Searcharxiv⌕ Search

arXiv subjects

Wei Li

Publications and source records attributed to Wei Li.

At least 757 records · Page 42Linked to original sources
arXiv2023

Balancing Logit Variation for Long-tailed Semantic Segmentation

Semantic segmentation usually suffers from a long-tail data distribution. Due to the imbalanced number of samples across categories, the features of those tail classes may get squeezed into a narrow area in the feature space. Towards a balanced feature distribution, we introduce category-wise variation into the network predictions in the training phase such that an instance is no longer projected to a feature point, but a small region instead. Such a perturbation is highly dependent on the category scale, which appears as assigning smaller variation to head classes and larger variation to tail classes. In this way, we manage to close the gap between the feature areas of different categories, resulting in a more balanced representation. It is noteworthy that the introduced variation is discarded at the inference stage to facilitate a confident prediction. Although with an embarrassingly simple implementation, our method manifests itself in strong generalizability to various datasets and task settings. Extensive experiments suggest that our plug-in design lends itself well to a range of state-of-the-art approaches and boosts the performance on top of them.

cs.CV↗
arXiv2023

On the Value of Myopic Behavior in Policy Reuse

Leveraging learned strategies in unfamiliar scenarios is fundamental to human intelligence. In reinforcement learning, rationally reusing the policies acquired from other tasks or human experts is critical for tackling problems that are difficult to learn from scratch. In this work, we present a framework called Selective Myopic bEhavior Control~(SMEC), which results from the insight that the short-term behaviors of prior policies are sharable across tasks. By evaluating the behaviors of prior policies via a hybrid value function architecture, SMEC adaptively aggregates the sharable short-term behaviors of prior policies and the long-term behaviors of the task policy, leading to coordinated decisions. Empirical results on a collection of manipulation and locomotion tasks demonstrate that SMEC outperforms existing methods, and validate the ability of SMEC to leverage related prior policies.

cs.LG↗
arXiv2023

WeCheck: Strong Factual Consistency Checker via Weakly Supervised Learning

A crucial issue of current text generation models is that they often uncontrollably generate factually inconsistent text with respective of their inputs. Limited by the lack of annotated data, existing works in evaluating factual consistency directly transfer the reasoning ability of models trained on other data-rich upstream tasks like question answering (QA) and natural language inference (NLI) without any further adaptation. As a result, they perform poorly on the real generated text and are biased heavily by their single-source upstream tasks. To alleviate this problem, we propose a weakly supervised framework that aggregates multiple resources to train a precise and efficient factual metric, namely WeCheck. WeCheck first utilizes a generative model to accurately label a real generated sample by aggregating its weak labels, which are inferred from multiple resources. Then, we train the target metric model with the weak supervision while taking noises into consideration. Comprehensive experiments on a variety of tasks demonstrate the strong performance of WeCheck, which achieves a 3.4\% absolute improvement over previous state-of-the-art methods on TRUE benchmark on average.

cs.CL↗
arXiv2023

Explore More Guidance: A Task-aware Instruction Network for Sign Language Translation Enhanced with Data Augmentation

Sign language recognition and translation first uses a recognition module to generate glosses from sign language videos and then employs a translation module to translate glosses into spoken sentences. Most existing works focus on the recognition step, while paying less attention to sign language translation. In this work, we propose a task-aware instruction network, namely TIN-SLT, for sign language translation, by introducing the instruction module and the learning-based feature fuse strategy into a Transformer network. In this way, the pre-trained model's language ability can be well explored and utilized to further boost the translation performance. Moreover, by exploring the representation space of sign language glosses and target spoken language, we propose a multi-level data augmentation scheme to adjust the data distribution of the training set. We conduct extensive experiments on two challenging benchmark datasets, PHOENIX-2014-T and ASLG-PC12, on which our method outperforms former best solutions by 1.65 and 1.42 in terms of BLEU-4. Our code is published at https://github.com/yongcaoplus/TIN-SLT.

