Han Wang

Liliang Ren, Zixuan Zhang, Han Wang et al. · microsoft-research

Modern large-scale Pre-trained Language Models (PLMs) have achieved tremendous success on a wide range of downstream tasks. However, most of the LM pre-training objectives only focus on text reconstruction, but have not sought to learn latent-level interpretable representations of sentences. In this paper, we manage to push the language models to obtain a deeper understanding of sentences by proposing a new pre-training objective, Sparse Latent Typing, which enables the model to sparsely extract sentence-level keywords with diverse latent types. Experimental results show that our model is able to learn interpretable latent type categories in a self-supervised manner without using any external knowledge. Besides, the language model pre-trained with such an objective also significantly improves Information Extraction related downstream tasks in both supervised and few-shot settings. Our code is publicly available at: https://github.com/renll/SparseLT.

Han Wang, Jun Tang, Xiaodong Liu et al.

Recent years have witnessed a trend of applying context frames to boost the performance of object detection as video object detection. Existing methods usually aggregate features at one stroke to enhance the feature. These methods, however, usually lack spatial information from neighboring frames and suffer from insufficient feature aggregation. To address the issues, we perform a progressive way to introduce both temporal information and spatial information for an integrated enhancement. The temporal information is introduced by the temporal feature aggregation model (TFAM), by conducting an attention mechanism between the context frames and the target frame (i.e., the frame to be detected). Meanwhile, we employ a Spatial Transition Awareness Model (STAM) to convey the location transition information between each context frame and target frame. Built upon a transformer-based detector DETR, our PTSEFormer also follows an end-to-end fashion to avoid heavy post-processing procedures while achieving 88.1% mAP on the ImageNet VID dataset. Codes are available at https://github.com/Hon-Wong/PTSEFormer.

Chenjun Xiao, Han Wang, Yangchen Pan et al. · deepmind

Reinforcement learning (RL) agents can leverage batches of previously collected data to extract a reasonable control policy. An emerging issue in this offline RL setting, however, is that the bootstrapping update underlying many of our methods suffers from insufficient action-coverage: standard max operator may select a maximal action that has not been seen in the dataset. Bootstrapping from these inaccurate values can lead to overestimation and even divergence. There are a growing number of methods that attempt to approximate an \emph{in-sample} max, that only uses actions well-covered by the dataset. We highlight a simple fact: it is more straightforward to approximate an in-sample \emph{softmax} using only actions in the dataset. We show that policy iteration based on the in-sample softmax converges, and that for decreasing temperatures it approaches the in-sample max. We derive an In-Sample Actor-Critic (AC), using this in-sample softmax, and show that it is consistently better or comparable to existing offline RL methods, and is also well-suited to fine-tuning.

Peng Cheng, Xianyuan Zhan, Zhihao Wu et al.

Offline reinforcement learning (RL) offers an appealing approach to real-world tasks by learning policies from pre-collected datasets without interacting with the environment. However, the performance of existing offline RL algorithms heavily depends on the scale and state-action space coverage of datasets. Real-world data collection is often expensive and uncontrollable, leading to small and narrowly covered datasets and posing significant challenges for practical deployments of offline RL. In this paper, we provide a new insight that leveraging the fundamental symmetry of system dynamics can substantially enhance offline RL performance under small datasets. Specifically, we propose a Time-reversal symmetry (T-symmetry) enforced Dynamics Model (TDM), which establishes consistency between a pair of forward and reverse latent dynamics. TDM provides both well-behaved representations for small datasets and a new reliability measure for OOD samples based on compliance with the T-symmetry. These can be readily used to construct a new offline RL algorithm (TSRL) with less conservative policy constraints and a reliable latent space data augmentation procedure. Based on extensive experiments, we find TSRL achieves great performance on small benchmark datasets with as few as 1% of the original samples, which significantly outperforms the recent offline RL algorithms in terms of data efficiency and generalizability.Code is available at: https://github.com/pcheng2/TSRL

Han Wang, Archit Sakhadeo, Adam White et al. · deepmind

The performance of reinforcement learning (RL) agents is sensitive to the choice of hyperparameters. In real-world settings like robotics or industrial control systems, however, testing different hyperparameter configurations directly on the environment can be financially prohibitive, dangerous, or time consuming. We propose a new approach to tune hyperparameters from offline logs of data, to fully specify the hyperparameters for an RL agent that learns online in the real world. The approach is conceptually simple: we first learn a model of the environment from the offline data, which we call a calibration model, and then simulate learning in the calibration model to identify promising hyperparameters. We identify several criteria to make this strategy effective, and develop an approach that satisfies these criteria. We empirically investigate the method in a variety of settings to identify when it is effective and when it fails.

