Showing 101–150 of 619 results for author: Cheung, C

Adaptive Label Smoothing with Self-Knowledge in Natural Language Generation

Authors: Dongkyu Lee, Ka Chun Cheung, Nevin L. Zhang

Abstract: Overconfidence has been shown to impair generalization and calibration of a neural network. Previous studies remedy this issue by adding a regularization term to a loss function, preventing a model from making a peaked distribution. Label smoothing smoothes target labels with a pre-defined prior label distribution; as a result, a model is learned to maximize the likelihood of predicting the soft l… ▽ More Overconfidence has been shown to impair generalization and calibration of a neural network. Previous studies remedy this issue by adding a regularization term to a loss function, preventing a model from making a peaked distribution. Label smoothing smoothes target labels with a pre-defined prior label distribution; as a result, a model is learned to maximize the likelihood of predicting the soft label. Nonetheless, the amount of smoothing is the same in all samples and remains fixed in training. In other words, label smoothing does not reflect the change in probability distribution mapped by a model over the course of training. To address this issue, we propose a regularization scheme that brings dynamic nature into the smoothing parameter by taking model probability distribution into account, thereby varying the parameter per instance. A model in training self-regulates the extent of smoothing on the fly during forward propagation. Furthermore, inspired by recent work in bridging label smoothing and knowledge distillation, our work utilizes self-knowledge as a prior label distribution in softening target labels, and presents theoretical support for the regularization effect by knowledge distillation and the dynamic smoothing parameter. Our regularizer is validated comprehensively, and the result illustrates marked improvements in model generalization and calibration, enhancing robustness and trustworthiness of a model. △ Less

Submitted 22 October, 2022; originally announced October 2022.

Comments: EMNLP 2022

Hard Gate Knowledge Distillation -- Leverage Calibration for Robust and Reliable Language Model

Authors: Dongkyu Lee, Zhiliang Tian, Yingxiu Zhao, Ka Chun Cheung, Nevin L. Zhang

Abstract: In knowledge distillation, a student model is trained with supervisions from both knowledge from a teacher and observations drawn from a training data distribution. Knowledge of a teacher is considered a subject that holds inter-class relations which send a meaningful supervision to a student; hence, much effort has been put to find such knowledge to be distilled. In this paper, we explore a quest… ▽ More In knowledge distillation, a student model is trained with supervisions from both knowledge from a teacher and observations drawn from a training data distribution. Knowledge of a teacher is considered a subject that holds inter-class relations which send a meaningful supervision to a student; hence, much effort has been put to find such knowledge to be distilled. In this paper, we explore a question that has been given little attention: "when to distill such knowledge." The question is answered in our work with the concept of model calibration; we view a teacher model not only as a source of knowledge but also as a gauge to detect miscalibration of a student. This simple and yet novel view leads to a hard gate knowledge distillation scheme that switches between learning from a teacher model and training data. We verify the gating mechanism in the context of natural language generation at both the token-level and the sentence-level. Empirical comparisons with strong baselines show that hard gate knowledge distillation not only improves model generalization, but also significantly lowers model calibration error. △ Less

Submitted 22 October, 2022; originally announced October 2022.

Comments: EMNLP 2022

Veneziano Variations: How Unique are String Amplitudes?

Authors: Clifford Cheung, Grant N. Remmen

Abstract: String theory offers an elegant and concrete realization of how to consistently couple states of arbitrarily high spin. But how unique is this construction? In this paper we derive a novel, multi-parameter family of four-point scattering amplitudes exhibiting i) polynomially bounded high-energy behavior and ii) exchange of an infinite tower of high-spin modes, albeit with a finite number of states… ▽ More String theory offers an elegant and concrete realization of how to consistently couple states of arbitrarily high spin. But how unique is this construction? In this paper we derive a novel, multi-parameter family of four-point scattering amplitudes exhibiting i) polynomially bounded high-energy behavior and ii) exchange of an infinite tower of high-spin modes, albeit with a finite number of states at each resonance. These amplitudes take an infinite-product form and, depending on parameters, exhibit mass spectra that are either unbounded or bounded, thus corresponding to generalizations of the Veneziano and Coon amplitudes, respectively. For the bounded case, masses converge to an accumulation point, a peculiar feature seen in the Coon amplitude but more recently understood to arise naturally in string theory. Importantly, our amplitudes contain free parameters allowing for the customization of the slope and offset of the spin-dependence in the Regge trajectory. We compute all partial waves for this multi-parameter class of amplitudes and identify unitary regions of parameter space. For the unbounded case, we apply similar methods to derive new deformations of the Veneziano and Virasoro-Shapiro amplitudes. △ Less

Submitted 30 January, 2023; v1 submitted 21 October, 2022; originally announced October 2022.

Comments: 32 pages, 3 figures

Journal ref: JHEP 2301:122,2023

Radio Emission of Nearby Early-type Galaxies at Low and Very-Low Radio Luminosity Range

Authors: Anna Wójtowicz, Łukasz Stawarz, C. C. Cheung, Norbert Werner, Dominik Rudka

Abstract: We analyze radio continuum emission of early-type galaxies with dynamical measurements of central super-massive black hole (SMBH) masses, and well-characterized large-scale environments, but regardless on the exact level of the nuclear activity. The 1.4 GHz radio fluxes collected with $\sim$arcmin resolution for 62 nearby targets (distances $\lesssim$ 153 Mpc), correspond to low and very low monoc… ▽ More We analyze radio continuum emission of early-type galaxies with dynamical measurements of central super-massive black hole (SMBH) masses, and well-characterized large-scale environments, but regardless on the exact level of the nuclear activity. The 1.4 GHz radio fluxes collected with $\sim$arcmin resolution for 62 nearby targets (distances $\lesssim$ 153 Mpc), correspond to low and very low monochromatic luminosities $L_{\rm r} \sim 10^{35} - 10^{41}$ erg s$^{-1}$. We quantify possible correlations between the radio properties with the main parameters of supermassive black holes, host galaxies, and hot gaseous halos, finding a general bimodality in the radio luminosity distribution, with the borderline between ``radio-bright'' and ``radio-dim'' populations $\log L_{\rm r} / L_{\rm Edd} \simeq -8.5$. We analyze the far-infrared data for the targets, finding that all radio-bright sources, and over a half of radio-dim ones, are over-luminous in radio with respect to the far-infrared--radio correlation. High-resolution radio maps reveal that the overwhelming majority of radio-dim sources are unresolved on arcsecond scale, while the bulk of radio-bright sources display extended jets and lobes of low- and intermediate-power radio galaxies; those jets dominate radio emission of radio-bright objects. Regarding the origin of the radio emission of radio-dim sources, we discuss the two main possibilities. One is the ADAF model, in which the radio and the nuclear X-ray radiative outputs at very low accretion rates, are both dominated by unresolved jets. The other possibility is that the radio-dim sources, unlike the radio-bright ones, are characterized by low values of SMBH spins, so that their radio emission is not related to the jets, but instead is due to a combination of starforming processes and past nuclear outbursts. △ Less

Submitted 24 January, 2023; v1 submitted 29 September, 2022; originally announced September 2022.

Comments: revised version accepted for publication in the Astrophysical Journal

Solar coronal heating from small-scale magnetic braids

Authors: L. P. Chitta, H. Peter, S. Parenti, D. Berghmans, F. Auchère, S. K. Solanki, R. Aznar Cuadrado, U. Schühle, L. Teriaca, S. Mandal, K. Barczynski, É. Buchlin, L. Harra, E. Kraaikamp, D. M. Long, L. Rodriguez, C. Schwanitz, P. J. Smith, C. Verbeeck, A. N. Zhukov, W. Liu, M. C. M. Cheung

Abstract: Relaxation of braided coronal magnetic fields through reconnection is thought to be a source of energy to heat plasma in active region coronal loops. However, observations of active region coronal heating associated with an untangling of magnetic braids remain sparse. One reason for this paucity could be the lack of coronal observations with a sufficiently high spatial and temporal resolution to c… ▽ More Relaxation of braided coronal magnetic fields through reconnection is thought to be a source of energy to heat plasma in active region coronal loops. However, observations of active region coronal heating associated with an untangling of magnetic braids remain sparse. One reason for this paucity could be the lack of coronal observations with a sufficiently high spatial and temporal resolution to capture this process in action. Using new observations with high spatial resolution (250-270 km on the Sun) and high cadence (3-10 s) from the Extreme Ultraviolet Imager (EUI) on board Solar Orbiter, we observed the untangling of small-scale coronal braids in different active regions. The untangling is associated with impulsive heating of the gas in these braided loops. We assess that coronal magnetic braids overlying cooler chromospheric filamentary structures are perhaps more common. Furthermore, our observations show signatures of spatially coherent and intermittent coronal heating during the relaxation of the magnetic braids. Our study reveals the operation of gentle and impulsive modes of magnetic reconnection in the solar corona. △ Less

Submitted 26 November, 2022; v1 submitted 25 September, 2022; originally announced September 2022.

Comments: Published in Astronomy & Astrophysics

Journal ref: A&A 667, A166 (2022)

NeuralMarker: A Framework for Learning General Marker Correspondence

Authors: Zhaoyang Huang, Xiaokun Pan, Weihong Pan, Weikang Bian, Yan Xu, Ka Chun Cheung, Guofeng Zhang, Hongsheng Li

Abstract: We tackle the problem of estimating correspondences from a general marker, such as a movie poster, to an image that captures such a marker. Conventionally, this problem is addressed by fitting a homography model based on sparse feature matching. However, they are only able to handle plane-like markers and the sparse features do not sufficiently utilize appearance information. In this paper, we pro… ▽ More We tackle the problem of estimating correspondences from a general marker, such as a movie poster, to an image that captures such a marker. Conventionally, this problem is addressed by fitting a homography model based on sparse feature matching. However, they are only able to handle plane-like markers and the sparse features do not sufficiently utilize appearance information. In this paper, we propose a novel framework NeuralMarker, training a neural network estimating dense marker correspondences under various challenging conditions, such as marker deformation, harsh lighting, etc. Besides, we also propose a novel marker correspondence evaluation method circumstancing annotations on real marker-image pairs and create a new benchmark. We show that NeuralMarker significantly outperforms previous methods and enables new interesting applications, including Augmented Reality (AR) and video editing. △ Less

Submitted 19 September, 2022; originally announced September 2022.

