Searcharxiv⌕ Search

arXiv subjects

Hao Ding

Publications and source records attributed to Hao Ding.

At least 91 records · Page 5Linked to original sources
arXiv2023

Enhancing the use of Galactic neutron stars as physical laboratories with precise astrometry

The existence of neutron stars was not confirmed until the discovery of pulsars at radio wavelengths in late 1960s. Since then, these highly compact and magnetized objects have been observed across the electromagnetic spectrum, and widely studied. However, lots of the studies related to neutron stars require precise determination of their distances and proper motions. This thesis focuses on high-precision astrometry of neutron stars using the data from the Very Long Baseline Array (VLBA) and the Gaia space telescope operating, respectively, at radio and optical frequencies. The neutron stars studied in the thesis include the extremely magnetized magnetars, the fast-spinning millisecond pulsars, the gravitational-wave-emitting double neutron stars and neutron star X-ray binaries. As a major accomplishment, this thesis presents the novel analysis and the results of the MSPSRpi project -- the largest astrometric survey of millisecond pulsars, then point out the abundant implications of the astrometric results. Additionally, the release of the astrometric results is bound to facilitate the detection of an ultra-low-frequency gravitational-wave background. Methodologically, this thesis applied advanced VLBI techniques to pulsar astrometry using the original data reduction pipeline psrvlbireduce, which leads to the first two significant magnetar parallaxes, and paves the way for studying magnetar formation channels with their velocity distribution. The astrometry Bayesian inference package sterne, developed during the PhD program, serves as a versatile and powerful tool for the inference of astrometric parameters.

astro-ph.HE↗
arXiv2022

Few-shot Medical Image Segmentation with Cycle-resemblance Attention

Recently, due to the increasing requirements of medical imaging applications and the professional requirements of annotating medical images, few-shot learning has gained increasing attention in the medical image semantic segmentation field. To perform segmentation with limited number of labeled medical images, most existing studies use Proto-typical Networks (PN) and have obtained compelling success. However, these approaches overlook the query image features extracted from the proposed representation network, failing to preserving the spatial connection between query and support images. In this paper, we propose a novel self-supervised few-shot medical image segmentation network and introduce a novel Cycle-Resemblance Attention (CRA) module to fully leverage the pixel-wise relation between query and support medical images. Notably, we first line up multiple attention blocks to refine more abundant relation information. Then, we present CRAPNet by integrating the CRA module with a classic prototype network, where pixel-wise relations between query and support features are well recaptured for segmentation. Extensive experiments on two different medical image datasets, e.g., abdomen MRI and abdomen CT, demonstrate the superiority of our model over existing state-of-the-art methods.

cs.CV↗
arXiv2022

Rethinking Causality-driven Robot Tool Segmentation with Temporal Constraints

Purpose: Vision-based robot tool segmentation plays a fundamental role in surgical robots and downstream tasks. CaRTS, based on a complementary causal model, has shown promising performance in unseen counterfactual surgical environments in the presence of smoke, blood, etc. However, CaRTS requires over 30 iterations of optimization to converge for a single image due to limited observability. Method: To address the above limitations, we take temporal relation into consideration and propose a temporal causal model for robot tool segmentation on video sequences. We design an architecture named Temporally Constrained CaRTS (TC-CaRTS). TC-CaRTS has three novel modules to complement CaRTS - temporal optimization pipeline, kinematics correction network, and spatial-temporal regularization. Results: Experiment results show that TC-CaRTS requires much fewer iterations to achieve the same or better performance as CaRTS. TC- CaRTS also has the same or better performance in different domains compared to CaRTS. All three modules are proven to be effective. Conclusion: We propose TC-CaRTS, which takes advantage of temporal constraints as additional observability. We show that TC-CaRTS outperforms prior work in the robot tool segmentation task with improved convergence speed on test datasets from different domains.