Mo Yu, Qiujing Wang, Shunchi Zhang et al. · ibm-research

When reading a story, humans can quickly understand new fictional characters with a few observations, mainly by drawing analogies to fictional and real people they already know. This reflects the few-shot and meta-learning essence of humans' inference of characters' mental states, i.e., theory-of-mind (ToM), which is largely ignored in existing research. We fill this gap with a novel NLP dataset, ToM-in-AMC, the first assessment of machines' meta-learning of ToM in a realistic narrative understanding scenario. Our dataset consists of ~1,000 parsed movie scripts, each corresponding to a few-shot character understanding task that requires models to mimic humans' ability of fast digesting characters with a few starting scenes in a new movie. We propose a novel ToM prompting approach designed to explicitly assess the influence of multiple ToM dimensions. It surpasses existing baseline models, underscoring the significance of modeling multiple ToM dimensions for our task. Our extensive human study verifies that humans are capable of solving our problem by inferring characters' mental states based on their previously seen movies. In comparison, our systems based on either state-of-the-art large language models (GPT-4) or meta-learning algorithms lags >20% behind, highlighting a notable limitation in existing approaches' ToM capabilities.

Han Wang, Canwen Xu, Julian McAuley

Prompt-based learning (i.e., prompting) is an emerging paradigm for exploiting knowledge learned by a pretrained language model. In this paper, we propose Automatic Multi-Label Prompting (AMuLaP), a simple yet effective method to automatically select label mappings for few-shot text classification with prompting. Our method exploits one-to-many label mappings and a statistics-based algorithm to select label mappings given a prompt template. Our experiments demonstrate that AMuLaP achieves competitive performance on the GLUE benchmark without human effort or external resources.

Habib Ammari, Daewon Chung, Hyeonbae Kang et al.

We derive transformation formulas for the generalized polarization tensors under rigid motions and scaling in three dimensions, and use them to construct an infinite number of invariants under those transformations. These invariants can be used as shape descriptors for dictionary matching.

Han Wang, Aritra Mitra, Hamed Hassani et al.

We initiate the study of federated reinforcement learning under environmental heterogeneity by considering a policy evaluation problem. Our setup involves $N$ agents interacting with environments that share the same state and action space but differ in their reward functions and state transition kernels. Assuming agents can communicate via a central server, we ask: Does exchanging information expedite the process of evaluating a common policy? To answer this question, we provide the first comprehensive finite-time analysis of a federated temporal difference (TD) learning algorithm with linear function approximation, while accounting for Markovian sampling, heterogeneity in the agents' environments, and multiple local updates to save communication. Our analysis crucially relies on several novel ingredients: (i) deriving perturbation bounds on TD fixed points as a function of the heterogeneity in the agents' underlying Markov decision processes (MDPs); (ii) introducing a virtual MDP to closely approximate the dynamics of the federated TD algorithm; and (iii) using the virtual MDP to make explicit connections to federated optimization. Putting these pieces together, we rigorously prove that in a low-heterogeneity regime, exchanging model estimates leads to linear convergence speedups in the number of agents.

Shangyu Xie, Han Wang, Shengbin Wang et al.

Smart grid has integrated an increasing number of distributed energy resources to improve the efficiency and flexibility of power generation and consumption as well as the resilience of the power grid. The energy consumers on the power grid (e.g., households) equipped with the distributed energy resources can be considered as "microgrids" that both generate and consume electricity. In this paper, we study the energy community discovery problems which identify multiple kinds of energy communities for the microgrids to facilitate energy management (e.g., power supply adjustment, load balancing, energy sharing) on the grid, such as homogeneous energy communities (HECs), mixed energy communities (MECs), and self-sufficient energy communities (SECs). Specifically, we present efficient algorithms to discover such communities of microgrids by taking into account not only their geo-locations but also their net energy over any period. Finally, we experimentally validate the performance of the algorithms using both synthetic and real datasets.

Junda Wang, Weijian Li, Han Wang et al.

Causal inference and model interpretability research are gaining increasing attention, especially in the domains of healthcare and bioinformatics. Despite recent successes in this field, decorrelating features under nonlinear environments with human interpretable representations has not been adequately investigated. To address this issue, we introduce a novel method with a variable decorrelation regularizer to handle both linear and nonlinear confounding. Moreover, we employ association rules as new representations using association rule mining based on the original features to further proximate human decision patterns to increase model interpretability. Extensive experiments are conducted on four healthcare datasets (one synthetically generated and three real-world collections on different diseases). Quantitative results in comparison to baseline approaches on parameter estimation and causality computation indicate the model's superior performance. Furthermore, expert evaluation given by healthcare professionals validates the effectiveness and interpretability of the proposed model.

An Zhang, Han Wang, Xiang Wang et al.