Comments: Accepted by ToG (SIGGRAPH Asia 2022). Project Page: https://drinkingcoder.github.io/publication/neuralmarker/

Multiwavelength Study of Dark Globule DC 314.8-5.1: Point Source Identification and Diffuse Emission Characterization

Authors: E. Kosmaczewski, L. Stawarz, C. C. Cheung, A. Bamba, A. Karska, W. R. M. Rocha

Abstract: We present an analysis of multi-wavelength observations of the dark globule DC\,314.8--5.1, using data from the Gaia optical, 2MASS near-infrared, and WISE mid-infrared surveys, dedicated imaging with the Spitzer Space Telescope, and X-ray data obtained with the Swift-XRT Telescope (XRT). The main goal was to identify possible pre-main sequence stars (PMSs) and young stellar objects (YSOs) associa… ▽ More We present an analysis of multi-wavelength observations of the dark globule DC\,314.8--5.1, using data from the Gaia optical, 2MASS near-infrared, and WISE mid-infrared surveys, dedicated imaging with the Spitzer Space Telescope, and X-ray data obtained with the Swift-XRT Telescope (XRT). The main goal was to identify possible pre-main sequence stars (PMSs) and young stellar objects (YSOs) associated with the globule. For this, we studied the infrared colors of all point sources within the boundaries of the cloud. After removing sources with non-stellar spectra, we investigated the Gaia parallaxes for the YSO candidates, and found that none are physically related to DC\,314.8--5.1. In addition, we searched for X-ray emission from pre-main sequence stars with Swift-XRT, and found no 0.5--10\,keV emission down to a luminosity level $\lesssim 10^{31}$erg\,s$^{-1}$, typical of a PMS with mass\,$\ge 2 M_\odot$. Our detailed inspection therefore supports a very young, ``pre-stellar core'' evolutionary stage for the cloud. Based on archival Planck and IRAS data, we moreover identify the presence of hot dust, with temperatures $\gtrsim 100$\,K, in addition to the dominant dust component at 14\,K, originating with the associated reflection nebula. △ Less

Submitted 25 October, 2023; v1 submitted 6 September, 2022; originally announced September 2022.

Comments: Accepted to AJ

Possible Gravitational Microlensing Events in the Optical Lightcurve of Active Galaxy S5 0716+714

Authors: D. Ł. Król, Ł. Stawarz, J. Krzesinski, C. C. Cheung

Abstract: A well-known active galaxy of the blazar type, S5 0716+714, is characterized by a particularly high variability duty cycle on short-time scales at optical frequencies. As such, the source was subjected to numerous monitoring programs, including both ground-based as well as space-borne telescopes. On closer inspection of the most recent accumulation of the data provided by the Transiting Exoplanet… ▽ More A well-known active galaxy of the blazar type, S5 0716+714, is characterized by a particularly high variability duty cycle on short-time scales at optical frequencies. As such, the source was subjected to numerous monitoring programs, including both ground-based as well as space-borne telescopes. On closer inspection of the most recent accumulation of the data provided by the Transiting Exoplanet Survey Satellite, we have noticed several conspicuous events with `volcano-like' symmetric shape, lasting all for several hours, which closely resemble the achromatic events detected with the previous Whole Earth Blazar Telescope campaigns targeting the source. We propose that those peculiar features could be due to the gravitational micro-lensing of the innermost segments of the precessing jet in the system, by a binary lens. We study the magnification pattern of the lens with the inverse-ray shooting method, and the source trajectory parameters with the Python package MuLensModel. In this way, we were able to fit successfully all the selected events with a single lens, adjusting slightly only the source trajectory parameters for each lensing event. Based on the fit results, we postulate the presence of a massive binary lens, containing an intermediate-mass black hole, possibly even a super-massive one, and a much less massive companion (by a factor of $\lesssim 0.01$), located within the host galaxy of the blazar, most likely the central kiloparsec region. We discuss the major physical implications of the proposed scenario regarding the quest for the intermediate-mass and dual supermassive black holes in active galaxies. △ Less

Submitted 22 December, 2022; v1 submitted 28 August, 2022; originally announced August 2022.

Comments: accepted for publication in the Astrophysical Journal

Exploring the Limits of Synthetic Creation of Solar EUV Images via Image-to-Image Translation

Authors: Valentina Salvatelli, Luiz F. G. dos Santos, Souvik Bose, Brad Neuberg, Mark C. M. Cheung, Miho Janvier, Meng Jin, Yarin Gal, Atilim Gunes Baydin

Abstract: The Solar Dynamics Observatory (SDO), a NASA multi-spectral decade-long mission that has been daily producing terabytes of observational data from the Sun, has been recently used as a use-case to demonstrate the potential of machine learning methodologies and to pave the way for future deep-space mission planning. In particular, the idea of using image-to-image translation to virtually produce ext… ▽ More The Solar Dynamics Observatory (SDO), a NASA multi-spectral decade-long mission that has been daily producing terabytes of observational data from the Sun, has been recently used as a use-case to demonstrate the potential of machine learning methodologies and to pave the way for future deep-space mission planning. In particular, the idea of using image-to-image translation to virtually produce extreme ultra-violet channels has been proposed in several recent studies, as a way to both enhance missions with less available channels and to alleviate the challenges due to the low downlink rate in deep space. This paper investigates the potential and the limitations of such a deep learning approach by focusing on the permutation of four channels and an encoder--decoder based architecture, with particular attention to how morphological traits and brightness of the solar surface affect the neural network predictions. In this work we want to answer the question: can synthetic images of the solar corona produced via image-to-image translation be used for scientific studies of the Sun? The analysis highlights that the neural network produces high-quality images over three orders of magnitude in count rate (pixel intensity) and can generally reproduce the covariance across channels within a 1% error. However the model performance drastically diminishes in correspondence of extremely high energetic events like flares, and we argue that the reason is related to the rareness of such events posing a challenge to model training. △ Less

Submitted 19 August, 2022; originally announced August 2022.

Comments: 16 pages, 8 figures. To be published on ApJ (submitted on Feb 21st, accepted on July 28th)

Journal ref: ApJ 937 (2022) 100

Expressing Multivariate Time Series as Graphs with Time Series Attention Transformer

Authors: William T. Ng, K. Siu, Albert C. Cheung, Michael K. Ng

Abstract: A reliable and efficient representation of multivariate time series is crucial in various downstream machine learning tasks. In multivariate time series forecasting, each variable depends on its historical values and there are inter-dependencies among variables as well. Models have to be designed to capture both intra- and inter-relationships among the time series. To move towards this goal, we pr… ▽ More A reliable and efficient representation of multivariate time series is crucial in various downstream machine learning tasks. In multivariate time series forecasting, each variable depends on its historical values and there are inter-dependencies among variables as well. Models have to be designed to capture both intra- and inter-relationships among the time series. To move towards this goal, we propose the Time Series Attention Transformer (TSAT) for multivariate time series representation learning. Using TSAT, we represent both temporal information and inter-dependencies of multivariate time series in terms of edge-enhanced dynamic graphs. The intra-series correlations are represented by nodes in a dynamic graph; a self-attention mechanism is modified to capture the inter-series correlations by using the super-empirical mode decomposition (SMD) module. We applied the embedded dynamic graphs to times series forecasting problems, including two real-world datasets and two benchmark datasets. Extensive experiments show that TSAT clearly outerperforms six state-of-the-art baseline methods in various forecasting horizons. We further visualize the embedded dynamic graphs to illustrate the graph representation power of TSAT. We share our code at https://github.com/RadiantResearch/TSAT. △ Less

Submitted 19 August, 2022; originally announced August 2022.

Comments: IJCAI'22 WORKSHOP AI4TS: AI FOR TIME SERIES ANALYSIS

Learning Degradation Representations for Image Deblurring

Authors: Dasong Li, Yi Zhang, Ka Chun Cheung, Xiaogang Wang, Hongwei Qin, Hongsheng Li

Abstract: In various learning-based image restoration tasks, such as image denoising and image super-resolution, the degradation representations were widely used to model the degradation process and handle complicated degradation patterns. However, they are less explored in learning-based image deblurring as blur kernel estimation cannot perform well in real-world challenging cases. We argue that it is part… ▽ More In various learning-based image restoration tasks, such as image denoising and image super-resolution, the degradation representations were widely used to model the degradation process and handle complicated degradation patterns. However, they are less explored in learning-based image deblurring as blur kernel estimation cannot perform well in real-world challenging cases. We argue that it is particularly necessary for image deblurring to model degradation representations since blurry patterns typically show much larger variations than noisy patterns or high-frequency textures.In this paper, we propose a framework to learn spatially adaptive degradation representations of blurry images. A novel joint image reblurring and deblurring learning process is presented to improve the expressiveness of degradation representations. To make learned degradation representations effective in reblurring and deblurring, we propose a Multi-Scale Degradation Injection Network (MSDI-Net) to integrate them into the neural networks. With the integration, MSDI-Net can handle various and complicated blurry patterns adaptively. Experiments on the GoPro and RealBlur datasets demonstrate that our proposed deblurring framework with the learned degradation representations outperforms state-of-the-art methods with appealing improvements. The code is released at https://github.com/dasongli1/Learning_degradation. △ Less

Submitted 10 August, 2022; originally announced August 2022.

Comments: Accepted to ECCV 2022

Journal ref: ECCV 2022

Ultraviolet Spectropolarimetry With Polstar: Using Polstar to test Magnetospheric Mass-loss Quenching

Authors: M. E. Shultz, R. Casini, M. C. M. Cheung, A. David-Uraz, T. del Pino Alemán, C. Erba, C. P. Folsom, K. Gayley, R. Ignace, Z. Keszthelyi, O. Kochukhov, Y. Nazé, C. Neiner, M. Oksala, V. Petit, P. A. Scowen, N. Sudnik, A. ud-Doula, J. S. Vink, G. A. Wade

Abstract: Polstar is a proposed NASA MIDEX space telescope that will provide high-resolution, simultaneous full-Stokes spectropolarimetry in the far ultraviolet, together with low-resolution linear polarimetry in the near ultraviolet. This observatory offers unprecedented capabilities to obtain unique information on the magnetic and plasma properties of the magnetospheres of hot stars. We describe an observ… ▽ More Polstar is a proposed NASA MIDEX space telescope that will provide high-resolution, simultaneous full-Stokes spectropolarimetry in the far ultraviolet, together with low-resolution linear polarimetry in the near ultraviolet. This observatory offers unprecedented capabilities to obtain unique information on the magnetic and plasma properties of the magnetospheres of hot stars. We describe an observing program making use of the known population of magnetic hot stars to test the fundamental hypothesis that magnetospheres should act to rapidly drain angular momentum, thereby spinning the star down, whilst simultaneously reducing the net mass-loss rate. Both effects are expected to lead to dramatic differences in the evolution of magnetic vs. non-magnetic stars. △ Less

Submitted 26 July, 2022; originally announced July 2022.

