Search

An important task in the study of fast radio bursts (FRBs) remains the automatic classification of repeating and non-repeating sources based on their morphological properties. We propose a statistical model that considers a modified logistic regression to classify FRB sources. The classical logistic regression model is modified to accommodate the small proportion of repeaters in the data, a feature that is likely due to the sampling procedure and duration and is not a characteristic of the population of FRB sources. The weighted logistic regression hinges on the choice of a tuning parameter that represents the true proportion $\tau$ of repeating FRB sources in the entire population. The proposed method has a sound statistical foundation, direct interpretability, and operates with only 5 parameters, enabling quicker retraining with added data. Using the CHIME/FRB Collaboration sample of repeating and non-repeating FRBs and numerical experiments, we achieve a classification accuracy for repeaters of nearly 75\% or higher when $\tau$ is set in the range of $50$ to $60$\%. This implies a tentative high proportion of repeaters, which is surprising, but is also in agreement with recent estimates of $\tau$ that are obtained using other methods., Comment: 16 pages, 7 figures. Submitted to ApJ

The identification of persistent radio sources (PRSs) coincident with two repeating fast radio bursts (FRBs) supports FRB theories requiring a compact central engine. However, deep non-detections in other cases highlight the diversity of repeating FRBs and their local environments. Here, we perform a systematic search for radio sources towards 37 CHIME/FRB repeaters using their arcminute localizations and a combination of archival surveys and targeted observations. Through multi-wavelength analysis of individual radio sources, we identify two (20181030A-S1 and 20190417A-S1) for which we disfavor an origin of either star formation or an active galactic nucleus in their host galaxies and thus consider them candidate PRSs. We do not find any associated PRSs for the majority of the repeating FRBs in our sample. For 8 FRB fields with Very Large Array imaging, we provide deep limits on the presence of PRSs that are 2--4 orders of magnitude fainter than the PRS associated with FRB\,20121102A. Using Very Large Array Sky Survey imaging of all 37 fields, we constrain the rate of luminous ($\gtrsim$10$^{40}$ erg s$^{-1}$) PRSs associated with repeating FRBs to be low. Within the context of FRB-PRS models, we find that 20181030A-S1 and 20190417A-S1 can be reasonably explained within the context of magnetar, hypernebulae, gamma-ray burst afterglow, or supernova ejecta models -- although we note that both sources follow the radio luminosity versus rotation measure relationship predicted in the nebula model framework. Future observations will be required to both further characterize and confirm the association of these PRS candidates with the FRBs., Comment: 37 pages, 12 figures

K-means is a widely used algorithm in clustering, however, its efficiency is primarily constrained by the computational cost of distance computing. Existing implementations suffer from suboptimal utilization of computational units and lack resilience against soft errors. To address these challenges, we introduce FT K-means, a high-performance GPU-accelerated implementation of K-means with online fault tolerance. We first present a stepwise optimization strategy that achieves competitive performance compared to NVIDIA's cuML library. We further improve FT K-means with a template-based code generation framework that supports different data types and adapts to different input shapes. A novel warp-level tensor-core error correction scheme is proposed to address the failure of existing fault tolerance methods due to memory asynchronization during copy operations. Our experimental evaluations on NVIDIA T4 GPU and A100 GPU demonstrate that FT K-means without fault tolerance outperforms cuML's K-means implementation, showing a performance increase of 10\%-300\% in scenarios involving irregular data shapes. Moreover, the fault tolerance feature of FT K-means introduces only an overhead of 11\%, maintaining robust performance even with tens of errors injected per second.

Late-time ($\sim$ year) radio follow-up of optically-discovered tidal disruption events (TDEs) is increasingly resulting in detections at radio wavelengths, and there is growing evidence for this late-time radio activity to be common to the broad class of sub-relativistic TDEs. Detailed studies of some of these TDEs at radio wavelengths are also challenging the existing models for radio emission. Using all-sky multi-epoch data from the Australian Square Kilometre Array Pathfinder (ASKAP), taken as a part of the Rapid ASKAP Continuum Survey (RACS), we searched for radio counterparts to a sample of optically-discovered TDEs. We detected late-time emission at RACS frequencies (742-1032\,MHz) in five TDEs, reporting the independent discovery of radio emission from TDE AT2019ahk and extending the time baseline out to almost 3000\,days for some events. Overall, we find that at least $22^{+15}_{-11}$\% of the population of optically-discovered TDEs has detectable radio emission in the RACS survey, while also noting that the true fraction can be higher given the limited cadence (2 epochs separated by $\sim 3\,$ years) of the survey. Finally, we project that the ongoing higher-cadence ($\sim 2$\,months) ASKAP Variable and Slow Transients (VAST) survey can detect $\sim 20$ TDEs in its operational span (4\,yrs), given the current rate from optical surveys., Comment: Accepted for publication in ApJ, comments welcome