cs.CL↗
arXiv2023

Language-Guided 3D Object Detection in Point Cloud for Autonomous Driving

This paper addresses the problem of 3D referring expression comprehension (REC) in autonomous driving scenario, which aims to ground a natural language to the targeted region in LiDAR point clouds. Previous approaches for REC usually focus on the 2D or 3D-indoor domain, which is not suitable for accurately predicting the location of the queried 3D region in an autonomous driving scene. In addition, the upper-bound limitation and the heavy computation cost motivate us to explore a better solution. In this work, we propose a new multi-modal visual grounding task, termed LiDAR Grounding. Then we devise a Multi-modal Single Shot Grounding (MSSG) approach with an effective token fusion strategy. It jointly learns the LiDAR-based object detector with the language features and predicts the targeted region directly from the detector without any post-processing. Moreover, the image feature can be flexibly integrated into our approach to provide rich texture and color information. The cross-modal learning enforces the detector to concentrate on important regions in the point cloud by considering the informative language expressions, thus leading to much better accuracy and efficiency. Extensive experiments on the Talk2Car dataset demonstrate the effectiveness of the proposed methods. Our work offers a deeper insight into the LiDAR-based grounding task and we expect it presents a promising direction for the autonomous driving community.

cs.CV↗
arXiv2023

Towards Tracing Code Provenance with Code Watermarking

Recent advances in large language models have raised wide concern in generating abundant plausible source code without scrutiny, and thus tracing the provenance of code emerges as a critical issue. To solve the issue, we propose CodeMark, a watermarking system that hides bit strings into variables respecting the natural and operational semantics of the code. For naturalness, we novelly introduce a contextual watermarking scheme to generate watermarked variables more coherent in the context atop graph neural networks. Each variable is treated as a node on the graph and the node feature gathers neighborhood (context) information through learning. Watermarks embedded into the features are thus reflected not only by the variables but also by the local contexts. We further introduce a pre-trained model on source code as a teacher to guide more natural variable generation. Throughout the embedding, the operational semantics are preserved as only variable names are altered. Beyond guaranteeing code-specific properties, CodeMark is superior in watermarking accuracy, capacity, and efficiency due to a more diversified pattern generated. Experimental results show CodeMark outperforms the SOTA watermarking systems with a better balance of the watermarking requirements.

cs.CR↗
arXiv2023

Quantification before Selection: Active Dynamics Preference for Robust Reinforcement Learning

Training a robust policy is critical for policy deployment in real-world systems or dealing with unknown dynamics mismatch in different dynamic systems. Domain Randomization~(DR) is a simple and elegant approach that trains a conservative policy to counter different dynamic systems without expert knowledge about the target system parameters. However, existing works reveal that the policy trained through DR tends to be over-conservative and performs poorly in target domains. Our key insight is that dynamic systems with different parameters provide different levels of difficulty for the policy, and the difficulty of behaving well in a system is constantly changing due to the evolution of the policy. If we can actively sample the systems with proper difficulty for the policy on the fly, it will stabilize the training process and prevent the policy from becoming over-conservative or over-optimistic. To operationalize this idea, we introduce Active Dynamics Preference~(ADP), which quantifies the informativeness and density of sampled system parameters. ADP actively selects system parameters with high informativeness and low density. We validate our approach in four robotic locomotion tasks with various discrepancies between the training and testing environments. Extensive results demonstrate that our approach has superior robustness for system inconsistency compared to several baselines.

cs.LG↗
arXiv2023

Open-Ended Diverse Solution Discovery with Regulated Behavior Patterns for Cross-Domain Adaptation

While Reinforcement Learning can achieve impressive results for complex tasks, the learned policies are generally prone to fail in downstream tasks with even minor model mismatch or unexpected perturbations. Recent works have demonstrated that a policy population with diverse behavior characteristics can generalize to downstream environments with various discrepancies. However, such policies might result in catastrophic damage during the deployment in practical scenarios like real-world systems due to the unrestricted behaviors of trained policies. Furthermore, training diverse policies without regulation of the behavior can result in inadequate feasible policies for extrapolating to a wide range of test conditions with dynamics shifts. In this work, we aim to train diverse policies under the regularization of the behavior patterns. We motivate our paradigm by observing the inverse dynamics in the environment with partial state information and propose Diversity in Regulation (DiR) training diverse policies with regulated behaviors to discover desired patterns that benefit the generalization. Considerable empirical results on various variations of different environments indicate that our method attains improvements over other diversity-driven counterparts.