Comments: 20 pages, 10 figures, accepted for publication in ApSS. arXiv admin note: substantial text overlap with arXiv:2111.06434

Fermi-LAT Gamma-ray Detection of the Recurrent Nova RS Ophiuchi during its 2021 Outburst

Authors: C. C. Cheung, T. J. Johnson, P. Jean, M. Kerr, K. L. Page, J. P. Osborne, A. P. Beardmore, K. V. Sokolovsky, F. Teyssier, S. Ciprini, G. Marti-Devesa, I. Mereu, S. Razzaque, K. S. Wood, S. N. Shore, S. Korotkiy, A. Levina, A. Blumenzweig

Abstract: We report the Fermi-LAT gamma-ray detection of the 2021 outburst of the symbiotic recurrent nova RS Ophiuchi. In this system, unlike classical novae from cataclysmic binaries, the ejecta from the white dwarf form shocks when interacting with the dense circumstellar wind environment of the red giant companion. We find the LAT spectra from 50 MeV to ~20-23 GeV, the highest-energy photons detected in… ▽ More We report the Fermi-LAT gamma-ray detection of the 2021 outburst of the symbiotic recurrent nova RS Ophiuchi. In this system, unlike classical novae from cataclysmic binaries, the ejecta from the white dwarf form shocks when interacting with the dense circumstellar wind environment of the red giant companion. We find the LAT spectra from 50 MeV to ~20-23 GeV, the highest-energy photons detected in some sub-intervals, are consistent with $π^{\rm 0}$-decay emission from shocks in the ejecta as proposed by Tatischeff & Hernanz (2007) for its previous 2006 outburst. The LAT light-curve displayed a fast rise to its peak >0.1 GeV flux of $\simeq$6x10^-6 ph cm^-2 s^-1 beginning on day 0.745 after its optically-constrained eruption epoch of 2021 August 8.50. The peak lasted for ~1 day, and exhibited a power-law decline up to the final LAT detection on day 45. We analyze the data on shorter timescales at early times and found evidence of an approximate doubling of emission over ~200 minutes at day 2.2, possibly indicating a localized shock-acceleration event. Comparing the data collected by the AAVSO, we measured a constant ratio of ~2.8x10^-3 between the gamma-ray and optical luminosities except for a ~5x smaller ratio within the first day of the eruption likely indicating attenuation of gamma rays by ejecta material and lower high-energy proton fluxes at the earliest stages of the shock development. The hard X-ray emission due to bremsstrahlung from shock-heated gas traced by the Swift-XRT 2-10 keV light-curve peaked at day ~6, later than at GeV and optical energies. Using X-ray derived temperatures to constrain the velocity profile, we find the hadronic model reproduces the observed >0.1 GeV light-curve. △ Less

Submitted 6 July, 2022; originally announced July 2022.

Comments: ApJ, accepted. 21 pages, 10 figures, 4 tables

Attributed Abnormality Graph Embedding for Clinically Accurate X-Ray Report Generation

Authors: Sixing Yan, William K. Cheung, Keith Chiu, Terence M. Tong, Charles K. Cheung, Simon See

Abstract: Automatic generation of medical reports from X-ray images can assist radiologists to perform the time-consuming and yet important reporting task. Yet, achieving clinically accurate generated reports remains challenging. Modeling the underlying abnormalities using the knowledge graph approach has been found promising in enhancing the clinical accuracy. In this paper, we introduce a novel fined-grai… ▽ More Automatic generation of medical reports from X-ray images can assist radiologists to perform the time-consuming and yet important reporting task. Yet, achieving clinically accurate generated reports remains challenging. Modeling the underlying abnormalities using the knowledge graph approach has been found promising in enhancing the clinical accuracy. In this paper, we introduce a novel fined-grained knowledge graph structure called an attributed abnormality graph (ATAG). The ATAG consists of interconnected abnormality nodes and attribute nodes, allowing it to better capture the abnormality details. In contrast to the existing methods where the abnormality graph was constructed manually, we propose a methodology to automatically construct the fine-grained graph structure based on annotations, medical reports in X-ray datasets, and the RadLex radiology lexicon. We then learn the ATAG embedding using a deep model with an encoder-decoder architecture for the report generation. In particular, graph attention networks are explored to encode the relationships among the abnormalities and their attributes. A gating mechanism is adopted and integrated with various decoders for the generation. We carry out extensive experiments based on the benchmark datasets, and show that the proposed ATAG-based deep model outperforms the SOTA methods by a large margin and can improve the clinical accuracy of the generated reports. △ Less

Submitted 5 July, 2022; v1 submitted 4 July, 2022; originally announced July 2022.

Comments: 14 pages, 7 figures

Question Personalization in an Intelligent Tutoring System

Authors: Sabina Elkins, Robert Belfer, Ekaterina Kochmar, Iulian Serban, Jackie C. K. Cheung

Abstract: This paper investigates personalization in the field of intelligent tutoring systems (ITS). We hypothesize that personalization in the way questions are asked improves student learning outcomes. Previous work on dialogue-based ITS personalization has yet to address question phrasing. We show that generating versions of the questions suitable for students at different levels of subject proficiency… ▽ More This paper investigates personalization in the field of intelligent tutoring systems (ITS). We hypothesize that personalization in the way questions are asked improves student learning outcomes. Previous work on dialogue-based ITS personalization has yet to address question phrasing. We show that generating versions of the questions suitable for students at different levels of subject proficiency improves student learning gains, using variants written by a domain expert and an experimental A/B test. This insight demonstrates that the linguistic realization of questions in an ITS affects the learning outcomes for students. △ Less

Submitted 25 May, 2022; originally announced June 2022.

Comments: To be published in AIED Late Breaking Results 2022

Ultraviolet Spectropolarimetric Diagnostics of Hot Star Magnetospheres

Authors: Asif ud-Doula, M. C. M. Cheung, A. David-Uraz, C. Erba, C. P. Folsom, K. Gayley, Y. Naze, C. Neiner, V. Petit, R. Prinja, M. E. Shultz, N. Sudnik, J. S. Vink, G. A. Wade

Abstract: Several space missions and instruments for UV spectropolarimetry are in preparation, such as the proposed NASA MIDEX Polstar project, the proposed ESA M mission Arago, and the Pollux instrument on the future LUVOIR-like NASA flagship mission. In the frame of Polstar, we have studied the capabilities these observatories would offer to gain information on the magnetic and plasma properties of the ma… ▽ More Several space missions and instruments for UV spectropolarimetry are in preparation, such as the proposed NASA MIDEX Polstar project, the proposed ESA M mission Arago, and the Pollux instrument on the future LUVOIR-like NASA flagship mission. In the frame of Polstar, we have studied the capabilities these observatories would offer to gain information on the magnetic and plasma properties of the magnetospheres of hot stars, helping us test the fundamental hypothesis that magnetospheres should act to rapidly drain angular momentum, thereby spinning the star down, whilst simultaneously reducing the net mass-loss rate. Both effects are expected to lead to dramatic differences in the evolution of magnetic vs. non-magnetic stars. △ Less

Submitted 26 June, 2022; originally announced June 2022.

Comments: Accepted for publication in Astrophysics and Space Science. arXiv admin note: substantial text overlap with arXiv:2111.06434

Report number: ASTR-D-22-00095R3

A Simple Baseline for Video Restoration with Grouped Spatial-temporal Shift

Authors: Dasong Li, Xiaoyu Shi, Yi Zhang, Ka Chun Cheung, Simon See, Xiaogang Wang, Hongwei Qin, Hongsheng Li

Abstract: Video restoration, which aims to restore clear frames from degraded videos, has numerous important applications. The key to video restoration depends on utilizing inter-frame information. However, existing deep learning methods often rely on complicated network architectures, such as optical flow estimation, deformable convolution, and cross-frame self-attention layers, resulting in high computati… ▽ More Video restoration, which aims to restore clear frames from degraded videos, has numerous important applications. The key to video restoration depends on utilizing inter-frame information. However, existing deep learning methods often rely on complicated network architectures, such as optical flow estimation, deformable convolution, and cross-frame self-attention layers, resulting in high computational costs. In this study, we propose a simple yet effective framework for video restoration. Our approach is based on grouped spatial-temporal shift, which is a lightweight and straightforward technique that can implicitly capture inter-frame correspondences for multi-frame aggregation. By introducing grouped spatial shift, we attain expansive effective receptive fields. Combined with basic 2D convolution, this simple framework can effectively aggregate inter-frame information. Extensive experiments demonstrate that our framework outperforms the previous state-of-the-art method, while using less than a quarter of its computational cost, on both video deblurring and video denoising tasks. These results indicate the potential for our approach to significantly reduce computational overhead while maintaining high-quality results. Code is avaliable at https://github.com/dasongli1/Shift-Net. △ Less

Submitted 22 May, 2023; v1 submitted 21 June, 2022; originally announced June 2022.

Comments: Accepted to CVPR2023

Journal ref: 2023 Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

Prospects of a thousand-ion Sn$^{2+}$ Coulomb-crystal clock with sub-$10^{-19}$ inaccuracy

Authors: David R. Leibrandt, Sergey G. Porsev, Charles Cheung, Marianna S. Safronova

Abstract: We propose a many-ion optical atomic clock based on three-dimensional Coulomb crystals of order one thousand Sn$^{2+}$ ions confined in a linear RF Paul trap. Sn$^{2+}$ has a unique combination of features that is not available in previously considered ions: a $^1$S$_0$ $\leftrightarrow$ $^3$P$_0$ clock transition between two states with zero electronic and nuclear angular momentum (I = J = F = 0)… ▽ More We propose a many-ion optical atomic clock based on three-dimensional Coulomb crystals of order one thousand Sn$^{2+}$ ions confined in a linear RF Paul trap. Sn$^{2+}$ has a unique combination of features that is not available in previously considered ions: a $^1$S$_0$ $\leftrightarrow$ $^3$P$_0$ clock transition between two states with zero electronic and nuclear angular momentum (I = J = F = 0) making it immune to nonscalar perturbations, a negative differential polarizability making it possible to operate the trap in a manner such that the two dominant shifts for three-dimensional ion crystals cancel each other, and a laser-accessible transition suitable for direct laser cooling and state readout. We present calculations of the differential polarizability, other relevant atomic properties, and the motion of ions in large Coulomb crystals, in order to estimate the achievable accuracy and precision of Sn$^{2+}$ Coulomb-crystal clocks. △ Less

Submitted 27 March, 2024; v1 submitted 30 May, 2022; originally announced May 2022.