Understanding how biological visual systems process information is challenging because of the nonlinear relationship between visual input and neuronal responses. Artificial neural networks allow computational neuroscientists to create predictive models that connect biological and machine vision. Machine learning has benefited tremendously from benchmarks that compare different model on the same task under standardized conditions. However, there was no standardized benchmark to identify state-of-the-art dynamic models of the mouse visual system. To address this gap, we established the Sensorium 2023 Benchmark Competition with dynamic input, featuring a new large-scale dataset from the primary visual cortex of ten mice. This dataset includes responses from 78,853 neurons to 2 hours of dynamic stimuli per neuron, together with the behavioral measurements such as running speed, pupil dilation, and eye movements. The competition ranked models in two tracks based on predictive performance for neuronal responses on a held-out test set: one focusing on predicting in-domain natural stimuli and another on out-of-distribution (OOD) stimuli to assess model generalization. As part of the NeurIPS 2023 competition track, we received more than 160 model submissions from 22 teams. Several new architectures for predictive models were proposed, and the winning teams improved the previous state-of-the-art model by 50%. Access to the dataset as well as the benchmarking infrastructure will remain online at www.sensorium-competition.net.

Recent works have constrained the binary fraction of evolved populations of massive stars in local galaxies such as red supergiants and Wolf-Rayet stars, but the binary fraction of yellow supergiants (YSGs) in the Hertzsprung Gap remains unconstrained. Binary evolution theory predicts that the Hertzsprung Gap is home to multiple populations of binary systems with varied evolutionary histories. In this paper, we develop a method to distinguish single YSGs from YSG plus O- or B-type main sequence binaries using optical and ultraviolet photometry, and then apply this method to identify candidate YSG binaries in the Magellanic Clouds. After constructing a set of combined stellar atmosphere models, we find that optical photometry is, given typical measurement and reddening uncertainties, sufficient to discern single YSGs from YSG+OB binaries if the OB-star is at least $\sim5$M$_{\odot}$ for T$_{\mathrm{eff,YSG}}\sim$ 4000 K, but requires a $\sim$20M$_{\odot}$ OB star for YSGs up to T$_{\mathrm{eff,YSG}}\sim$ 9000 K. For these hotter YSG temperatures, ultraviolet photometry allows binaries with OB companions as small as $\sim$7M$_{\odot}$ to be identified. We use color-color spaces developed from these models to search for evidence of excess blue or ultraviolet light in a set of $\sim$1000 YSG candidates in the Magellanic Clouds. We identify hundreds of candidate YSG binary systems and report a preliminary fraction of YSGs that show a blue/UV color excess of 20-60\%. Spectroscopic follow-up is now required to confirm the true nature of this population., Comment: 21 pages, 10 figures, 3 tables. Submitted to The Astrophysical Journal on June 24th 2024, Accepted on September 3rd

Tuning searches are pivotal in High-Performance Computing (HPC), addressing complex optimization challenges in computational applications. The complexity arises not only from finely tuning parameters within routines but also potential interdependencies among them, rendering traditional optimization methods inefficient. Instead of scrutinizing interdependencies among parameters and routines, practitioners often face the dilemma of conducting independent tuning searches for each routine, thereby overlooking interdependence, or pursuing a more resource-intensive joint search for all routines. This decision is driven by the consideration that some interdependence analysis and high-dimensional decomposition techniques in literature may be prohibitively expensive in HPC tuning searches. Our methodology adapts and refines these methods to ensure computational feasibility while maximizing performance gains in real-world scenarios. Our methodology leverages a cost-effective interdependence analysis to decide whether to merge several tuning searches into a joint search or conduct orthogonal searches. Tested on synthetic functions with varying levels of parameter interdependence, our methodology efficiently explores the search space. In comparison to Bayesian-optimization-based full independent or fully joint searches, our methodology suggested an optimized breakdown of independent and merged searches that led to final configurations up to 8% more accurate, reducing the search time by up to 95%. When applied to GPU-offloaded Real-Time Time-Dependent Density Functional Theory (RT-TDDFT), an application in computational materials science that challenges modern HPC autotuners, our methodology achieved an effective tuning search. Its adaptability and efficiency extend beyond RT-TDDFT, making it valuable for related applications in HPC.