cs.LG↗
arXiv2023

CodeVIO: Visual-Inertial Odometry with Learned Optimizable Dense Depth

In this work, we present a lightweight, tightly-coupled deep depth network and visual-inertial odometry (VIO) system, which can provide accurate state estimates and dense depth maps of the immediate surroundings. Leveraging the proposed lightweight Conditional Variational Autoencoder (CVAE) for depth inference and encoding, we provide the network with previously marginalized sparse features from VIO to increase the accuracy of initial depth prediction and generalization capability. The compact encoded depth maps are then updated jointly with navigation states in a sliding window estimator in order to provide the dense local scene geometry. We additionally propose a novel method to obtain the CVAE's Jacobian which is shown to be more than an order of magnitude faster than previous works, and we additionally leverage First-Estimate Jacobian (FEJ) to avoid recalculation. As opposed to previous works relying on completely dense residuals, we propose to only provide sparse measurements to update the depth code and show through careful experimentation that our choice of sparse measurements and FEJs can still significantly improve the estimated depth maps. Our full system also exhibits state-of-the-art pose estimation accuracy, and we show that it can run in real-time with single-thread execution while utilizing GPU acceleration only for the network and code Jacobian.

cs.CV↗
arXiv2023

Phonetic and Prosody-aware Self-supervised Learning Approach for Non-native Fluency Scoring

Speech fluency/disfluency can be evaluated by analyzing a range of phonetic and prosodic features. Deep neural networks are commonly trained to map fluency-related features into the human scores. However, the effectiveness of deep learning-based models is constrained by the limited amount of labeled training samples. To address this, we introduce a self-supervised learning (SSL) approach that takes into account phonetic and prosody awareness for fluency scoring. Specifically, we first pre-train the model using a reconstruction loss function, by masking phones and their durations jointly on a large amount of unlabeled speech and text prompts. We then fine-tune the pre-trained model using human-annotated scoring data. Our experimental results, conducted on datasets such as Speechocean762 and our non-native datasets, show that our proposed method outperforms the baseline systems in terms of Pearson correlation coefficients (PCC). Moreover, we also conduct an ablation study to better understand the contribution of phonetic and prosody factors during the pre-training stage.

cs.CL↗
arXiv2023

In-orbit performance of LE onboard Insight-HXMT in the first 5 years

Purpose: The Low-Energy X-ray telescope (LE) is a main instrument of the Insight-HXMT mission and consists of 96 Swept Charge Devices (SCD) covering the 1-10 keV energy band. The energy gain and resolution are continuously calibrated by analysing Cassiopeia A (Cas A) and blank sky data, while the effective areas are also calibrated with the observations of the Crab Nebula. In this paper, we present the evolution of the in-orbit performances of LE in the first 5 years since launch. Methods: The Insight-HXMT Data Analysis Software package (HXMTDAS) is utilized to extract the spectra of Cas A, blank sky, and Crab Nebula using different Good Time Interval (GTI) selections. We fit a model with a power-law continuum and several Gaussian lines to different ranges of Cas A and blank sky spectra to get peak energies of their lines through xspec. After updating the energy gain calibration in CALibration DataBase (CALDB), we rerun the Cas A data to obtain the energy resolution. An empirical function is used to modify the simulated effective areas so that the background-subtracted spectrum of the Crab Nebula can best match the standard model of the Crab Nebula. Results: The energy gain, resolution, and effective areas are calibrated every month. The corresponding calibration results are duly updated in CALDB, which can be downloaded and used for the analysis of Insight-HXMT data. Simultaneous observations with NuSTAR and NICER can also be used to verify our derived results. Conclusion: LE is a well calibrated X-ray telescope working in 1-10 keV band. The uncertainty of LE gain is less than 20 eV in 2-9 keV band and the uncertainty of LE resolution is less than 15eV. The systematic errors of LE, compared to the model of the Crab Nebula, are lower than 1.5% in 1-10 keV.

astro-ph.IM↗
arXiv2023

Electrically programmable magnetic coupling in an Ising network exploiting solid-state ionic gating

Two-dimensional arrays of magnetically coupled nanomagnets provide a mesoscopic platform for exploring collective phenomena as well as realizing a broad range of spintronic devices. In particular, the magnetic coupling plays a critical role in determining the nature of the cooperative behaviour and providing new functionalities in nanomagnet-based devices. Here, we create coupled Ising-like nanomagnets in which the coupling between adjacent nanomagnetic regions can be reversibly converted between parallel and antiparallel through solid-state ionic gating. This is achieved with the voltage-control of magnetic anisotropies in a nanosized region where the symmetric exchange interaction favours parallel alignment and the antisymmetric exchange interaction, namely the Dzyaloshinskii-Moriya interaction, favours antiparallel alignment. Applying this concept to a two-dimensional lattice, we demonstrate a voltage-controlled phase transition in artificial spin ices. Furthermore, we achieve an addressable control of the individual couplings and realize an electrically programmable Ising network, which opens up new avenues to design nanomagnet-based logic devices and neuromorphic computers