cs.RO↗
arXiv2022

CaRTS: Causality-driven Robot Tool Segmentation from Vision and Kinematics Data

Vision-based segmentation of the robotic tool during robot-assisted surgery enables downstream applications, such as augmented reality feedback, while allowing for inaccuracies in robot kinematics. With the introduction of deep learning, many methods were presented to solve instrument segmentation directly and solely from images. While these approaches made remarkable progress on benchmark datasets, fundamental challenges pertaining to their robustness remain. We present CaRTS, a causality-driven robot tool segmentation algorithm, that is designed based on a complementary causal model of the robot tool segmentation task. Rather than directly inferring segmentation masks from observed images, CaRTS iteratively aligns tool models with image observations by updating the initially incorrect robot kinematic parameters through forward kinematics and differentiable rendering to optimize image feature similarity end-to-end. We benchmark CaRTS with competing techniques on both synthetic as well as real data from the dVRK, generated in precisely controlled scenarios to allow for counterfactual synthesis. On training-domain test data, CaRTS achieves a Dice score of 93.4 that is preserved well (Dice score of 91.8) when tested on counterfactually altered test data, exhibiting low brightness, smoke, blood, and altered background patterns. This compares favorably to Dice scores of 95.0 and 86.7, respectively, of the SOTA image-based method. Future work will involve accelerating CaRTS to achieve video framerate and estimating the impact occlusion has in practice. Despite these limitations, our results are promising: In addition to achieving high segmentation accuracy, CaRTS provides estimates of the true robot kinematics, which may benefit applications such as force estimation. Code is available at: https://github.com/hding2455/CaRTS

cs.RO↗
arXiv2022

Hankel Spectrum Analysis: A novel signal decomposition method and its geophysical applications

To analyze non-stationary harmonic signals typically contained in geophysical observables is a quest that has seen continual advances in numerical techniques over the decades. In this paper, based on transient z-pole estimation (in Hankel matrices), a novel state-space analysis referred to as Hankel Spectral Analysis (HSA), was developed. Depended on the Hankel total least square (HTLS), the HSA incorporates truncated singular value decomposition (TSVD) and its shift-invariant property in robustly decomposing the closely-spaced sinusoids. Resorted to a sliding window processing, HSA can be used to analyze non-stationary sequential structures, in the support of consecutive quaternary parameters {Ai, αi, fi, θi}. Based on a series of experiments with special features commonly in real measurements, the availabilities of HSA in complex harmonic constituents (e.g., the time-variant amplitude/frequency, mutation, the episodic recording signals) with low Signal-to-Noise Ratio are confirmed. In real applications, we use HSA to analyze both global geophysical observables, including polar motion (PM) and earth's dynamic oblateness (ΔJ2), and some new findings are obtained. In the PM series since the 1900s, a total of triple jumps from Chandler wobble (CW) are firstly confirmed; and all of them are synchronized by the sharp decrease of Chandler intensity and period. In the ΔJ2 series, two decadal signals (18.6 yr, 10.5 yr) are identified to be associated with the tide effect, and solar activity; and its interannual-to-decadal oscillations contribute to multiple global gravity anomalies. These findings implied the great potential of the HSA in searching hitherto signals of geophysical observations.

physics.geo-ph↗
arXiv2022

Context Uncertainty in Contextual Bandits with Applications to Recommender Systems

Recurrent neural networks have proven effective in modeling sequential user feedbacks for recommender systems. However, they usually focus solely on item relevance and fail to effectively explore diverse items for users, therefore harming the system performance in the long run. To address this problem, we propose a new type of recurrent neural networks, dubbed recurrent exploration networks (REN), to jointly perform representation learning and effective exploration in the latent space. REN tries to balance relevance and exploration while taking into account the uncertainty in the representations. Our theoretical analysis shows that REN can preserve the rate-optimal sublinear regret even when there exists uncertainty in the learned representations. Our empirical study demonstrates that REN can achieve satisfactory long-term rewards on both synthetic and real-world recommendation datasets, outperforming state-of-the-art models.

cs.LG↗
arXiv2022

Probing magnetar formation channels with high-precision astrometry: The progress of VLBA astrometry of the fastest-spinning magnetar Swift J1818.0-1607

Boasting supreme magnetic strengths, magnetars are among the prime candidates to generate fast radio bursts. Several theories have been proposed for the formation mechanism of magnetars, but have not yet been fully tested. As different magnetar formation theories expect distinct magnetar space velocity distributions, high-precision astrometry of Galactic magnetars can serve as a probe for the formation theories. In addition, magnetar astrometry can refine the understanding of the distribution of Galactic magnetars. This distribution can be compared against fast radio bursts (FRBs) localized in spiral galaxies, in order to test the link between FRBs and magnetars. Swift J1818.0-1607 is the hitherto fastest-spinning magnetar and the fifth discovered radio magnetar. In an ongoing astrometric campaign, we have observed Swift J1818.0-1607 for one year using the Very Long Baseline Array, and have determined a precise proper motion as well as a tentative parallax for the magnetar.