Class activation maps (CAMs) are commonly employed in weakly supervised semantic segmentation (WSSS) to produce pseudo-labels. Due to incomplete or excessive class activation, existing studies often resort to offline CAM refinement, introducing additional stages or proposing offline modules. This can cause optimization difficulties for single-stage methods and limit generalizability. In this study, we aim to reduce the observed CAM inconsistency and error to mitigate reliance on refinement processes. We propose an end-to-end WSSS model incorporating guided CAMs, wherein our segmentation model is trained while concurrently optimizing CAMs online. Our method, Co-training with Swapping Assignments (CoSA), leverages a dual-stream framework, where one sub-network learns from the swapped assignments generated by the other. We introduce three techniques: i) soft perplexity-based regularization to penalize uncertain regions; ii) a threshold-searching approach to dynamically revise the confidence threshold; and iii) contrastive separation to address the coexistence problem. CoSA demonstrates exceptional performance, achieving mIoU of 76.2\% and 51.0\% on VOC and COCO validation datasets, respectively, surpassing existing baselines by a substantial margin. Notably, CoSA is the first single-stage approach to outperform all existing multi-stage methods including those with additional supervision. Code is avilable at \url{https://github.com/youshyee/CoSA}., Comment: Accepted at ECCV24

We present 3D Points Splatting Hand Reconstruction (3D-PSHR), a real-time and photo-realistic hand reconstruction approach. We propose a self-adaptive canonical points upsampling strategy to achieve high-resolution hand geometry representation. This is followed by a self-adaptive deformation that deforms the hand from the canonical space to the target pose, adapting to the dynamic changing of canonical points which, in contrast to the common practice of subdividing the MANO model, offers greater flexibility and results in improved geometry fitting. To model texture, we disentangle the appearance color into the intrinsic albedo and pose-aware shading, which are learned through a Context-Attention module. Moreover, our approach allows the geometric and the appearance models to be trained simultaneously in an end-to-end manner. We demonstrate that our method is capable of producing animatable, photorealistic and relightable hand reconstructions using multiple datasets, including monocular videos captured with handheld smartphones and large-scale multi-view videos featuring various hand poses. We also demonstrate that our approach achieves real-time rendering speeds while simultaneously maintaining superior performance compared to existing state-of-the-art methods.

Personal sensing, leveraging data passively and near-continuously collected with wearables from patients in their ecological environment, is a promising paradigm to monitor mood disorders (MDs), a major determinant of worldwide disease burden. However, collecting and annotating wearable data is very resource-intensive. Studies of this kind can thus typically afford to recruit only a couple dozens of patients. This constitutes one of the major obstacles to applying modern supervised machine learning techniques to MDs detection. In this paper, we overcome this data bottleneck and advance the detection of MDs acute episode vs stable state from wearables data on the back of recent advances in self-supervised learning (SSL). This leverages unlabelled data to learn representations during pre-training, subsequently exploited for a supervised task. First, we collected open-access datasets recording with an Empatica E4 spanning different, unrelated to MD monitoring, personal sensing tasks -- from emotion recognition in Super Mario players to stress detection in undergraduates -- and devised a pre-processing pipeline performing on-/off-body detection, sleep-wake detection, segmentation, and (optionally) feature extraction. With 161 E4-recorded subjects, we introduce E4SelfLearning, the largest to date open access collection, and its pre-processing pipeline. Second, we show that SSL confidently outperforms fully-supervised pipelines using either our novel E4-tailored Transformer architecture (E4mer) or classical baseline XGBoost: 81.23% against 75.35% (E4mer) and 72.02% (XGBoost) correctly classified recording segments from 64 (half acute, half stable) patients. Lastly, we illustrate that SSL performance is strongly associated with the specific surrogate task employed for pre-training as well as with unlabelled data availability.