Domain Generalization (DG), designed to enhance out-of-distribution (OOD) generalization, is all about learning invariance against domain shifts utilizing sufficient supervision signals. Yet, the scarcity of such labeled data has led to the rise of unsupervised domain generalization (UDG) - a more important yet challenging task in that models are trained across diverse domains in an unsupervised manner and eventually tested on unseen domains. UDG is fast gaining attention but is still far from well-studied. To close the research gap, we propose a novel learning framework designed for UDG, termed the Disentangled Masked Auto Encoder (DisMAE), aiming to discover the disentangled representations that faithfully reveal the intrinsic features and superficial variations without access to the class label. At its core is the distillation of domain-invariant semantic features, which cannot be distinguished by domain classifier, while filtering out the domain-specific variations (for example, color schemes and texture patterns) that are unstable and redundant. Notably, DisMAE co-trains the asymmetric dual-branch architecture with semantic and lightweight variation encoders, offering dynamic data manipulation and representation level augmentation capabilities. Extensive experiments on four benchmark datasets (i.e., DomainNet, PACS, VLCS, Colored MNIST) with both DG and UDG tasks demonstrate that DisMAE can achieve competitive OOD performance compared with the state-of-the-art DG and UDG baselines, which shed light on potential research line in improving the generalization ability with large-scale unlabeled data.

Yangjun Wu, Kebin Fang, Dongxiang Zhang et al.

Structured dropout approaches, such as attention dropout and DropHead, have been investigated to regularize the multi-head attention mechanism in Transformers. In this paper, we propose a new regularization scheme based on token-level rather than structure-level to reduce overfitting. Specifically, we devise a novel Token-Level Masking (TLM) training strategy for Transformers to regularize the connections of self-attention, which consists of two masking techniques that are effective and easy to implement. The underlying idea is to manipulate the connections between tokens in the multi-head attention via masking, where the networks are forced to exploit partial neighbors' information to produce a meaningful representation. The generality and effectiveness of TLM are thoroughly evaluated via extensive experiments on 4 diversified NLP tasks across 18 datasets, including natural language understanding benchmark GLUE, ChineseGLUE, Chinese Grammatical Error Correction, and data-to-text generation. The results indicate that TLM can consistently outperform attention dropout and DropHead, e.g., it increases by 0.5 points relative to DropHead with BERT-large on GLUE. Moreover, TLM can establish a new record on the data-to-text benchmark Rotowire (18.93 BLEU). Our code will be publicly available at https://github.com/Young1993/tlm.

Anil N. Hirani, Kaushik Kalyanaraman, Han Wang et al.

We describe algorithms for finding harmonic cochains, an essential ingredient for solving elliptic partial differential equations in exterior calculus. Harmonic cochains are also useful in computational topology and computer graphics. We focus on finding harmonic cochains cohomologous to a given cocycle. Amongst other things, this allows localization near topological features of interest. We derive a weighted least squares method by proving a discrete Hodge-deRham theorem on the isomorphism between the space of harmonic cochains and cohomology. The solution obtained either satisfies the Whitney form finite element exterior calculus equations or the discrete exterior calculus equations for harmonic cochains, depending on the discrete Hodge star used.

Zhijian Liu, Nian Cai, Wensheng Ouyang et al.

Automatic hardhat wearing detection can strengthen the safety management in construction sites, which is still challenging due to complicated video surveillance scenes. To deal with the poor generalization of previous deep learning based methods, a novel anchor-free deep learning framework called CA-CentripetalNet is proposed for hardhat wearing detection. Two novel schemes are proposed to improve the feature extraction and utilization ability of CA-CentripetalNet, which are vertical-horizontal corner pooling and bounding constrained center attention. The former is designed to realize the comprehensive utilization of marginal features and internal features. The latter is designed to enforce the backbone to pay attention to internal features, which is only used during the training rather than during the detection. Experimental results indicate that the CA-CentripetalNet achieves better performance with the 86.63% mAP (mean Average Precision) with less memory consumption at a reasonable speed than the existing deep learning based methods, especially in case of small-scale hardhats and non-worn-hardhats.

Guangchen Lan, Han Wang, James Anderson et al.

Federated reinforcement learning (FedRL) enables agents to collaboratively train a global policy without sharing their individual data. However, high communication overhead remains a critical bottleneck, particularly for natural policy gradient (NPG) methods, which are second-order. To address this issue, we propose the FedNPG-ADMM framework, which leverages the alternating direction method of multipliers (ADMM) to approximate global NPG directions efficiently. We theoretically demonstrate that using ADMM-based gradient updates reduces communication complexity from ${O}({d^{2}})$ to ${O}({d})$ at each iteration, where $d$ is the number of model parameters. Furthermore, we show that achieving an $ε$-error stationary convergence requires ${O}(\frac{1}{(1-γ)^{2}ε})$ iterations for discount factor $γ$, demonstrating that FedNPG-ADMM maintains the same convergence rate as the standard FedNPG. Through evaluation of the proposed algorithms in MuJoCo environments, we demonstrate that FedNPG-ADMM maintains the reward performance of standard FedNPG, and that its convergence rate improves when the number of federated agents increases.

Han Wang, Jiayuan Zhang, Lipeng Wan et al.