Comments: 15 pages, 5 figures, 3 tables

Global geomagnetic perturbation forecasting using Deep Learning

Authors: Vishal Upendran, Panagiotis Tigas, Banafsheh Ferdousi, Teo Bloch, Mark C. M. Cheung, Siddha Ganju, Asti Bhatt, Ryan M. McGranaghan, Yarin Gal

Abstract: Geomagnetically Induced Currents (GICs) arise from spatio-temporal changes to Earth's magnetic field which arise from the interaction of the solar wind with Earth's magnetosphere, and drive catastrophic destruction to our technologically dependent society. Hence, computational models to forecast GICs globally with large forecast horizon, high spatial resolution and temporal cadence are of increasi… ▽ More Geomagnetically Induced Currents (GICs) arise from spatio-temporal changes to Earth's magnetic field which arise from the interaction of the solar wind with Earth's magnetosphere, and drive catastrophic destruction to our technologically dependent society. Hence, computational models to forecast GICs globally with large forecast horizon, high spatial resolution and temporal cadence are of increasing importance to perform prompt necessary mitigation. Since GIC data is proprietary, the time variability of horizontal component of the magnetic field perturbation (dB/dt) is used as a proxy for GICs. In this work, we develop a fast, global dB/dt forecasting model, which forecasts 30 minutes into the future using only solar wind measurements as input. The model summarizes 2 hours of solar wind measurement using a Gated Recurrent Unit, and generates forecasts of coefficients which are folded with a spherical harmonic basis to enable global forecasts. When deployed, our model produces results in under a second, and generates global forecasts for horizontal magnetic perturbation components at 1-minute cadence. We evaluate our model across models in literature for two specific storms of 5 August 2011 and 17 March 2015, while having a self-consistent benchmark model set. Our model outperforms, or has consistent performance with state-of-the-practice high time cadence local and low time cadence global models, while also outperforming/having comparable performance with the benchmark models. Such quick inferences at high temporal cadence and arbitrary spatial resolutions may ultimately enable accurate forewarning of dB/dt for any place on Earth, resulting in precautionary measures to be taken in an informed manner. △ Less

Submitted 12 May, 2022; originally announced May 2022.

Comments: 23 pages, 8 figures, 5 tables; accepted for publication in AGU: Spaceweather

MaskEval: Weighted MLM-Based Evaluation for Text Summarization and Simplification

Authors: Yu Lu Liu, Rachel Bawden, Thomas Scialom, Benoît Sagot, Jackie Chi Kit Cheung

Abstract: In text summarization and simplification, system outputs must be evaluated along multiple dimensions such as relevance, factual consistency, fluency, and grammaticality, and a wide range of possible outputs could be of high quality. These properties make the development of an adaptable, reference-less evaluation metric both necessary and challenging. We introduce MaskEval, a reference-less metric… ▽ More In text summarization and simplification, system outputs must be evaluated along multiple dimensions such as relevance, factual consistency, fluency, and grammaticality, and a wide range of possible outputs could be of high quality. These properties make the development of an adaptable, reference-less evaluation metric both necessary and challenging. We introduce MaskEval, a reference-less metric for text summarization and simplification that operates by performing masked language modeling (MLM) on the concatenation of the candidate and the source texts. It features an attention-like weighting mechanism to modulate the relative importance of each MLM step, which crucially allows it to be adapted to evaluate different quality dimensions. We demonstrate its effectiveness on English summarization and simplification in terms of correlations with human judgments, and explore transfer scenarios between the two tasks. △ Less

Submitted 13 October, 2022; v1 submitted 24 May, 2022; originally announced May 2022.

Subarcsecond view on the high-redshift blazar GB 1508+5714 by the International LOFAR Telescope

Authors: A. Kappes, P. R. Burd, M. Kadler, G. Ghisellini, E. Bonnassieux, M. Perucho, M. Brüggen, C. C. Cheung, B. Ciardi, E. Gallo, F. Haardt, L. K. Morabito, T. Sbarrato, A. Drabent, J. Harwood, N. Jackson, J. Moldon

Abstract: Studies of the most distant AGNs allow us to test our current understanding of the physics present in radio-jetted AGNs across a range of environments. The decrease in apparent luminosity with distance is the primary difficulty to overcome in the study of these distant AGNs, which requires highly sensitive instruments. Our goal is to employ new long wavelength radio data to better parametrise the… ▽ More Studies of the most distant AGNs allow us to test our current understanding of the physics present in radio-jetted AGNs across a range of environments. The decrease in apparent luminosity with distance is the primary difficulty to overcome in the study of these distant AGNs, which requires highly sensitive instruments. Our goal is to employ new long wavelength radio data to better parametrise the broad-band SED of GB 1508+5714, a high-redshift (z=4.30) AGN. Its high redshift, high intrinsic luminosity, and classification as a blazar allow us to test emission models that consider the efficient cooling of jet electrons via inverse Compton losses in interactions with the dense CMB photon field at high redshifts. A significant detection of this effect in GB 1508+5714 may partly explain the apparent sparsity of high-redshift radio galaxies in wide-field surveys; detections of this kind are only becoming possible with the current generation of SKA precursors. We used international LOFAR telescope to image the long wavelength radio emission around the high-redshift blazar GB 1508+5714 on arcsecond scales at frequencies between 128 MHz and 160 MHz. This allowed us to compare the spatially resolved structure with higher frequency observations, and to construct spectral index maps. The LOFAR image shows a compact unresolved core and two resolved emission regions around 2 arcsec to the east and to the west of the radio core. We find structure consistent with previous VLA observations, as well as a previously unreported emission region to the east. We interpret the arcsecond-scale radio structure of GB 1508+5714 as a FR II-like radio galaxy at a small viewing angle. Our SED modelling shows that a scenario featuring significant quenching effects caused by interaction with the CMB provides a good description of the data, and notably explains the suppressed radio emission. △ Less

Submitted 23 May, 2022; originally announced May 2022.

Comments: 11 pages, 10 figures, 2 tables

MSC Class: 85-02

Journal ref: A&A 663, A44 (2022)

MPPNet: Multi-Frame Feature Intertwining with Proxy Points for 3D Temporal Object Detection

Authors: Xuesong Chen, Shaoshuai Shi, Benjin Zhu, Ka Chun Cheung, Hang Xu, Hongsheng Li

Abstract: Accurate and reliable 3D detection is vital for many applications including autonomous driving vehicles and service robots. In this paper, we present a flexible and high-performance 3D detection framework, named MPPNet, for 3D temporal object detection with point cloud sequences. We propose a novel three-hierarchy framework with proxy points for multi-frame feature encoding and interactions to ach… ▽ More Accurate and reliable 3D detection is vital for many applications including autonomous driving vehicles and service robots. In this paper, we present a flexible and high-performance 3D detection framework, named MPPNet, for 3D temporal object detection with point cloud sequences. We propose a novel three-hierarchy framework with proxy points for multi-frame feature encoding and interactions to achieve better detection. The three hierarchies conduct per-frame feature encoding, short-clip feature fusion, and whole-sequence feature aggregation, respectively. To enable processing long-sequence point clouds with reasonable computational resources, intra-group feature mixing and inter-group feature attention are proposed to form the second and third feature encoding hierarchies, which are recurrently applied for aggregating multi-frame trajectory features. The proxy points not only act as consistent object representations for each frame, but also serve as the courier to facilitate feature interaction between frames. The experiments on large Waymo Open dataset show that our approach outperforms state-of-the-art methods with large margins when applied to both short (e.g., 4-frame) and long (e.g., 16-frame) point cloud sequences. Code is available at https://github.com/open-mmlab/OpenPCDet. △ Less

Submitted 2 September, 2022; v1 submitted 12 May, 2022; originally announced May 2022.

Comments: Accepted by ECCV 2022

Calculation of energies and hyperfine structure constants of 233U^+ and 233U

Authors: S. G. Porsev, C. Cheung, M. S. Safronova

Abstract: We carried out calculations of the energies and magnetic dipole hyperfine structure constants of the low-lying states of 233U^+ and 233U using two different approaches. With six valence electrons and a very heavy core, uranium represents a major challenge for precision atomic theory even using large-scale computational resources. The first approach combines configuration interaction (CI) with a me… ▽ More We carried out calculations of the energies and magnetic dipole hyperfine structure constants of the low-lying states of 233U^+ and 233U using two different approaches. With six valence electrons and a very heavy core, uranium represents a major challenge for precision atomic theory even using large-scale computational resources. The first approach combines configuration interaction (CI) with a method allowing us to include core-valence correlations to all orders of the perturbation theory over residual Coulomb interaction. The second approach is a pure CI method which allows the use of different initial approximations. We present a detailed analysis of all calculated properties and discuss the advantages and disadvantages of each of these methods. We report a preliminary value of the U nuclear magnetic moment and outline the need for further experiments. △ Less

Submitted 13 October, 2022; v1 submitted 27 April, 2022; originally announced April 2022.

Comments: 7 pages

Non-perturbative Double Copy in Flatland

Authors: Clifford Cheung, James Mangan, Julio Parra-Martinez, Nabha Shah

Abstract: We derive a non-perturbative, Lagrangian-level formulation of the double copy in two spacetime dimensions. Our results elucidate the field theoretic underpinnings of the double copy in a broad class of scalar theories which can include masses and higher-dimension operators. An immediate corollary is the amplitudes-level double copy at all orders in perturbation theory. Applied to certain integrabl… ▽ More We derive a non-perturbative, Lagrangian-level formulation of the double copy in two spacetime dimensions. Our results elucidate the field theoretic underpinnings of the double copy in a broad class of scalar theories which can include masses and higher-dimension operators. An immediate corollary is the amplitudes-level double copy at all orders in perturbation theory. Applied to certain integrable models, the double copy defines an isomorphism between Lax connections, Wilson lines, and infinite towers of conserved currents. We also implement the double copy at the level of non-perturbative classical solutions, both analytically and numerically, and present a generalization of the double copy map that includes a fixed tower of higher-dimension corrections given by the Moyal algebra. △ Less

Submitted 6 December, 2022; v1 submitted 14 April, 2022; originally announced April 2022.

Comments: Updated to match published version. 9 pages + 1 figure + 1 animation also accessible at https://bit.ly/3OdGIo4

Report number: CALT-TH 2022-015

Journal ref: Phys. Rev. Lett. 129 (2022) 22, 221602

Using Interactive Feedback to Improve the Accuracy and Explainability of Question Answering Systems Post-Deployment

Authors: Zichao Li, Prakhar Sharma, Xing Han Lu, Jackie C. K. Cheung, Siva Reddy

Abstract: Most research on question answering focuses on the pre-deployment stage; i.e., building an accurate model for deployment. In this paper, we ask the question: Can we improve QA systems further \emph{post-}deployment based on user interactions? We focus on two kinds of improvements: 1) improving the QA system's performance itself, and 2) providing the model with the ability to explain the correctnes… ▽ More Most research on question answering focuses on the pre-deployment stage; i.e., building an accurate model for deployment. In this paper, we ask the question: Can we improve QA systems further \emph{post-}deployment based on user interactions? We focus on two kinds of improvements: 1) improving the QA system's performance itself, and 2) providing the model with the ability to explain the correctness or incorrectness of an answer. We collect a retrieval-based QA dataset, FeedbackQA, which contains interactive feedback from users. We collect this dataset by deploying a base QA system to crowdworkers who then engage with the system and provide feedback on the quality of its answers. The feedback contains both structured ratings and unstructured natural language explanations. We train a neural model with this feedback data that can generate explanations and re-score answer candidates. We show that feedback data not only improves the accuracy of the deployed QA system but also other stronger non-deployed systems. The generated explanations also help users make informed decisions about the correctness of answers. Project page: https://mcgill-nlp.github.io/feedbackqa/ △ Less

Submitted 6 April, 2022; originally announced April 2022.