We present a novel, computationally efficient approach to accelerate quantum optimal control calculations of large multi-qubit systems used in a variety of quantum computing applications. By leveraging the intrinsic symmetry of finite groups, the Hilbert space can be decomposed and the Hamiltonians block-diagonalized to enable extremely fast quantum optimal control calculations. Our approach reduces the Hamiltonian size of an $n$-qubit system from 2^n by 2^n to O(n by n) or O((2^n / n) by (2^n / n)) under Sn or Dn symmetry, respectively. Most importantly, this approach reduces the computational runtime of qubit optimal control calculations by orders of magnitude while maintaining the same accuracy as the conventional method. As prospective applications, we show that (1) symmetry-protected subspaces can be potential platforms for quantum error suppression and simulation of other quantum Hamiltonians, and (2) Lie-Trotter-Suzuki decomposition approaches can generalize our method to a general variety of multi-qubit systems., Comment: 15 pages, 5 figures. This article may be downloaded for personal use only. Any other use requires prior permission of the author and AIP Publishing. This article appeared in AVS Quantum Science and may be found at https://doi.org/10.1116/5.0162455. Find the PDF of the Supplementary Material in the source files

We present a new velocity-gauge real-time, time-dependent density functional tight-binding (VG-rtTDDFTB) implementation in the open-source DFTB+ software package (https://dftbplus.org) for probing electronic excitations in large, condensed matter systems. Our VG-rtTDDFTB approach enables real-time electron dynamics simulations of large, periodic, condensed matter systems containing thousands of atoms with a favorable computational scaling as a function of system size. We provide computational details and benchmark calculations to demonstrate its accuracy and computational parallelizability on a variety of large material systems. As a representative example, we calculate laser-induced electron dynamics in a 512-atom amorphous silicon supercell to highlight the large periodic systems that can be examined with our implementation. Taken together, our VG-rtTDDFTB approach enables new electron dynamics simulations of complex systems that require large periodic supercells, such as crystal defects, complex surfaces, nanowires, and amorphous materials., Comment: 23 pages, 6 figures

We report ten fast radio bursts (FRBs) detected in the far side-lobe region (i.e., $\geq 5^\circ$ off-meridian) of the Canadian Hydrogen Intensity Mapping Experiment (CHIME) from 2018 August 28 to 2021 August 31. We localize the bursts by fitting their spectra with a model of the CHIME/FRB synthesized beam response. We find that the far side-lobe events have on average ~500 times greater fluxes than events detected in CHIME's main lobe. We show that the side-lobe sample is therefore statistically ~20 times closer than the main-lobe sample. We find promising host galaxy candidates (P$_{\rm cc}$ < 1%) for two of the FRBs, 20190112B and 20210310B, at distances of 38 and 16 Mpc, respectively. CHIME/FRB did not observe repetition of similar brightness from the uniform sample of 10 side-lobe FRBs in a total exposure time of 35580 hours. Under the assumption of Poisson-distributed bursts, we infer that the mean repetition interval above the detection threshold of the far side-lobe events is longer than 11880 hours, which is at least 2380 times larger than the interval from known CHIME/FRB detected repeating sources, with some caveats, notably that very narrow-band events could have been missed. Our results from these far side-lobe events suggest one of two scenarios: either (1) all FRBs repeat and the repetition intervals span a wide range, with high-rate repeaters being a rare subpopulation, or (2) non-repeating FRBs are a distinct population different from known repeaters., Comment: 27 pages, 20 figures. This version is the result of the merger of arxiv:2307.05262 and the previous version of this paper. Accepted for publication in ApJ