Manipulation relationship detection (MRD) aims to guide the robot to grasp objects in the right order, which is important to ensure the safety and reliability of grasping in object stacked scenes. Previous works infer manipulation relationship by deep neural network trained with data collected from a predefined view, which has limitation in visual dislocation in unstructured environments. Multi-view data provide more comprehensive information in space, while a challenge of multi-view MRD is domain shift. In this paper, we propose a novel multi-view fusion framework, namely multi-view MRD network (MMRDN), which is trained by 2D and 3D multi-view data. We project the 2D data from different views into a common hidden space and fit the embeddings with a set of Von-Mises-Fisher distributions to learn the consistent representations. Besides, taking advantage of position information within the 3D data, we select a set of $K$ Maximum Vertical Neighbors (KMVN) points from the point cloud of each object pair, which encodes the relative position of these two objects. Finally, the features of multi-view 2D and 3D data are concatenated to predict the pairwise relationship of objects. Experimental results on the challenging REGRAD dataset show that MMRDN outperforms the state-of-the-art methods in multi-view MRD tasks. The results also demonstrate that our model trained by synthetic data is capable to transfer to real-world scenarios.

Han Wang, Archiki Prasad, Elias Stengel-Eskin et al.

Generations from large language models (LLMs) can be improved by sampling and scoring multiple solutions to select a final answer. Current "sample and select" methods such as self-consistency (SC) rely on majority voting to score answers. However, when tasks have many distinct and valid answers, selection by voting requires a large number of samples. This makes SC prohibitively expensive for interactive tasks that involve generating multiple actions (answers) sequentially. After establishing that majority voting fails to provide consistent gains on such tasks, we demonstrate how to increase success rates by softening the scoring criterion. We introduce Soft Self-Consistency (SOFT-SC), which replaces SC's discontinuous scoring with a continuous score computed from model likelihoods, allowing for selection even when actions are sparsely distributed. SOFT-SC improves both performance and efficiency on long-horizon interactive tasks, requiring half as many samples as SC for comparable or better performance. For a fixed number of samples, SOFT-SC leads to a 1.3% increase over SC in absolute success rate on writing bash programs, a 6.6% increase on online shopping (WebShop), and a 4.7% increase for an interactive household game (ALFWorld). Finally, we show that SOFT-SC can be applied to both open-source and black-box models.

Peiyuan Chen, Zecheng Zhang, Yiping Dong et al.

Visual Question Answering (VQA) is a challenging task that requires systems to provide accurate answers to questions based on image content. Current VQA models struggle with complex questions due to limitations in capturing and integrating multimodal information effectively. To address these challenges, we propose the Rank VQA model, which leverages a ranking-inspired hybrid training strategy to enhance VQA performance. The Rank VQA model integrates high-quality visual features extracted using the Faster R-CNN model and rich semantic text features obtained from a pre-trained BERT model. These features are fused through a sophisticated multimodal fusion technique employing multi-head self-attention mechanisms. Additionally, a ranking learning module is incorporated to optimize the relative ranking of answers, thus improving answer accuracy. The hybrid training strategy combines classification and ranking losses, enhancing the model's generalization ability and robustness across diverse datasets. Experimental results demonstrate the effectiveness of the Rank VQA model. Our model significantly outperforms existing state-of-the-art models on standard VQA datasets, including VQA v2.0 and COCO-QA, in terms of both accuracy and Mean Reciprocal Rank (MRR). The superior performance of Rank VQA is evident in its ability to handle complex questions that require understanding nuanced details and making sophisticated inferences from the image and text. This work highlights the effectiveness of a ranking-based hybrid training strategy in improving VQA performance and lays the groundwork for further research in multimodal learning methods.

Li Zhang, Wenhao Li, Haotian Guan et al.

The identification of compound-protein interactions (CPI) plays a critical role in drug screening, drug repurposing, and combination therapy studies. The effectiveness of CPI prediction relies heavily on the features extracted from both compounds and target proteins. While various prediction methods employ different feature combinations, both molecular-based and network-based models encounter the common obstacle of incomplete feature representations. Thus, a promising solution to this issue is to fully integrate all relevant CPI features. This study proposed a novel model named MCPI, which is designed to improve the prediction performance of CPI by integrating multiple sources of information, including the PPI network, CCI network, and structural features of CPI. The results of the study indicate that the MCPI model outperformed other existing methods for predicting CPI on public datasets. Furthermore, the study has practical implications for drug development, as the model was applied to search for potential inhibitors among FDA-approved drugs in response to the SARS-CoV-2 pandemic. The prediction results were then validated through the literature, suggesting that the MCPI model could be a useful tool for identifying potential drug candidates. Overall, this study has the potential to advance our understanding of CPI and guide drug development efforts.

Leonardo F. Toso, Han Wang, James Anderson

We investigate the problem of learning an $ε$-approximate solution for the discrete-time Linear Quadratic Regulator (LQR) problem via a Stochastic Variance-Reduced Policy Gradient (SVRPG) approach. Whilst policy gradient methods have proven to converge linearly to the optimal solution of the model-free LQR problem, the substantial requirement for two-point cost queries in gradient estimations may be intractable, particularly in applications where obtaining cost function evaluations at two distinct control input configurations is exceptionally costly. To this end, we propose an oracle-efficient approach. Our method combines both one-point and two-point estimations in a dual-loop variance-reduced algorithm. It achieves an approximate optimal solution with only $O\left(\log\left(1/ε\right)^β\right)$ two-point cost information for $β\in (0,1)$.