astro-ph.HE↗
arXiv2021

The orbital-decay test of general relativity to the 2% level with 6-year VLBA astrometry of the double neutron star PSR J1537+1155

PSR J1537+1155, also known as PSR B1534+12, is the second discovered double neutron star (DNS) binary. More than 20 years of timing observations of PSR J1537+1155 have offered some of the most precise tests of general relativity (GR) in the strong-field regime. As one of these tests, the gravitational-wave emission predicted by GR has been probed with the significant orbital decay ($\dot{P}_\mathrm{b}$) of PSR J1537+1155. However, compared to most GR tests provided with the post-Keplerian parameters, the orbital-decay test was lagging behind in terms of both precision and consistency with GR, limited by the uncertain distance of PSR J1537+1155. With an astrometric campaign spanning 6 years using the Very Long Baseline Array, we measured an annual geometric parallax of $1.063\pm0.075$ mas for PSR J1537+1155, corresponding to a distance of $0.94^{+0.07}_{-0.06}$ kpc. This is the most tightly-constrained model-independent distance achieved for a DNS to date. After obtaining $\dot{P}_\mathrm{b}^\mathrm{Gal}$ (i.e., the orbital decay caused by Galactic gravitational potential) with a combination of 4 Galactic mass distribution models, we updated the ratio of the observed intrinsic orbital decay to the GR prediction to $0.977\pm0.020$, three times more precise than the previous orbital-decay test ($0.91\pm0.06$) made with PSR J1537+1155.

astro-ph.HE↗
arXiv2021

Zero-Shot Recommender Systems

Performance of recommender systems (RS) relies heavily on the amount of training data available. This poses a chicken-and-egg problem for early-stage products, whose amount of data, in turn, relies on the performance of their RS. On the other hand, zero-shot learning promises some degree of generalization from an old dataset to an entirely new dataset. In this paper, we explore the possibility of zero-shot learning in RS. We develop an algorithm, dubbed ZEro-Shot Recommenders (ZESRec), that is trained on an old dataset and generalize to a new one where there are neither overlapping users nor overlapping items, a setting that contrasts typical cross-domain RS that has either overlapping users or items. Different from categorical item indices, i.e., item ID, in previous methods, ZESRec uses items' natural-language descriptions (or description embeddings) as their continuous indices, and therefore naturally generalize to any unseen items. In terms of users, ZESRec builds upon recent advances on sequential RS to represent users using their interactions with items, thereby generalizing to unseen users as well. We study three pairs of real-world RS datasets and demonstrate that ZESRec can successfully enable recommendations in such a zero-shot setting, opening up new opportunities for resolving the chicken-and-egg problem for data-scarce startups or early-stage products.

cs.LG↗
arXiv2021

On the Sins of Image Synthesis Loss for Self-supervised Depth Estimation

Scene depth estimation from stereo and monocular imagery is critical for extracting 3D information for downstream tasks such as scene understanding. Recently, learning-based methods for depth estimation have received much attention due to their high performance and flexibility in hardware choice. However, collecting ground truth data for supervised training of these algorithms is costly or outright impossible. This circumstance suggests a need for alternative learning approaches that do not require corresponding depth measurements. Indeed, self-supervised learning of depth estimation provides an increasingly popular alternative. It is based on the idea that observed frames can be synthesized from neighboring frames if accurate depth of the scene is known - or in this case, estimated. We show empirically that - contrary to common belief - improvements in image synthesis do not necessitate improvement in depth estimation. Rather, optimizing for image synthesis can result in diverging performance with respect to the main prediction objective - depth. We attribute this diverging phenomenon to aleatoric uncertainties, which originate from data. Based on our experiments on four datasets (spanning street, indoor, and medical) and five architectures (monocular and stereo), we conclude that this diverging phenomenon is independent of the dataset domain and not mitigated by commonly used regularization techniques. To underscore the importance of this finding, we include a survey of methods which use image synthesis, totaling 127 papers over the last six years. This observed divergence has not been previously reported or studied in depth, suggesting room for future improvement of self-supervised approaches which might be impacted the finding.