Comments: ACL 2022 Findings

Journal ref: Findings of the Association for Computational Linguistics: ACL (2022) 926-937

M5-branes wrapped on four-dimensional orbifolds

Authors: K. C. Matthew Cheung, Jacob H. T. Fry, Jerome P. Gauntlett, James Sparks

Abstract: We construct supersymmetric $AdS_3$ solutions of $D=11$ supergravity, dual to $d=2$, $\mathcal{N}=(0,2)$ SCFTs, that are associated with M5-branes wrapping two different four-dimensional orbifolds. In one case the orbifold is a spindle fibred over another spindle, while in the other it is a spindle fibred over a Riemann surface with genus $g>1$. We show that the central charges of the $d=2$ SCFTs… ▽ More We construct supersymmetric $AdS_3$ solutions of $D=11$ supergravity, dual to $d=2$, $\mathcal{N}=(0,2)$ SCFTs, that are associated with M5-branes wrapping two different four-dimensional orbifolds. In one case the orbifold is a spindle fibred over another spindle, while in the other it is a spindle fibred over a Riemann surface with genus $g>1$. We show that the central charges of the $d=2$ SCFTs calculated from the gravity solutions agree with field theory computations using anomaly polynomials. The new $D=11$ solutions are obtained after constructing a new consistent Kaluza-Klein truncation of maximal $D=7$ gauged supergravity reduced on a spindle down to $D=5$ minimal gauged supergravity. △ Less

Submitted 5 August, 2022; v1 submitted 6 April, 2022; originally announced April 2022.

Comments: 37 pages. Very minor changes. Published version

Report number: Imperial/TP/2022/JG/01

Why Exposure Bias Matters: An Imitation Learning Perspective of Error Accumulation in Language Generation

Authors: Kushal Arora, Layla El Asri, Hareesh Bahuleyan, Jackie Chi Kit Cheung

Abstract: Current language generation models suffer from issues such as repetition, incoherence, and hallucinations. An often-repeated hypothesis is that this brittleness of generation models is caused by the training and the generation procedure mismatch, also referred to as exposure bias. In this paper, we verify this hypothesis by analyzing exposure bias from an imitation learning perspective. We show th… ▽ More Current language generation models suffer from issues such as repetition, incoherence, and hallucinations. An often-repeated hypothesis is that this brittleness of generation models is caused by the training and the generation procedure mismatch, also referred to as exposure bias. In this paper, we verify this hypothesis by analyzing exposure bias from an imitation learning perspective. We show that exposure bias leads to an accumulation of errors, analyze why perplexity fails to capture this accumulation, and empirically show that this accumulation results in poor generation quality. Source code to reproduce these experiments is available at https://github.com/kushalarora/quantifying_exposure_bias △ Less

Submitted 9 January, 2023; v1 submitted 3 April, 2022; originally announced April 2022.

Comments: Accepted in Findings of ACL 2022. v2: Equation 7 updated, typo fixes

FlowFormer: A Transformer Architecture for Optical Flow

Authors: Zhaoyang Huang, Xiaoyu Shi, Chao Zhang, Qiang Wang, Ka Chun Cheung, Hongwei Qin, Jifeng Dai, Hongsheng Li

Abstract: We introduce optical Flow transFormer, dubbed as FlowFormer, a transformer-based neural network architecture for learning optical flow. FlowFormer tokenizes the 4D cost volume built from an image pair, encodes the cost tokens into a cost memory with alternate-group transformer (AGT) layers in a novel latent space, and decodes the cost memory via a recurrent transformer decoder with dynamic positio… ▽ More We introduce optical Flow transFormer, dubbed as FlowFormer, a transformer-based neural network architecture for learning optical flow. FlowFormer tokenizes the 4D cost volume built from an image pair, encodes the cost tokens into a cost memory with alternate-group transformer (AGT) layers in a novel latent space, and decodes the cost memory via a recurrent transformer decoder with dynamic positional cost queries. On the Sintel benchmark, FlowFormer achieves 1.159 and 2.088 average end-point-error (AEPE) on the clean and final pass, a 16.5% and 15.5% error reduction from the best published result (1.388 and 2.47). Besides, FlowFormer also achieves strong generalization performance. Without being trained on Sintel, FlowFormer achieves 1.01 AEPE on the clean pass of Sintel training set, outperforming the best published result (1.29) by 21.7%. △ Less

Submitted 21 September, 2022; v1 submitted 30 March, 2022; originally announced March 2022.

Comments: Accepted to ECCV 2022. Project Page: https://drinkingcoder.github.io/publication/flowformer/

Type IIA embeddings of $D=5$ minimal gauged supergravity via Non-Abelian T-duality

Authors: K. C. Matthew Cheung, Rahim Leung

Abstract: In this note, we construct explicit Type IIA uplifts of $D=5$ minimal gauged supergravity, by T-dualising known Type IIB uplifts on $N_5 = S^5$, $T^{1,1}$ and $Y^{p,q}$ along their $SU(2)$ isometries. When the $D=5$ gauge field is set to zero, our uplifts recover precisely the known non-Abelian T-duals of the $AdS_5\times N_5$ solutions. As an application, we obtain new supersymmetric… ▽ More In this note, we construct explicit Type IIA uplifts of $D=5$ minimal gauged supergravity, by T-dualising known Type IIB uplifts on $N_5 = S^5$, $T^{1,1}$ and $Y^{p,q}$ along their $SU(2)$ isometries. When the $D=5$ gauge field is set to zero, our uplifts recover precisely the known non-Abelian T-duals of the $AdS_5\times N_5$ solutions. As an application, we obtain new supersymmetric $AdS_3\timesΣ\times M_5$ solutions in Type IIA, where $Σ= \mathbb{WCP}^1_{[n_-,n_+]}$ is a weighted projective space. Existing holographic results of T-dualised AdS solutions suggest that our solutions capture features of $d = 2$ SCFTs with $\mathcal{N}=(0, 2)$ supersymmetry. △ Less

Submitted 28 March, 2022; originally announced March 2022.

Comments: 41 pages, 1 figure

Coronal Mass Ejections and Dimmings: A Comparative Study using MHD Simulations and SDO Observations

Authors: Meng Jin, Mark C. M. Cheung, Marc L. DeRosa, Nariaki V. Nitta, Carolus J. Schrijver

Abstract: Solar coronal dimmings have been observed extensively in the past two decades. Due to their close association with coronal mass ejections (CMEs), there is a critical need to improve our understanding of the physical processes that cause dimmings as well as their relationship with CMEs. In this study, we investigate coronal dimmings by combining simulation and observational efforts. By utilizing a… ▽ More Solar coronal dimmings have been observed extensively in the past two decades. Due to their close association with coronal mass ejections (CMEs), there is a critical need to improve our understanding of the physical processes that cause dimmings as well as their relationship with CMEs. In this study, we investigate coronal dimmings by combining simulation and observational efforts. By utilizing a data-constrained global magnetohydrodynamics model (AWSoM: Alfven-wave Solar Model), we simulate coronal dimmings resulting from different CME energetics and flux rope configurations. We synthesize the emissions of different EUV spectral bands/lines and compare with SDO/AIA and EVE observations. A detailed analysis of the simulation and observation data suggests that the transient dimming / brightening are related to plasma heating processes, while the long-lasting core and remote dimmings are caused by mass loss process induced by the CME. Moreover, the interaction between the erupting flux rope with different orientations and the global solar corona could significantly influence the coronal dimming patterns. Using metrics such as dimming depth and dimming slope, we investigate the relationship between dimmings and CME properties (e.g., CME mass, CME speed) in the simulation. Our result suggests that coronal dimmings encode important information about the associated CMEs, which provides a physical basis for detecting stellar CMEs from distant solar-like stars. △ Less

Submitted 25 February, 2022; originally announced February 2022.

Comments: 16 pages, 9 figures, accepted for publication in ApJ

Geometry-Kinematics Duality

Authors: Clifford Cheung, Andreas Helset, Julio Parra-Martinez

Abstract: We propose a mapping between geometry and kinematics that implies the classical equivalence of any theory of massless bosons -- including spin and exhibiting arbitrary derivative or potential interactions -- to a nonlinear sigma model (NLSM) with a momentum-dependent metric in field space. From this kinematic metric we construct a corresponding kinematic connection, covariant derivative, and curva… ▽ More We propose a mapping between geometry and kinematics that implies the classical equivalence of any theory of massless bosons -- including spin and exhibiting arbitrary derivative or potential interactions -- to a nonlinear sigma model (NLSM) with a momentum-dependent metric in field space. From this kinematic metric we construct a corresponding kinematic connection, covariant derivative, and curvature, all of which transform appropriately under general field redefinitions, even including derivatives. We show explicitly how all tree-level on-shell scattering amplitudes of massless bosons are equal to those of the NLSM via the replacement of geometry with kinematics. Lastly, we describe how the recently introduced geometric soft theorem of the NLSM, which universally encodes all leading and subleading soft scalar theorems, also captures the soft photon theorems. △ Less

Submitted 14 February, 2022; originally announced February 2022.

Comments: 5 pages

Report number: CALT-TH 2022-006

Massively parallel pixel-by-pixel nanophotonic optimization using a Green's function formalism

Authors: Jiahui Wang, Alfred K. C. Cheung, Aleksandra Spyra, Ian A. D. Williamson, Jian Guan, Martin F. Schubert

Abstract: We introduce an efficient parallelization scheme to implement pixel-by-pixel nanophotonic optimization using a Green's function based formalism. The crucial insight in our proposal is the reframing of the optimization algorithm as a large-scale data processing pipeline, which allows for the efficient distribution of computational tasks across thousands of workers. We demonstrate the utility of our… ▽ More We introduce an efficient parallelization scheme to implement pixel-by-pixel nanophotonic optimization using a Green's function based formalism. The crucial insight in our proposal is the reframing of the optimization algorithm as a large-scale data processing pipeline, which allows for the efficient distribution of computational tasks across thousands of workers. We demonstrate the utility of our implementation by exercising it to optimize a high numerical aperture focusing metalens at problem sizes that would otherwise be far out of reach for the Green's function based method. Finally, we highlight the connection to powerful ideas from reinforcement learning as a natural corollary of reinterpreting the nanophotonic inverse design problem as a graph traversal enabled by the pixel-by-pixel optimization paradigm. △ Less

Submitted 10 February, 2022; originally announced February 2022.