Lingwei Zhu, Han Wang, Yukie Nagai

Sparse continuous policies are distributions that can choose some actions at random yet keep strictly zero probability for the other actions, which are radically different from the Gaussian. They have important real-world implications, e.g. in modeling safety-critical tasks like medicine. The combination of offline reinforcement learning and sparse policies provides a novel paradigm that enables learning completely from logged datasets a safety-aware sparse policy. However, sparse policies can cause difficulty with the existing offline algorithms which require evaluating actions that fall outside of the current support. In this paper, we propose the first offline policy optimization algorithm that tackles this challenge: Fat-to-Thin Policy Optimization (FtTPO). Specifically, we maintain a fat (heavy-tailed) proposal policy that effectively learns from the dataset and injects knowledge to a thin (sparse) policy, which is responsible for interacting with the environment. We instantiate FtTPO with the general $q$-Gaussian family that encompasses both heavy-tailed and sparse policies and verify that it performs favorably in a safety-critical treatment simulation and the standard MuJoCo suite. Our code is available at \url{https://github.com/lingweizhu/fat2thin}.

Stephen H. Bach, Victor Sanh, Zheng-Xin Yong et al.

PromptSource is a system for creating, sharing, and using natural language prompts. Prompts are functions that map an example from a dataset to a natural language input and target output. Using prompts to train and query language models is an emerging area in NLP that requires new tools that let users develop and refine these prompts collaboratively. PromptSource addresses the emergent challenges in this new setting with (1) a templating language for defining data-linked prompts, (2) an interface that lets users quickly iterate on prompt development by observing outputs of their prompts on many examples, and (3) a community-driven set of guidelines for contributing new prompts to a common pool. Over 2,000 prompts for roughly 170 datasets are already available in PromptSource. PromptSource is available at https://github.com/bigscience-workshop/promptsource.

Mahdi Abolfazli Esfahani, Han Wang

Sun Glare widely exists in the images captured by unmanned ground and aerial vehicles performing in outdoor environments. The existence of such artifacts in images will result in wrong feature extraction and failure of autonomous systems. Humans will try to adapt their view once they observe a glare (especially when driving), and this behavior is an essential requirement for the next generation of autonomous vehicles. The source of glare is not limited to the sun, and glare can be seen in the images captured during the nighttime and in indoor environments, which is due to the presence of different light sources; reflective surfaces also influence the generation of such artifacts. The glare's visual characteristics are different on images captured by various cameras and depend on several factors such as the camera's shutter speed and exposure level. Hence, it is challenging to introduce a general - robust and accurate - algorithm for glare detection that can perform well in various captured images. This research aims to introduce the first dataset for glare detection, which includes images captured by different cameras. Besides, the effect of multiple image representations and their combination in glare detection is examined using the proposed deep network architecture. The released dataset is available at https://github.com/maesfahani/glaredetection

Hanbo Zhang, Deyu Yang, Han Wang et al.

Despite the impressive progress achieved in robotic grasping, robots are not skilled in sophisticated tasks (e.g. search and grasp a specified target in clutter). Such tasks involve not only grasping but the comprehensive perception of the world (e.g. the object relationships). Recently, encouraging results demonstrate that it is possible to understand high-level concepts by learning. However, such algorithms are usually data-intensive, and the lack of data severely limits their performance. In this paper, we present a new dataset named REGRAD for the learning of relationships among objects and grasps. We collect the annotations of object poses, segmentations, grasps, and relationships for the target-driven relational grasping tasks. Our dataset is collected in both forms of 2D images and 3D point clouds. Moreover, since all the data are generated automatically, it is free to import new objects for data generation. We also released a real-world validation dataset to evaluate the sim-to-real performance of models trained on REGRAD. Finally, we conducted a series of experiments, showing that the models trained on REGRAD could generalize well to the realistic scenarios, in terms of both relationship and grasp detection. Our dataset and code could be found at: https://github.com/poisonwine/REGRAD

Han Wang, Chen Wang, Lihua Xie

As an essential component of visual simultaneous localization and mapping (SLAM), place recognition is crucial for robot navigation and autonomous driving. Existing methods often formulate visual place recognition as feature matching, which is computationally expensive for many robotic applications with limited computing power, e.g., autonomous driving and cleaning robot. Inspired by the fact that human beings always recognize a place by remembering salient regions or landmarks that are more attractive or interesting than others, we formulate visual place recognition as saliency re-identification. In the meanwhile, we propose to perform both saliency detection and re-identification in frequency domain, in which all operations become element-wise. The experiments show that our proposed method achieves competitive accuracy and much higher speed than the state-of-the-art feature-based methods. The proposed method is open-sourced and available at https://github.com/wh200720041/SRLCD.git.