cs.CV↗
arXiv2021

Influence Selection for Active Learning

The existing active learning methods select the samples by evaluating the sample's uncertainty or its effect on the diversity of labeled datasets based on different task-specific or model-specific criteria. In this paper, we propose the Influence Selection for Active Learning(ISAL) which selects the unlabeled samples that can provide the most positive Influence on model performance. To obtain the Influence of the unlabeled sample in the active learning scenario, we design the Untrained Unlabeled sample Influence Calculation(UUIC) to estimate the unlabeled sample's expected gradient with which we calculate its Influence. To prove the effectiveness of UUIC, we provide both theoretical and experimental analyses. Since the UUIC just depends on the model gradients, which can be obtained easily from any neural network, our active learning algorithm is task-agnostic and model-agnostic. ISAL achieves state-of-the-art performance in different active learning settings for different tasks with different datasets. Compared with previous methods, our method decreases the annotation cost at least by 12%, 13% and 16% on CIFAR10, VOC2012 and COCO, respectively.

cs.LG↗
arXiv2021

Gaia EDR3 parallaxes of type I X-ray bursters and their implications on the models of type I X-ray bursts: a generic approach to the Gaia parallax zero-point and its uncertainty

Light curves of photospheric radius expansion (PRE) bursts, a subset of type I X-ray bursts, have been used as standard candles to estimate the "nominal PRE distances" for 63% of PRE bursters (bursters), assuming PRE burst emission is spherically symmetric. Model-independent geometric parallaxes of bursters provide a valuable chance to test models of PRE bursts (PRE models), and can be provided in some cases by Gaia astrometry of the donor stars in bursters. We searched for counterparts to 115 known bursters in the Gaia Early Data Release 3, and confirmed 4 bursters with Gaia counterparts that have detected (>3sigma, prior to zero-point correction) parallaxes. We describe a generic approach to the Gaia parallax zero point as well as its uncertainty using an ensemble of Gaia quasars individually determined for each target. Assuming the spherically symmetric PRE model is correct, we refined the resultant nominal PRE distances of three bursters (i.e. Cen X-4, Cyg X-2 and 4U 0919-54), and put constraints on their compositions of the nuclear fuel powering the bursts. Finally, we describe a method for testing the correctness of the spherically symmetric PRE model using parallax measurements, and provide preliminary results.

astro-ph.HE↗
arXiv2021

Learning-based Efficient Graph Similarity Computation via Multi-Scale Convolutional Set Matching

Graph similarity computation is one of the core operations in many graph-based applications, such as graph similarity search, graph database analysis, graph clustering, etc. Since computing the exact distance/similarity between two graphs is typically NP-hard, a series of approximate methods have been proposed with a trade-off between accuracy and speed. Recently, several data-driven approaches based on neural networks have been proposed, most of which model the graph-graph similarity as the inner product of their graph-level representations, with different techniques proposed for generating one embedding per graph. However, using one fixed-dimensional embedding per graph may fail to fully capture graphs in varying sizes and link structures, a limitation that is especially problematic for the task of graph similarity computation, where the goal is to find the fine-grained difference between two graphs. In this paper, we address the problem of graph similarity computation from another perspective, by directly matching two sets of node embeddings without the need to use fixed-dimensional vectors to represent whole graphs for their similarity computation. The model, GraphSim, achieves the state-of-the-art performance on four real-world graph datasets under six out of eight settings (here we count a specific dataset and metric combination as one setting), compared to existing popular methods for approximate Graph Edit Distance (GED) and Maximum Common Subgraph (MCS) computation.

cs.LG↗
arXiv2021

Deeply Shape-guided Cascade for Instance Segmentation

The key to a successful cascade architecture for precise instance segmentation is to fully leverage the relationship between bounding box detection and mask segmentation across multiple stages. Although modern instance segmentation cascades achieve leading performance, they mainly make use of a unidirectional relationship, i.e., mask segmentation can benefit from iteratively refined bounding box detection. In this paper, we investigate an alternative direction, i.e., how to take the advantage of precise mask segmentation for bounding box detection in a cascade architecture. We propose a Deeply Shape-guided Cascade (DSC) for instance segmentation, which iteratively imposes the shape guidances extracted from mask prediction at the previous stage on bounding box detection at current stage. It forms a bi-directional relationship between the two tasks by introducing three key components: (1) Initial shape guidance: A mask-supervised Region Proposal Network (mPRN) with the ability to generate class-agnostic masks; (2) Explicit shape guidance: A mask-guided region-of-interest (RoI) feature extractor, which employs mask segmentation at previous stage to focus feature extraction at current stage within a region aligned well with the shape of the instance-of-interest rather than a rectangular RoI; (3) Implicit shape guidance: A feature fusion operation which feeds intermediate mask features at previous stage to the bounding box head at current stage. Experimental results show that DSC outperforms the state-of-the-art instance segmentation cascade, Hybrid Task Cascade (HTC), by a large margin and achieves 51.8 box AP and 45.5 mask AP on COCO test-dev. The code is released at: https://github.com/hding2455/DSC.