cond-mat.mes-hall↗
arXiv2023

Masked Frequency Modeling for Self-Supervised Visual Pre-Training

We present Masked Frequency Modeling (MFM), a unified frequency-domain-based approach for self-supervised pre-training of visual models. Instead of randomly inserting mask tokens to the input embeddings in the spatial domain, in this paper, we shift the perspective to the frequency domain. Specifically, MFM first masks out a portion of frequency components of the input image and then predicts the missing frequencies on the frequency spectrum. Our key insight is that predicting masked components in the frequency domain is more ideal to reveal underlying image patterns rather than predicting masked patches in the spatial domain, due to the heavy spatial redundancy. Our findings suggest that with the right configuration of mask-and-predict strategy, both the structural information within high-frequency components and the low-level statistics among low-frequency counterparts are useful in learning good representations. For the first time, MFM demonstrates that, for both ViT and CNN, a simple non-Siamese framework can learn meaningful representations even using none of the following: (i) extra data, (ii) extra model, (iii) mask token. Experimental results on image classification and semantic segmentation, as well as several robustness benchmarks show the competitive performance and advanced robustness of MFM compared with recent masked image modeling approaches. Furthermore, we also comprehensively investigate the effectiveness of classical image restoration tasks for representation learning from a unified frequency perspective and reveal their intriguing relations with our MFM approach.

cs.CV↗
arXiv2023

An Iterative Classification and Semantic Segmentation Network for Old Landslide Detection Using High-Resolution Remote Sensing Images

Huge challenges exist for old landslide detection because their morphology features have been partially or strongly transformed over a long time and have little difference from their surrounding. Besides, small-sample problem also restrict in-depth learning. In this paper, an iterative classification and semantic segmentation network (ICSSN) is developed, which can greatly enhance both object-level and pixel-level classification performance by iteratively upgrading the feature extractor shared by two network. An object-level contrastive learning (OCL) strategy is employed in the object classification sub-network featuring a siamese network to realize the global features extraction, and a sub-object-level contrastive learning (SOCL) paradigm is designed in the semantic segmentation sub-network to efficiently extract salient features from boundaries of landslides. Moreover, an iterative training strategy is elaborated to fuse features in semantic space such that both object-level and pixel-level classification performance are improved. The proposed ICSSN is evaluated on the real landslide data set, and the experimental results show that ICSSN can greatly improve the classification and segmentation accuracy of old landslide detection. For the semantic segmentation task, compared to the baseline, the F1 score increases from 0.5054 to 0.5448, the mIoU improves from 0.6405 to 0.6610, the landslide IoU improved from 0.3381 to 0.3743, and the object-level detection accuracy of old landslides is enhanced from 0.55 to 0.9. For the object classification task, the F1 score increases from 0.8846 to 0.9230, and the accuracy score is up from 0.8375 to 0.8875.

cs.CV↗
arXiv2023

Exploring a CNN Model for Earthquake Magnitude Estimation using HR-GNSS data

High rate Global Navigation Satellite System (HR GNSS) data can be highly useful for earthquake analysis as it provides continuous high-rate measurements of ground motion. This data can be used to estimate the magnitude, to assess the potential of an earthquake for generating tsunamis, and to analyze diverse parameters related to the seismic source. Particularly, in this work, we present the first results of a deep learning model based on a convolutional neural network for earthquake magnitude estimation, using HR GNSS displacement time series. The influence of different dataset configurations, such as station numbers, epicentral distances, signal duration, and earthquake size, were analyzed to figure out how the model can be adapted to various scenarios. We explored the potential of the model for global application and compared its performance using both synthetic and real data from different seismogenic regions. The performance of our model at this stage was satisfactory in estimating earthquake magnitude from synthetic data. Comparable results were observed in tests using synthetic data from a different tectonic region than the training data. Furthermore, the model was tested using real data from different regions and magnitudes, resulting in a good accuracy, provided that the data from a particular group of stations had similar epicentral distance constraints to those used during the model training. The robustness of the DL model can be improved to work independently from the window size of the time series and the number of stations, enabling faster estimation by the model using only near field data. Overall, this study provides insights for the development of future DL approaches for earthquake magnitude estimation with HR-GNSS data, emphasizing the importance of proper handling and careful data selection for further model improvements.