Guoyi Zhang, Siyang Chen, Guangsheng Xu et al.

Small moving target detection is crucial for many defense applications but remains highly challenging due to low signal-to-noise ratios, ambiguous visual cues, and cluttered backgrounds. In this work, we propose a novel deep learning framework that differs fundamentally from existing approaches, which often rely on target-specific features or motion cues and tend to lack robustness in complex environments. Our key insight is that small target detection and background discrimination are inherently coupled, even cluttered video backgrounds often exhibit strong low-rank structures that can serve as stable priors for detection. We reformulate the task as a tensor-based low-rank and sparse decomposition problem and conduct a theoretical analysis of the background, target, and noise components to guide model design. Building on these insights, we introduce TenRPCANet, a deep neural network that requires minimal assumptions about target characteristics. Specifically, we propose a tokenization strategy that implicitly enforces multi-order tensor low-rank priors through a self-attention mechanism. This mechanism captures both local and non-local self-similarity to model the low-rank background without relying on explicit iterative optimization. In addition, inspired by the sparse component update in tensor RPCA, we design a feature refinement module to enhance target saliency. The proposed method achieves state-of-the-art performance on two highly distinct and challenging tasks: multi-frame infrared small target detection and space object detection. These results demonstrate both the effectiveness and the generalizability of our approach.

Han Wang, Yuang Zhang, Yuhong Zhang et al.

Colorization is a traditional computer vision task and it plays an important role in many time-consuming tasks, such as old film restoration. Existing methods suffer from unsaturated color and temporally inconsistency. In this paper, we propose a novel pipeline to overcome the challenges. We regard the colorization task as a generative task and introduce Stable Video Diffusion (SVD) as our base model. We design a palette-based color guider to assist the model in generating vivid and consistent colors. The color context introduced by the palette not only provides guidance for color generation, but also enhances the stability of the generated colors through a unified color context across multiple sequences. Experiments demonstrate that the proposed method can provide vivid and stable colors for videos, surpassing previous methods.

Enze Hou, Xiaoyang Wang, Han Wang

Lead (Pb) is a typical low-melting-point ductile metal and serves as an important model material in the study of dynamic responses. Under shock-wave loading, its dynamic mechanical behavior comprises two key phenomena: plastic deformation and shock induced phase transitions. The underlying mechanisms of these processes are still poorly understood. Revealing these mechanisms remains challenging for experimental approaches. Non-equilibrium molecular dynamics (NEMD) simulations are an alternative theoretical tool for studying dynamic responses, as they capture atomic-scale mechanisms such as defect evolution and deformation pathways. However, due to the limited accuracy of empirical interatomic potentials, the reliability of previous NEMD studies is questioned. Using our newly developed machine learning potential for Pb-Sn alloys, we revisited the microstructure evolution in response to shock loading under various shock orientations. The results reveal that shock loading along the [001] orientation of Pb exhibits a fast, reversible, and massive phase transition and stacking fault evolution. The behavior of Pb differs from previous studies by the absence of twinning during plastic deformation. Loading along the [011] orientation leads to slow, irreversible plastic deformation, and a localized FCC-BCC phase transition in the Pitsch orientation relationship. This study provides crucial theoretical insights into the dynamic mechanical response of Pb, offering a theoretical input for understanding the microstructure-performance relationship under extreme conditions.

Yaron Veksler, Sharon Hornstein, Han Wang et al.

The proliferation of connected automated vehicles represents an unprecedented opportunity for improving driving efficiency and alleviating traffic congestion. However, existing research fails to address realistic multi-lane highway scenarios without assuming connectivity, perception, and control capabilities that are typically unavailable in current vehicles. This paper proposes a novel AI system that is the first to improve highway traffic efficiency compared with human-like traffic in realistic, simulated multi-lane scenarios, while relying on existing connectivity, perception, and control capabilities. At the core of our approach is a reinforcement learning based controller that dynamically communicates time-headways to automated vehicles near bottlenecks based on real-time traffic conditions. These desired time-headways are then used by adaptive cruise control (ACC) systems to adjust their following distance. By (i) integrating existing traffic estimation technology and low-bandwidth vehicle-to-infrastructure connectivity, (ii) leveraging safety-certified ACC systems, and (iii) targeting localized bottleneck challenges that can be addressed independently in different locations, we propose a potentially practical, safe, and scalable system that can positively impact numerous road users.

Raphael Chekroun, Han Wang, Jonathan Lee et al. · berkeley

Accurate real-time traffic state forecasting plays a pivotal role in traffic control research. In particular, the CIRCLES consortium project necessitates predictive techniques to mitigate the impact of data source delays. After the success of the MegaVanderTest experiment, this paper aims at overcoming the current system limitations and develop a more suited approach to improve the real-time traffic state estimation for the next iterations of the experiment. In this paper, we introduce the SA-LSTM, a deep forecasting method integrating Self-Attention (SA) on the spatial dimension with Long Short-Term Memory (LSTM) yielding state-of-the-art results in real-time mesoscale traffic forecasting. We extend this approach to multi-step forecasting with the n-step SA-LSTM, which outperforms traditional multi-step forecasting methods in the trade-off between short-term and long-term predictions, all while operating in real-time.