We study the 10 fast radio bursts (FRBs) detected in the far side-lobe region of the CHIME telescope from 2018 August 28 to 2021 August 31. We find that the far side-lobe events have on average $\sim$500 times greater fluxes than events detected in CHIME's main lobe. We show that the side-lobe sample is therefore statistically $\sim$20 times closer than the main-lobe sample. The median dispersion measure (DM) excess, after removing the Galactic disk component using the NE2001 for the free electron density distribution of the Milky Way, of the 10 far side-lobe and 471 non-repeating main-lobe FRBs in the first CHIME/FRB catalog is 183.0 and 433.9 pc\;cm$^{-3}$, respectively. By comparing the DM excesses of the two populations under reasonable assumptions, we statistically constrain that the local degenerate contributions (from the Milky Way halo and the host galaxy) and the intergalactic contribution to the excess DM of the 471 non-repeating main-lobe FRBs for the NE2001 model are 131.2$-$158.3 and 302.7$-$275.6 pc cm$^{-3}$, respectively, which corresponds to a median redshift for the main-lobe FRB sample of $\sim$0.3. These constraints are useful for population studies of FRBs, and in particular for constraining the location of the missing baryons., Comment: This paper has been merged with arxiv:2307.05261. Refer to that article for the latest version

General Matrix Multiplication (GEMM) is a crucial algorithm for various applications such as machine learning and scientific computing, and an efficient GEMM implementation is essential for the performance of these systems. While researchers often strive for faster performance by using large compute platforms, the increased scale of these systems can raise concerns about hardware and software reliability. In this paper, we present a design for a high-performance GEMM with algorithm-based fault tolerance for use on GPUs. We describe fault-tolerant designs for GEMM at the thread, warp, and threadblock levels, and also provide a baseline GEMM implementation that is competitive with or faster than the state-of-the-art, proprietary cuBLAS GEMM. We present a kernel fusion strategy to overlap and mitigate the memory latency due to fault tolerance with the original GEMM computation. To support a wide range of input matrix shapes and reduce development costs, we present a template-based approach for automatic code generation for both fault-tolerant and non-fault-tolerant GEMM implementations. We evaluate our work on NVIDIA Tesla T4 and A100 server GPUs. Experimental results demonstrate that our baseline GEMM presents comparable or superior performance compared to the closed-source cuBLAS. The fault-tolerant GEMM incurs only a minimal overhead (8.89\% on average) compared to cuBLAS even with hundreds of errors injected per minute. For irregularly shaped inputs, the code generator-generated kernels show remarkable speedups of $160\% \sim 183.5\%$ and $148.55\% \sim 165.12\%$ for fault-tolerant and non-fault-tolerant GEMMs, outperforming cuBLAS by up to $41.40\%$., Comment: 11 pages, 2023 International Conference on Supercomputing

Recently, deep learning based approaches have shown promising results in 3D hand reconstruction from a single RGB image. These approaches can be roughly divided into model-based approaches, which are heavily dependent on the model's parameter space, and model-free approaches, which require large numbers of 3D ground truths to reduce depth ambiguity and struggle in weakly-supervised scenarios. To overcome these issues, we propose a novel probabilistic model to achieve the robustness of model-based approaches and reduced dependence on the model's parameter space of model-free approaches. The proposed probabilistic model incorporates a model-based network as a prior-net to estimate the prior probability distribution of joints and vertices. An Attention-based Mesh Vertices Uncertainty Regression (AMVUR) model is proposed to capture dependencies among vertices and the correlation between joints and mesh vertices to improve their feature representation. We further propose a learning based occlusion-aware Hand Texture Regression model to achieve high-fidelity texture reconstruction. We demonstrate the flexibility of the proposed probabilistic model to be trained in both supervised and weakly-supervised scenarios. The experimental results demonstrate our probabilistic model's state-of-the-art accuracy in 3D hand and texture reconstruction from a single image in both training schemes, including in the presence of severe occlusions.

Accurate predictive models of the visual cortex neural response to natural visual stimuli remain a challenge in computational neuroscience. In this work, we introduce V1T, a novel Vision Transformer based architecture that learns a shared visual and behavioral representation across animals. We evaluate our model on two large datasets recorded from mouse primary visual cortex and outperform previous convolution-based models by more than 12.7% in prediction performance. Moreover, we show that the self-attention weights learned by the Transformer correlate with the population receptive fields. Our model thus sets a new benchmark for neural response prediction and can be used jointly with behavioral and neural recordings to reveal meaningful characteristic features of the visual cortex., Comment: updated references and added link to code repository; add analysis on generalization and visualize aRFs; updated with TMLR publication