periodic pulsations (QPPs) appear to be a common feature observed in the light curves of both solar and stellar flares. However, their quasi-periodic nature, along with the fact that they can be small in amplitude and short-lived, makes QPPs difficult to unequivocally detect. In this paper, we test the strengths and limitations of state-of-the-art methods for detecting QPPs using a series of hare-and-hounds exercises. The hare simulated a set of flares, both with and without QPPs of a variety of forms, while the hounds attempted to detect QPPs in blind tests. We use the results of these exercises to create a blueprint for anyone who wishes to detect QPPs in real solar and stellar data. We present eight clear recommendations to be kept in mind for future QPP detections, with the plethora of solar and stellar flare data from new and future satellites. These recommendations address the key pitfalls in QPP detection, including detrending, trimming data, accounting for colored noise, detecting stationary-period QPPs, detecting QPPs with nonstationary periods, and ensuring that detections are robust and false detections are minimized. We find that QPPs can be detected reliably and robustly by a variety of methods, which are clearly identified and described, if the appropriate care and due diligence are taken. △ Less

Submitted 18 October, 2019; originally announced October 2019.

Comments: Accepted for publication in ApJSS, 46 pages, 29 figures, 7 tables

REMIND Your Neural Network to Prevent Catastrophic Forgetting

Authors: Tyler L. Hayes, Kushal Kafle, Robik Shrestha, Manoj Acharya, Christopher Kanan

Abstract: People learn throughout life. However, incrementally updating conventional neural networks leads to catastrophic forgetting. A common remedy is replay, which is inspired by how the brain consolidates memory. Replay involves fine-tuning a network on a mixture of new and old instances. While there is neuroscientific evidence that the brain replays compressed memories, existing methods for convolutio… ▽ More People learn throughout life. However, incrementally updating conventional neural networks leads to catastrophic forgetting. A common remedy is replay, which is inspired by how the brain consolidates memory. Replay involves fine-tuning a network on a mixture of new and old instances. While there is neuroscientific evidence that the brain replays compressed memories, existing methods for convolutional networks replay raw images. Here, we propose REMIND, a brain-inspired approach that enables efficient replay with compressed representations. REMIND is trained in an online manner, meaning it learns one example at a time, which is closer to how humans learn. Under the same constraints, REMIND outperforms other methods for incremental class learning on the ImageNet ILSVRC-2012 dataset. We probe REMIND's robustness to data ordering schemes known to induce catastrophic forgetting. We demonstrate REMIND's generality by pioneering online learning for Visual Question Answering (VQA). △ Less

Submitted 13 July, 2020; v1 submitted 6 October, 2019; originally announced October 2019.

Comments: To appear in the European Conference on Computer Vision (ECCV-2020)