cs.CV↗
arXiv2021

Tuning interactions between spins in a superconductor

Novel many-body and topological electronic phases can be created in assemblies of interacting spins coupled to a superconductor, such as one-dimensional topological superconductors with Majorana zero modes (MZMs) at their ends. Understanding and controlling interactions between spins and the emergent band structure of the in-gap Yu-Shiba-Rusinov (YSR) states they induce in a superconductor are fundamental for engineering such phases. Here, by precisely positioning magnetic adatoms with a scanning tunneling microscope (STM), we demonstrate both the tunability of exchange interaction between spins and precise control of the hybridization of YSR states they induce on the surface of a bismuth (Bi) thin film that is made superconducting with the proximity effect. In this platform, depending on the separation of spins, the interplay between Ruderman-Kittel-Kasuya-Yosida (RKKY) interaction, spin-orbit coupling, and surface magnetic anisotropy stabilizes different types of spin alignments. Using high-resolution STM spectroscopy at millikelvin temperatures, we probe these spin alignments through monitoring the spin-induced YSR states and their energy splitting. Such measurements also reveal a quantum phase transition between the ground states with different electron number parity for a pair of spins in a superconductor tuned by their separation. Experiments on larger assemblies show that spin-spin interactions can be mediated in a superconductor over long distances. Our results show that controlling hybridization of the YSR states in this platform provides the possibility of engineering the band structure of such states for creating topological phases.

cond-mat.supr-con↗
arXiv2021

Gender Inequality in Research Productivity During the COVID-19 Pandemic

We study the disproportionate impact of the lockdown as a result of the COVID-19 outbreak on female and male academics' research productivity in social science. The lockdown has caused substantial disruptions to academic activities, requiring people to work from home. How this disruption affects productivity and the related gender equity is an important operations and societal question. We collect data from the largest open-access preprint repository for social science on 41,858 research preprints in 18 disciplines produced by 76,832 authors across 25 countries over a span of two years. We use a difference-in-differences approach leveraging the exogenous pandemic shock. Our results indicate that, in the 10 weeks after the lockdown in the United States, although the total research productivity increased by 35%, female academics' productivity dropped by 13.9% relative to that of male academics. We also show that several disciplines drive such gender inequality. Finally, we find that this intensified productivity gap is more pronounced for academics in top-ranked universities, and the effect exists in six other countries. Our work points out the fairness issue in productivity caused by the lockdown, a finding that universities will find helpful when evaluating faculty productivity. It also helps organizations realize the potential unintended consequences that can arise from telecommuting.

cs.DL↗
arXiv2020

New evidences for the fluctuation characteristic of intradecadal periodic signals in length-of-day variation

The intradecadal fluctuations in the length-of-day variation (dLOD) are considered likely to play an important role in core motions. Two intradecadal oscillations, with 5.9yr and 8.5yr periods (referred to as SYO and EYO, respectively), have been detected in previous studies. However, whether the SYO and the EYO have stable damping trends since 1962 and whether geomagnetic jerks are possible excitation sources for the SYO/EYO are still debated. In this study, based on different methods and dLOD records with different time span, we show robust evidences to prove that the SYO and the EYO have no stable damping trends since 1962, and we find that there is also a possible 7.6yr signal. To prove whether it is a periodic signal, we use the optimal sequence estimation method to stack 35 global geomagnetic records, the results also show an 7.6yr periodic signal which has an Y2,-2 spatial distribution, and it has a high degree of consistent synchronicity with the 7.6yr signal in dLOD. After confirming that the jerks have no special consistency with the peaks/valleys of the EYO/SYO, we confirm that the geomagnetic jerks seem to be related to sudden changes in the SYO/EYO time series and their excitation series; so we finally suggest that jerks are possible excitation sources of the SYO/EYO. Meanwhile, after using a deconvolution method, we estimate that the period P and quality factor Q of the SYO and the EYO are [P=5.85+/-0.06yr, Q larger than 180] and [P=8.455+/-0.17yr, Q larger than 350], respectively.

physics.geo-ph↗