Han Wang, Erfan Miahi, Martha White et al.

In this paper we investigate the properties of representations learned by deep reinforcement learning systems. Much of the early work on representations for reinforcement learning focused on designing fixed-basis architectures to achieve properties thought to be desirable, such as orthogonality and sparsity. In contrast, the idea behind deep reinforcement learning methods is that the agent designer should not encode representational properties, but rather that the data stream should determine the properties of the representation -- good representations emerge under appropriate training schemes. In this paper we bring these two perspectives together, empirically investigating the properties of representations that support transfer in reinforcement learning. We introduce and measure six representational properties over more than 25 thousand agent-task settings. We consider Deep Q-learning agents with different auxiliary losses in a pixel-based navigation environment, with source and transfer tasks corresponding to different goal locations. We develop a method to better understand why some representations work better for transfer, through a systematic approach varying task similarity and measuring and correlating representation properties with transfer performance. We demonstrate the generality of the methodology by investigating representations learned by a Rainbow agent that successfully transfer across games modes in Atari 2600.

Han Wang, James Anderson

This work considers the problem of learning the Markov parameters of a linear system from observed data. Recent non-asymptotic system identification results have characterized the sample complexity of this problem in the single and multi-rollout setting. In both instances, the number of samples required in order to obtain acceptable estimates can produce optimization problems with an intractably large number of decision variables for a second-order algorithm. We show that a randomized and distributed Newton algorithm based on Hessian-sketching can produce $ε$-optimal solutions and converges geometrically. Moreover, the algorithm is trivially parallelizable. Our results hold for a variety of sketching matrices and we illustrate the theory with numerical examples.

Han Wang

Users' website browsing history contains sensitive information, like health conditions, political interests, financial situations, etc. Some recent studies have demonstrated the possibility of inferring website fingerprints based on important usage information such as traffic, cache usage, memory usage, CPU activity, power consumption, and hardware performance counters information. However, existing website fingerprinting attacks demand a high sampling rate which causes high performance overheads and large network traffic, and/or they require launching an additional malicious website by the user, which is not guaranteed. As a result, such drawbacks make the existing attacks more noticeable to users and corresponding fingerprinting detection mechanisms. In response, in this work, we propose Leaked-Web, a novel accurate and efficient machine learning-based website fingerprinting attack through processor's Hardware Performance Counters (HPCs). Leaked-Web efficiently collects hardware performance counters in users' computer systems at a significantly low granularity monitoring rate and sends the samples to the remote attack's server for further classification. Leaked-Web examines the web browsers' microarchitectural features using various advanced machine learning algorithms ranging from classical, boosting, deep learning, and time-series models. Our experimental results indicate that Leaked-Web based on a LogitBoost ML classifier using only the top 4 HPC features achieves 91% classification accuracy outperforming the state-of-the-art attacks by nearly 5%. Furthermore, our proposed attack obtains a negligible performance overhead (only <1%), around 12% lower than the existing hardware-assisted website fingerprinting attacks.

Guiju Ping, Mahdi Abolfazli Esfahani, Han Wang

Solving the challenging problem of 3D object reconstruction from a single image appropriately gives existing technologies the ability to perform with a single monocular camera rather than requiring depth sensors. In recent years, thanks to the development of deep learning, 3D reconstruction of a single image has demonstrated impressive progress. Existing researches use Chamfer distance as a loss function to guide the training of the neural network. However, the Chamfer loss will give equal weights to all points inside the 3D point clouds. It tends to sacrifice fine-grained and thin structures to avoid incurring a high loss, which will lead to visually unsatisfactory results. This paper proposes a framework that can recover a detailed three-dimensional point cloud from a single image by focusing more on boundaries (edge and corner points). Experimental results demonstrate that the proposed method outperforms existing techniques significantly, both qualitatively and quantitatively, and has fewer training parameters.

Shangyu Xie, Han Wang, Yu Kong et al.

Widely deployed deep neural network (DNN) models have been proven to be vulnerable to adversarial perturbations in many applications (e.g., image, audio and text classifications). To date, there are only a few adversarial perturbations proposed to deviate the DNN models in video recognition systems by simply injecting 2D perturbations into video frames. However, such attacks may overly perturb the videos without learning the spatio-temporal features (across temporal frames), which are commonly extracted by DNN models for video recognition. To our best knowledge, we propose the first black-box attack framework that generates universal 3-dimensional (U3D) perturbations to subvert a variety of video recognition systems. U3D has many advantages, such as (1) as the transfer-based attack, U3D can universally attack multiple DNN models for video recognition without accessing to the target DNN model; (2) the high transferability of U3D makes such universal black-box attack easy-to-launch, which can be further enhanced by integrating queries over the target model when necessary; (3) U3D ensures human-imperceptibility; (4) U3D can bypass the existing state-of-the-art defense schemes; (5) U3D can be efficiently generated with a few pre-learned parameters, and then immediately injected to attack real-time DNN-based video recognition systems. We have conducted extensive experiments to evaluate U3D on multiple DNN models and three large-scale video datasets. The experimental results demonstrate its superiority and practicality.