Comments: 10 pages, 7 figures

Inverse design of photonic devices with strict foundry fabrication constraints

Authors: Martin F. Schubert, Alfred K. C. Cheung, Ian A. D. Williamson, Aleksandra Spyra, David H. Alexander

Abstract: We introduce a new method for inverse design of nanophotonic devices which guarantees that resulting designs satisfy strict length scale constraints - including minimum width and spacing constraints required by commercial semiconductor foundries. The method adopts several concepts from machine learning to transform the problem of topology optimization with strict length scale constraints to an unc… ▽ More We introduce a new method for inverse design of nanophotonic devices which guarantees that resulting designs satisfy strict length scale constraints - including minimum width and spacing constraints required by commercial semiconductor foundries. The method adopts several concepts from machine learning to transform the problem of topology optimization with strict length scale constraints to an unconstrained stochastic gradient optimization problem. Specifically, we introduce a conditional generator for feasible designs and adopt a straight-through estimator for backpropagation of gradients to a latent design. We demonstrate the performance and reliability of our method by designing several common integrated photonic components. △ Less

Submitted 13 June, 2022; v1 submitted 30 January, 2022; originally announced January 2022.

Comments: 16 pages, 17 figures

Journal ref: ACS Photonics, vol. 9, no. 7, pp. 2327-2336, Jun. 2022

Incremental Fermi Large Area Telescope Fourth Source Catalog

Authors: Fermi-LAT collaboration, :, Soheila Abdollahi, Fabio Acero, Luca Baldini, Jean Ballet, Denis Bastieri, Ronaldo Bellazzini, Bijan Berenji, Alessandra Berretta, Elisabetta Bissaldi, Roger D. Blandford, Elliott Bloom, Raffaella Bonino, Ari Brill, Richard J. Britto, Philippe Bruel, Toby H. Burnett, Sara Buson, Rob A. Cameron, Regina Caputo, Patrizia A. Caraveo, Daniel Castro, Sylvain Chaty, Teddy C. Cheung , et al. (116 additional authors not shown)

Abstract: We present an incremental version (4FGL-DR3, for Data Release 3) of the fourth Fermi-LAT catalog of gamma-ray sources. Based on the first twelve years of science data in the energy range from 50 MeV to 1 TeV, it contains 6658 sources. The analysis improves on that used for the 4FGL catalog over eight years of data: more sources are fit with curved spectra, we introduce a more robust spectral param… ▽ More We present an incremental version (4FGL-DR3, for Data Release 3) of the fourth Fermi-LAT catalog of gamma-ray sources. Based on the first twelve years of science data in the energy range from 50 MeV to 1 TeV, it contains 6658 sources. The analysis improves on that used for the 4FGL catalog over eight years of data: more sources are fit with curved spectra, we introduce a more robust spectral parameterization for pulsars, and we extend the spectral points to 1 TeV. The spectral parameters, spectral energy distributions, and associations are updated for all sources. Light curves are rebuilt for all sources with 1 yr intervals (not 2 month intervals). Among the 5064 original 4FGL sources, 16 were deleted, 112 are formally below the detection threshold over 12 yr (but are kept in the list), while 74 are newly associated, 10 have an improved association, and seven associations were withdrawn. Pulsars are split explicitly between young and millisecond pulsars. Pulsars and binaries newly detected in LAT sources, as well as more than 100 newly classified blazars, are reported. We add three extended sources and 1607 new point sources, mostly just above the detection threshold, among which eight are considered identified, and 699 have a plausible counterpart at other wavelengths. We discuss degree-scale residuals to the global sky model and clusters of soft unassociated point sources close to the Galactic plane, which are possibly related to limitations of the interstellar emission model and missing extended sources. △ Less

Submitted 10 May, 2022; v1 submitted 26 January, 2022; originally announced January 2022.

Comments: accepted in ApJS; follow-up paper to 1902.10045

Journal ref: ApJS 260, 53 (2022)

Multivariate matrix-exponential affine mixtures and their applications in risk theory

Authors: Eric C. K. Cheung, Oscar Peralta, Jae-Kyung Woo

Abstract: In this paper, a class of multivariate matrix-exponential affine mixtures with matrix-exponential marginals is proposed. The class is shown to possess various attractive properties such as closure under size-biased Esscher transform, order statistics, residual lifetime and higher order equilibrium distributions. This allows for explicit calculations of various actuarial quantities of interest. The… ▽ More In this paper, a class of multivariate matrix-exponential affine mixtures with matrix-exponential marginals is proposed. The class is shown to possess various attractive properties such as closure under size-biased Esscher transform, order statistics, residual lifetime and higher order equilibrium distributions. This allows for explicit calculations of various actuarial quantities of interest. The results are applied in a wide range of actuarial problems including multivariate risk measures, aggregate loss, large claims reinsurance, weighted premium calculations and risk capital allocation. Furthermore, a multiplicative background risk model with dependent risks is considered and its capital allocation rules are provided as well. We finalize by discussing a calibration scheme based on complete data and potential avenues of research. △ Less

Submitted 17 December, 2021; originally announced January 2022.

New Measurement Resolves Key Astrophysical Fe XVII Oscillator Strength Problem

Authors: Steffen Kühn, Charles Cheung, Natalia S. Oreshkina, René Steinbrügge, Moto Togawa, Sonja Bernitt, Lukas Berger, Jens Buck, Moritz Hoesch, Jörn Seltmann, Florian Trinter, Christoph H. Keitel, Mikhail G. Kozlov, Sergey G. Porsev, Ming Feng Gu, F. Scott Porter, Thomas Pfeifer, Maurice A. Leutenegger, Zoltán Harman, Marianna S. Safronova, José R. Crespo López-Urrutia, Chintan Shah

Abstract: One of the most enduring and intensively studied problems of X-ray astronomy is the disagreement of state-of-the art theory and observations for the intensity ratio of two Fe XVII transitions of crucial value for plasma diagnostics, dubbed 3C and 3D. We unravel this conundrum at the PETRA III synchrotron facility by increasing the resolving power two and a half times and the signal-to-noise ratio… ▽ More One of the most enduring and intensively studied problems of X-ray astronomy is the disagreement of state-of-the art theory and observations for the intensity ratio of two Fe XVII transitions of crucial value for plasma diagnostics, dubbed 3C and 3D. We unravel this conundrum at the PETRA III synchrotron facility by increasing the resolving power two and a half times and the signal-to-noise ratio thousand-fold compared to our previous work. The Lorentzian wings had hitherto been indistinguishable from the background and were thus not modeled, resulting in a biased line-strength estimation. The present experimental oscillator-strength ratio $R_\mathrm{exp}=f_{\mathrm{3C}}/f_{\mathrm{3D}}=3.51(2)_{\mathrm{stat}}(7)_{\mathrm{sys}}$ agrees with our state-of-the-art calculation of $R_\mathrm{th}=3.55(2)$, as well as with some previous theoretical predictions. To further rule out any uncertainties associated with the measured ratio, we also determined the individual natural linewidths and oscillator strengths of 3C and 3D transitions, which also agree well with the theory. This finally resolves the decades-old mystery of Fe XVII oscillator strengths. △ Less

Submitted 6 December, 2022; v1 submitted 22 January, 2022; originally announced January 2022.

Comments: Main manuscript and supplemental material at https://journals.aps.org/prl/supplemental/10.1103/PhysRevLett.129.245001/LN17392_Supplemental_Material.pdf

Journal ref: Physical Review Letters 129, 245001 (2022)

On-shell Correlators and Color-Kinematics Duality in Curved Symmetric Spacetimes

Authors: Clifford Cheung, Julio Parra-Martinez, Allic Sivaramakrishnan

Abstract: We define a perturbatively calculable quantity--the on-shell correlator--which furnishes a unified description of particle dynamics in curved spacetime. Specializing to the case of flat and anti-de Sitter space, on-shell correlators coincide precisely with on-shell scattering amplitudes and boundary correlators, respectively. Remarkably, we find that symmetric manifolds admit a generalization of o… ▽ More We define a perturbatively calculable quantity--the on-shell correlator--which furnishes a unified description of particle dynamics in curved spacetime. Specializing to the case of flat and anti-de Sitter space, on-shell correlators coincide precisely with on-shell scattering amplitudes and boundary correlators, respectively. Remarkably, we find that symmetric manifolds admit a generalization of on-shell kinematics in which the corresponding momenta are literally the isometry generators of the spacetime acting on the external kinematic data. These isometric momenta are intrinsically non-commutative but exhibit on-shell conditions that are identical to those of flat space, thus providing a common language for computing and representing on-shell correlators which is agnostic about the underlying geometry. Afterwards, we compute tree-level on-shell correlators for biadjoint scalar (BAS) theory and the nonlinear sigma model (NLSM) and learn that color-kinematics duality is manifested at the level of fields under a mapping of the color algebra to the algebra of gauged isometries on the spacetime manifold. Last but not least, we present a field theoretic derivation of the fundamental BCJ relations for on-shell correlators following from the existence of certain conserved currents in BAS theory and the NLSM. △ Less

Submitted 13 January, 2022; originally announced January 2022.

Comments: 42 pages + refs

Report number: CALT-TH-2022-002

Does Pre-training Induce Systematic Inference? How Masked Language Models Acquire Commonsense Knowledge

Authors: Ian Porada, Alessandro Sordoni, Jackie Chi Kit Cheung

Abstract: Transformer models pre-trained with a masked-language-modeling objective (e.g., BERT) encode commonsense knowledge as evidenced by behavioral probes; however, the extent to which this knowledge is acquired by systematic inference over the semantics of the pre-training corpora is an open question. To answer this question, we selectively inject verbalized knowledge into the minibatches of a BERT mod… ▽ More Transformer models pre-trained with a masked-language-modeling objective (e.g., BERT) encode commonsense knowledge as evidenced by behavioral probes; however, the extent to which this knowledge is acquired by systematic inference over the semantics of the pre-training corpora is an open question. To answer this question, we selectively inject verbalized knowledge into the minibatches of a BERT model during pre-training and evaluate how well the model generalizes to supported inferences. We find generalization does not improve over the course of pre-training, suggesting that commonsense knowledge is acquired from surface-level, co-occurrence patterns rather than induced, systematic reasoning. △ Less

Submitted 15 December, 2021; originally announced December 2021.