physics.geo-ph↗
arXiv2023

Hybrid plasmonic modes for enhanced refractive index sensing

Compared to single nanoparticles, strongly coupled plasmonic nanoparticles provide attractive advantages owing to their ability to exhibit multiple resonances with unique spectral features and higher local field intensity. These enhanced plasmonic properties of coupled metal nanoparticles have been used for various applications including realization of strong light-matter interaction, photocatalysis, and sensing applications. In this article, we review the basic physics of hybrid plasmonic modes in coupled metallic nanodimers and assess their potentials for refractive index sensing. In particular, we overview various modes of hybrid plasmons including bonding and antibonding modes in symmetric nanodimers, Fano resonances in asymmetric nanodimers, charge transfer plasmons in linked nanoparticle dimers, hybrid plasmon modes in nanoshells, and gap modes in particle-on-mirror configurations. Beyond the dimeric nanosystems, we also showcase the potentials of hybrid plasmonic modes in periodic nanoparticle arrays for sensing applications. Finally, based on the critical assessment of the recent researches on coupled plasmonic modes, the outlook on the future prospects of hybrid plasmon based refractometric sensing are discussed We believe that, given their tunable resonances and ultranarrow spectral signatures, coupled metal nanoparticles are expected to play key roles in developing precise plasmonic nanodevices with extreme sensitivity.

physics.optics↗
arXiv2023

Nematic chiral spin liquid in a Kitaev magnet under external magnetic field

The possible existence of a quantum spin liquid (QSL) phase, an exotic state of matter with long-range quantum entanglement and fractionalized excitations, in $α$-RuCl$_3$ has sparked widespread interests in exploring QSLs in various Kitaev models under magnetic fields. Recently, a $K$-$J$-$Γ$-$Γ'$ model has been proposed to accurately describe the compound by fitting the measured thermodynamic data, where $J$ is Heisenberg interaction, and $Γ$, $Γ'$ are off-diagonal exchanges on top of the dominant Kitaev coupling $K$. Based on this effective model, an intermediate QSL phase in presence of an out-of-plane fields along the $[1 1 1]$ direction, between the low-field zigzag order and high-field polarized phase, has been predicted. By combining density matrix renormalization group (DMRG), exponential tensor renormalization group (XTRG), and variational Monte Carlo (VMC) calculations, we address the nature of this QSL phase in the honeycomb $K$-$J$-$Γ$-$Γ'$ model under the $[1 1 1]$-direction field. Our DMRG calculations find the algebraic-like decay of spin correlation function, the finite spin scalar chiral order, and lattice nematic order. Together with the XTRG results of power-law specific heat at low temperature, our findings naturally suggest a gapless nematic chiral spin liquid. On the other hand, our VMC study finds a gapped nematic chiral spin liquid with the variational energy very close to that obtained in DMRG. As VMC finds a very small gap that is beyond the current resolution of both ground-state DMRG and finite-temperature XTRG calculations on finite-width cylinders, we resort the full clarification for the nature of the intermediate QSL to future studies. Lastly, we discuss the implications of our results to the recent experiment on $α$-RuCl$_3$ and the QSL-like phase in a similar $K$-$Γ$-$Γ'$ model.

cond-mat.str-el↗
arXiv2023

Spontaneous rotational symmetry breaking in KTaO$_3$ heterointerface superconductors

Broken symmetries play a fundamental role in superconductivity and influence many of its properties in a profound way. Understanding these symmetry breaking states is essential to elucidate the various exotic quantum behaviors in non-trivial superconductors. Here, we report an experimental observation of spontaneous rotational symmetry breaking of superconductivity at the heterointerface of amorphous (a)-YAlO$_3$/KTaO$_3$(111) with a superconducting transition temperature of 1.86 K. Both the magnetoresistance and superconducting critical field in an in-plane field manifest striking twofold symmetric oscillations deep inside the superconducting state, whereas the anisotropy vanishes in the normal state, demonstrating that it is an intrinsic property of the superconducting phase. We attribute this behavior to the mixed-parity superconducting state, which is an admixture of \emph{s}-wave and \emph{p}-wave pairing components induced by strong spin-orbit coupling inherent to inversion symmetry breaking at the heterointerface of a-YAlO$_3$/KTaO$_3$. Our work suggests an unconventional nature of the underlying pairing interaction in the KTaO$_3$ heterointerface superconductors, and brings a new broad of perspective on understanding non-trivial superconducting properties at the artificial heterointerfaces.

cond-mat.supr-con↗