Laser Spectroscopy of the y$^7$P$_J^{\circ}$ states of Cr I

Authors: E. B. Norrgard, D. S. Barker, S. P. Eckel, S. G. Porsev, C. Cheung, M. G. Kozlov, I. I. Tupitsyn, M. S. Safronova

Abstract: Here we report measured and calculated values of decay rates of the 3d$^4$($^5$D)4s4p($^3$P$^{\rm{o}}$)\ y$^7$P$^{\rm{o}}_{2,3,4}$ states of Cr I. The decay rates are measured using time-correlated single photon counting with roughly 1% total uncertainty. In addition, the isotope shifts for these transitions are measured by laser induced fluorescence to roughly 0.5% uncertainty. The decay rate cal… ▽ More Here we report measured and calculated values of decay rates of the 3d$^4$($^5$D)4s4p($^3$P$^{\rm{o}}$)\ y$^7$P$^{\rm{o}}_{2,3,4}$ states of Cr I. The decay rates are measured using time-correlated single photon counting with roughly 1% total uncertainty. In addition, the isotope shifts for these transitions are measured by laser induced fluorescence to roughly 0.5% uncertainty. The decay rate calculations are carried out by a hybrid approach that combines configuration interaction and the linearized coupled cluster method (CI+all-order method). The measurements provide a much needed precision benchmark for testing the accuracy of the CI+all-order approach for such complicated systems with six valence electrons, allowing to significantly expand its applicability. These measurements also demonstrate operation of a cryogenic buffer gas beam source for future experiments with MgF molecules toward quantum blackbody thermometry. △ Less

Submitted 23 November, 2021; originally announced November 2021.

Ultraviolet Spectropolarimetry With Polstar: Hot Star Magnetospheres

Authors: M. E. Shultz, R. Casini, M. C. M. Cheung, A. David-Uraz, T. del Pino Alemán, C. Erba, C. P. Folsom, K. Gayley, R. Ignace, Z. Keszthelyi, O. Kochukhov, Y. Nazé, C. Neiner, M. Oksala, V. Petit, P. A. Scowen, N. Sudnik, A. ud-Doula, J. S. Vink, G. A. Wade

Abstract: Polstar is a proposed NASA MIDEX space telescope that will provide high-resolution, simultaneous full-Stokes spectropolarimetry in the far ultraviolet, together with low-resolution linear polarimetry in the near ultraviolet. In this white paper, we describe the unprecedented capabilities this observatory would offer in order to obtain unique information on the magnetic and plasma properties of the… ▽ More Polstar is a proposed NASA MIDEX space telescope that will provide high-resolution, simultaneous full-Stokes spectropolarimetry in the far ultraviolet, together with low-resolution linear polarimetry in the near ultraviolet. In this white paper, we describe the unprecedented capabilities this observatory would offer in order to obtain unique information on the magnetic and plasma properties of the magnetospheres of hot stars. This would enable a test of the fundamental hypothesis that magnetospheres should act to rapidly drain angular momentum, thereby spinning the star down, whilst simultaneously reducing the net mass-loss rate. Both effects are expected to lead to dramatic differences in the evolution of magnetic vs. non-magnetic stars. △ Less

Submitted 9 December, 2021; v1 submitted 11 November, 2021; originally announced November 2021.

Comments: White paper, 40 pages

The solar internetwork. III. Unipolar versus bipolar flux appearance

Authors: Milan Gošić, Luis R. Bellot Rubio, Mark C. M. Cheung, David Orozco Suárez, Yukio Katsukawa, Jose Carlos Del Toro Iniesta

Abstract: Small-scale internetwork (IN) magnetic fields are considered to be the main building blocks of the quiet Sun magnetism. For this reason, it is crucial to understand how they appear on the solar surface. Here, we employ a high-resolution, high-sensitivity, long-duration Hinode/NFI magnetogram sequence to analyze the appearance modes and spatio-temporal evolution of individual IN magnetic elements i… ▽ More Small-scale internetwork (IN) magnetic fields are considered to be the main building blocks of the quiet Sun magnetism. For this reason, it is crucial to understand how they appear on the solar surface. Here, we employ a high-resolution, high-sensitivity, long-duration Hinode/NFI magnetogram sequence to analyze the appearance modes and spatio-temporal evolution of individual IN magnetic elements inside a supergranular cell at the disk center. From identification of flux patches and magnetofrictional simulations, we show that there are two distinct populations of IN flux concentrations: unipolar and bipolar features. Bipolar features tend to be bigger and stronger than unipolar features. They also live longer and carry more flux per feature. Both types of flux concentrations appear uniformly over the solar surface. However, we argue that bipolar features truly represent the emergence of new flux on the solar surface, while unipolar features seem to be formed by coalescence of background flux. Magnetic bipoles appear at a faster rate than unipolar features (68 as opposed to 55 Mx cm$^{-2}$ day$^{-1}$), and provide about 70% of the total instantaneous IN flux detected in the interior of the supergranule. △ Less

Submitted 4 November, 2021; originally announced November 2021.

Comments: 14 pages, 11 figures. Accepted for publication in ApJ. Animations are available at https://www.lmsal.com/~mgosic/download/animations_apj_2021b.tgz

Geometric Soft Theorems

Authors: Clifford Cheung, Andreas Helset, Julio Parra-Martinez

Abstract: We derive a universal soft theorem for every scattering amplitude with at least one massless particle in an arbitrary theory of scalars. Our results follow from the geometry of field space and are valid for any choice of mass spectrum, potential terms, and higher-derivative interactions. For a vanishing potential, the soft limit of every amplitude is equal to the field-space covariant derivative o… ▽ More We derive a universal soft theorem for every scattering amplitude with at least one massless particle in an arbitrary theory of scalars. Our results follow from the geometry of field space and are valid for any choice of mass spectrum, potential terms, and higher-derivative interactions. For a vanishing potential, the soft limit of every amplitude is equal to the field-space covariant derivative of an amplitude with one fewer particle. Furthermore, the Adler zero and the dilaton soft theorem are special cases of our results. We also discuss more exotic scenarios in which the soft limit is non-trivial but still universal. Last but not least, we derive new theorems for multiple-soft limits which directly probe the field-space curvature, as well as on-shell recursion relations applicable to two-derivative scalar field theories exhibiting no symmetries whatsoever. △ Less

Submitted 4 November, 2021; originally announced November 2021.

Comments: 32 pages + refs, 2 figures

Report number: CALT-TH-2021-038

Atomistic hartree theory and crystal field of twisted double bilayer graphene near the magic angle

Authors: Christopher T. S. Cheung, Zachary A. H. Goodwin, Valerio Vitale, Johannes Lischner, Arash A. Mostofi

Abstract: Twisted double bilayer graphene (tDBLG) is a moiré material that has recently generated significant interest because of the observation of correlated phases near the magic angle. We carry out atomistic Hartree theory calculations to study the role of electron-electron interactions in the normal state. In contrast to twisted bilayer graphene (tBLG), we find that such interactions do not result in s… ▽ More Twisted double bilayer graphene (tDBLG) is a moiré material that has recently generated significant interest because of the observation of correlated phases near the magic angle. We carry out atomistic Hartree theory calculations to study the role of electron-electron interactions in the normal state. In contrast to twisted bilayer graphene (tBLG), we find that such interactions do not result in significant doping-dependent deformations of the electronic band structure. However, interactions play an important role for the electronic structure in the presence of a perpendicular electric field as they screen the external field. Finally, we analyze the contribution of the Hartree potential to the crystal field, i.e. the on-site energy difference between the inner and outer layers. We find that the on-site energy obtained from Hartree theory has the same sign, but a smaller magnitude compared to previous studies in which the on-site energy was determined by fitting tight-binding results to ab initio density-functional theory (DFT) band structures. To understand this quantitative difference, we analyze the ab initio Kohn-Sham potential obtained from DFT and find that a subtle interplay of electron-electron and electron-ion interactions determines the magnitude of the on-site potential. △ Less

Submitted 9 March, 2022; v1 submitted 4 November, 2021; originally announced November 2021.

Comments: 10 pages, 8 figures

Variability and Spectral Characteristics of Three Flaring Gamma-ray Quasars Observed by VERITAS and Fermi-LAT

Authors: C. B. Adams, J. Batshoun, W. Benbow, A. Brill, J. H. Buckley, M. Capasso, B. Cavins, J. L. Christiansen, P. Coppi, M. Errando, K. A Farrell, Q. Feng, J. P. Finley, G. M. Foote, L. Fortson, A. Furniss, A. Gent, C. Giuri, D. Hanna, T. Hassan, O. Hervet, J. Holder, M. Houck, T. B. Humensky, W. Jin , et al. (41 additional authors not shown)

Abstract: Flat spectrum radio quasars (FSRQs) are the most luminous blazars at GeV energies, but only rarely emit detectable fluxes of TeV gamma rays, typically during bright GeV flares. We explore the gamma-ray variability and spectral characteristics of three FSRQs that have been observed at GeV and TeV energies by Fermi-LAT and VERITAS, making use of almost 100 hours of VERITAS observations spread over 1… ▽ More Flat spectrum radio quasars (FSRQs) are the most luminous blazars at GeV energies, but only rarely emit detectable fluxes of TeV gamma rays, typically during bright GeV flares. We explore the gamma-ray variability and spectral characteristics of three FSRQs that have been observed at GeV and TeV energies by Fermi-LAT and VERITAS, making use of almost 100 hours of VERITAS observations spread over 10 years: 3C 279, PKS 1222+216, and Ton 599. We explain the GeV flux distributions of the sources in terms of a model derived from a stochastic differential equation describing fluctuations in the magnetic field in the accretion disk, and estimate the timescales of magnetic flux accumulation and stochastic instabilities in their accretion disks. We identify distinct flares using a procedure based on Bayesian blocks and analyze their daily and sub-daily variability and gamma-ray energy spectra. Using observations from VERITAS as well as Fermi, Swift, and the Steward Observatory, we model the broadband spectral energy distributions of PKS 1222+216 and Ton 599 during VHE-detected flares in 2014 and 2017, respectively, strongly constraining the jet Doppler factors and gamma-ray emission region locations during these events. Finally, we place theoretical constraints on the potential production of PeV-scale neutrinos during these VHE flares. △ Less

Submitted 25 October, 2021; originally announced October 2021.

Comments: 34 pages, 13 figures. Accepted for publication in the Astrophysical Journal

Achieving the Pareto Frontier of Regret Minimization and Best Arm Identification in Multi-Armed Bandits

Authors: Zixin Zhong, Wang Chi Cheung, Vincent Y. F. Tan

Abstract: We study the Pareto frontier of two archetypal objectives in multi-armed bandits, namely, regret minimization (RM) and best arm identification (BAI) with a fixed horizon. It is folklore that the balance between exploitation and exploration is crucial for both RM and BAI, but exploration is more critical in achieving the optimal performance for the latter objective. To this end, we design and analy… ▽ More We study the Pareto frontier of two archetypal objectives in multi-armed bandits, namely, regret minimization (RM) and best arm identification (BAI) with a fixed horizon. It is folklore that the balance between exploitation and exploration is crucial for both RM and BAI, but exploration is more critical in achieving the optimal performance for the latter objective. To this end, we design and analyze the BoBW-lil'UCB$(γ)$ algorithm. Complementarily, by establishing lower bounds on the regret achievable by any algorithm with a given BAI failure probability, we show that (i) no algorithm can simultaneously perform optimally for both the RM and BAI objectives, and (ii) BoBW-lil'UCB$(γ)$ achieves order-wise optimal performance for RM or BAI under different values of $γ$. Our work elucidates the trade-off more precisely by showing how the constants in previous works depend on certain hardness parameters. Finally, we show that BoBW-lil'UCB outperforms a close competitor UCB$_α$ (Degenne et al., 2019) in terms of the time complexity and the regret on diverse datasets such as MovieLens and Published Kinase Inhibitor Set. △ Less

Submitted 9 June, 2023; v1 submitted 16 October, 2021; originally announced October 2021.

Comments: 43 pages, 10 figures

The Magnetic Origin of Solar Campfires

Authors: Navdeep K. Panesar, Sanjiv K. Tiwari, David Berghmans, Mark C. M. Cheung, Daniel Muller, Frederic Auchere, Andrei Zhukov

Abstract: Solar campfires are fine-scale heating events, recently observed by Extreme Ultraviolet Imager (EUI), onboard Solar Orbiter. Here we use EUI 174Å images, together with EUV images from SDO/AIA, and line-of-sight magnetograms from SDO/HMI to investigate the magnetic origin of 52 randomly selected campfires in the quiet solar corona. We find that (i) the campfires are rooted at the edges of photosphe… ▽ More Solar campfires are fine-scale heating events, recently observed by Extreme Ultraviolet Imager (EUI), onboard Solar Orbiter. Here we use EUI 174Å images, together with EUV images from SDO/AIA, and line-of-sight magnetograms from SDO/HMI to investigate the magnetic origin of 52 randomly selected campfires in the quiet solar corona. We find that (i) the campfires are rooted at the edges of photospheric magnetic network lanes; (ii) most of the campfires reside above the neutral line between majority-polarity magnetic flux patch and a merging minority-polarity flux patch, with a flux cancelation rate of $\sim$10$^{18}$Mx hr$^{-1}$; (iii) some of the campfires occur repeatedly from the same neutral line; (iv) in the large majority of instances, campfires are preceded by a cool-plasma structure, analogous to minifilaments in coronal jets; and (v) although many campfires have `complex' structure, most campfires resemble small-scale jets, dots, or loops. Thus, `campfire' is a general term that includes different types of small-scale solar dynamic features. They contain sufficient magnetic energy ($\sim$10$^{26}$-10$^{27}$ erg) to heat the solar atmosphere locally to 0.5--2.5MK. Their lifetimes range from about a minute to over an hour, with most of the campfires having a lifetime of $<$10 minutes. The average lengths and widths of the campfires are 5400$\pm$2500km and 1600$\pm$640km, respectively. Our observations suggest that (a) the presence of magnetic flux ropes may be ubiquitous in the solar atmosphere and not limited to coronal jets and larger-scale eruptions that make CMEs, and (b) magnetic flux cancelation is the fundamental process for the formation and triggering of most campfires. △ Less

Submitted 13 October, 2021; originally announced October 2021.

Comments: Accepted for publication in ApJ Letters, 20 Pages, 1 Table, 12 Figures

Spatial and temporal dynamics of RhoA activities of single breast tumor cells in a 3D environment revealed by a machine learning-assisted FRET technique

Authors: Brian CH Cheung, Louis Hodgson, Jeffrey E Segall, Mingming Wu

Abstract: One of the hallmarks of cancer cells is their exceptional ability to migrate within the extracellular matrix (ECM) for gaining access to the circulatory system, a critical step of cancer metastasis. RhoA, a small GTPase, is known to be a key molecular switch that toggles between actomyosin contractility and lamellipodial protrusion during cell migration. Current understanding of RhoA activity in c… ▽ More One of the hallmarks of cancer cells is their exceptional ability to migrate within the extracellular matrix (ECM) for gaining access to the circulatory system, a critical step of cancer metastasis. RhoA, a small GTPase, is known to be a key molecular switch that toggles between actomyosin contractility and lamellipodial protrusion during cell migration. Current understanding of RhoA activity in cell migration has been largely derived from studies of cells plated on a two-dimensional (2D) substrate using a FRET biosensor. There has been increasing evidence that cells behave differently in a more physiologically relevant three-dimensional (3D) environment, however, studies of RhoA activities in 3D have been hindered by low signal-to-noise ratio in fluorescence imaging. In this paper, we present a machine learning-assisted FRET technique to follow the spatiotemporal dynamics of RhoA activities of single breast tumor cells (MDA-MB-231) migrating in a 3D as well as a 2D environment using a RhoA biosensor. We found that RhoA activity is more polarized along the long axis of the cell for single cells migrating on 2D fibronectin-coated glass versus those embedded in 3D collagen matrices. In particular, RhoA activities of cells in 2D exhibit a distinct front-to-back and back-to-front movement during migration in contrast to those in 3D. Finally, regardless of dimensionality, RhoA polarization is found to be correlated with cell shape. △ Less

Submitted 29 September, 2021; originally announced September 2021.

Comments: 10 pages, 5 figures, 7 supplementary figures

Hallucinated but Factual! Inspecting the Factuality of Hallucinations in Abstractive Summarization

Authors: Meng Cao, Yue Dong, Jackie Chi Kit Cheung

Abstract: State-of-the-art abstractive summarization systems often generate \emph{hallucinations}; i.e., content that is not directly inferable from the source text. Despite being assumed incorrect, we find that much hallucinated content is factual, namely consistent with world knowledge. These factual hallucinations can be beneficial in a summary by providing useful background information. In this work, we… ▽ More State-of-the-art abstractive summarization systems often generate \emph{hallucinations}; i.e., content that is not directly inferable from the source text. Despite being assumed incorrect, we find that much hallucinated content is factual, namely consistent with world knowledge. These factual hallucinations can be beneficial in a summary by providing useful background information. In this work, we propose a novel detection approach that separates factual from non-factual hallucinations of entities. Our method utilizes an entity's prior and posterior probabilities according to pre-trained and finetuned masked language models, respectively. Empirical results suggest that our approach vastly outperforms two baselines %in both accuracy and F1 scores and strongly correlates with human judgments. % on factuality classification tasks. Furthermore, we show that our detector, when used as a reward signal in an off-line reinforcement learning (RL) algorithm, significantly improves the factuality of summaries while maintaining the level of abstractiveness. △ Less

Submitted 6 December, 2021; v1 submitted 30 August, 2021; originally announced September 2021.

A kernel-based least-squares collocation method for surface diffusion

Authors: Meng Chen, Ka Chun Cheung, Leevan Ling

Abstract: There are plenty of applications and analysis for time-independent elliptic partial differential equations in the literature hinting at the benefits of overtesting by using more collocation conditions than the number of basis functions. Overtesting not only reduces the problem size, but is also known to be necessary for stability and convergence of widely used unsymmetric Kansa-type strong-form co… ▽ More There are plenty of applications and analysis for time-independent elliptic partial differential equations in the literature hinting at the benefits of overtesting by using more collocation conditions than the number of basis functions. Overtesting not only reduces the problem size, but is also known to be necessary for stability and convergence of widely used unsymmetric Kansa-type strong-form collocation methods. We consider kernel-based meshfree methods, which is a method of lines with collocation and overtesting spatially, for solving parabolic partial differential equations on surfaces without parametrization. In this paper, we extend the time-independent convergence theories for overtesting techniques to the parabolic equations on smooth and closed surfaces. △ Less

Submitted 4 May, 2023; v1 submitted 7 September, 2021; originally announced September 2021.

Comments: 4 figures, 21 pages

MSC Class: 65D15; 65N35; 65N40; 41A63

Journal ref: SIAM Journal on Numerical Analysis,2023,Vol.61(3): 1386-1404

SynthIA: A Synthetic Inversion Approximation for the Stokes Vector Fusing SDO and Hinode into a Virtual Observatory

Authors: Richard E. L. Higgins, David F. Fouhey, Spiro K. Antiochos, Graham Barnes, Mark C. M. Cheung, J. Todd Hoeksema, KD Leka, Yang Liu, Peter W. Schuck, Tamas I. Gombosi

Abstract: Both NASA's Solar Dynamics Observatory (SDO) and the JAXA/NASA Hinode mission include spectropolarimetric instruments designed to measure the photospheric magnetic field. SDO's Helioseismic and Magnetic Imager (HMI) emphasizes full-disk high-cadence and good spatial resolution data acquisition while Hinode's Solar Optical Telescope Spectro-Polarimeter (SOT-SP) focuses on high spatial resolution an… ▽ More Both NASA's Solar Dynamics Observatory (SDO) and the JAXA/NASA Hinode mission include spectropolarimetric instruments designed to measure the photospheric magnetic field. SDO's Helioseismic and Magnetic Imager (HMI) emphasizes full-disk high-cadence and good spatial resolution data acquisition while Hinode's Solar Optical Telescope Spectro-Polarimeter (SOT-SP) focuses on high spatial resolution and spectral sampling at the cost of a limited field of view and slower temporal cadence. This work introduces a deep-learning system named SynthIA (Synthetic Inversion Approximation), that can enhance both missions by capturing the best of each instrument's characteristics. We use SynthIA to produce a new magnetogram data product, SynodeP (Synthetic Hinode Pipeline), that mimics magnetograms from the higher spectral resolution Hinode/SOT-SP pipeline, but is derived from full-disk, high-cadence, and lower spectral-resolution SDO/HMI Stokes observations. Results on held-out data show that SynodeP has good agreement with the Hinode/SOT-SP pipeline inversions, including magnetic fill fraction, which is not provided by the current SDO/HMI pipeline. SynodeP further shows a reduction in the magnitude of the 24-hour oscillations present in the SDO/HMI data. To demonstrate SynthIA's generality, we show the use of SDO/AIA data and subsets of the HMI data as inputs, which enables trade-offs between fidelity to the Hinode/SOT-SP inversions, number of observations used, and temporal artifacts. We discuss possible generalizations of SynthIA and its implications for space weather modeling. This work is part of the NASA Heliophysics DRIVE Science Center (SOLSTICE) at the University of Michigan under grant NASA 80NSSC20K0600E, and will be open-sourced. △ Less

Submitted 27 August, 2021; originally announced August 2021.