Han Wang, Chen Wang, Chun-Lin Chen et al.

Simultaneous Localization and Mapping (SLAM) has wide robotic applications such as autonomous driving and unmanned aerial vehicles. Both computational efficiency and localization accuracy are of great importance towards a good SLAM system. Existing works on LiDAR based SLAM often formulate the problem as two modules: scan-to-scan match and scan-to-map refinement. Both modules are solved by iterative calculation which are computationally expensive. In this paper, we propose a general solution that aims to provide a computationally efficient and accurate framework for LiDAR based SLAM. Specifically, we adopt a non-iterative two-stage distortion compensation method to reduce the computational cost. For each scan input, the edge and planar features are extracted and matched to a local edge map and a local plane map separately, where the local smoothness is also considered for iterative pose optimization. Thorough experiments are performed to evaluate its performance in challenging scenarios, including localization for a warehouse Automated Guided Vehicle (AGV) and a public dataset on autonomous driving. The proposed method achieves a competitive localization accuracy with a processing rate of more than 10 Hz in the public dataset evaluation, which provides a good trade-off between performance and computational cost for practical applications.

Jinglin Sun, Zhipeng Wu, Han Wang et al.

Eye-tracking technologies have been widely used in applications like psychological studies and human computer interactions (HCI). However, most current eye trackers focus on 2D point of gaze (PoG) estimation and cannot provide accurate gaze depth.Concerning future applications such as HCI with 3D displays, we propose a novel binocular eye tracking device with stereo stimuli to provide highly accurate 3D PoG estimation. In our device, the 3D stereo imaging system can provide users with a friendly and immersive 3D visual experience without wearing any accessories. The eye capturing system can directly record the users eye movements under 3D stimuli without disturbance. A regression based 3D eye tracking model is built based on collected eye movement data under stereo stimuli. Our model estimates users 2D gaze with features defined by eye region landmarks and further estimates 3D PoG with a multi source feature set constructed by comprehensive eye movement features and disparity features from stereo stimuli. Two test stereo scenes with different depths of field are designed to verify the model effectiveness. Experimental results show that the average error for 2D gaze estimation was 0.66\degree and for 3D PoG estimation, the average errors are 1.85~cm/0.15~m over the workspace volume 50~cm $\times$ 30~cm $\times$ 75~cm/2.4~m $\times$ 4.0~m $\times$ 7.9~m separately.

Xuda Ding, Han Wang, Hongbo Li et al.

The mobile robotic testbed is an essential and critical support to verify the effectiveness of mobile robotics research. This paper introduces a novel multi-robot testbed, named Robopheus, which exploits the ideas of virtual-physical modeling in digital-twin. Unlike most existing testbeds, the developed Robopheus constructs a bridge that connects the traditional physical hardware and virtual simulation testbeds, providing scalable, interactive, and high-fidelity simulations-tests on both sides. Another salient feature of the Robopheus is that it enables a new form to learn the actual models from the physical environment dynamically and is compatible with heterogeneous robot chassis and controllers. In turn, the virtual world's learned models are further leveraged to approximate the robot dynamics online on the physical side. Extensive experiments demonstrate the extraordinary performance of the Robopheus. Significantly, the physical-virtual interaction design increases the trajectory accuracy of a real robot by 300%, compared with that of not using the interaction.

Wang Han, Wesley Yeung, Angeline Tung et al.

Clinical performance audits are routinely performed in Emergency Medical Services (EMS) to ensure adherence to treatment protocols, to identify individual areas of weakness for remediation, and to discover systemic deficiencies to guide the development of the training syllabus. At present, these audits are performed by manual chart review which is time-consuming and laborious. In this paper, we present an automatic audit system based on both the structured and unstructured ambulance case records and clinical notes with a deep neural network-based named entities recognition model. The dataset used in this study contained 58,898 unlabelled ambulance incidents encountered by the Singapore Civil Defence Force from 1st April 2019 to 30th June 2019. A weakly-supervised training approach was adopted to label the sentences. Later on, we trained three different models to perform the NER task. All three models achieve F1 scores of around 0.981 under entity type matching evaluation and around 0.976 under strict evaluation, while the BiLSTM-CRF model is 1~2 orders of magnitude lighter and faster than our BERT-based models. Overall, our approach yielded a named entity recognition model that could reliably identify clinical entities from unstructured paramedic free-text reports. Our proposed system may improve the efficiency of clinical performance audits and can also help with EMS database research.

Yifan Chen, Han Wang, Xiaolu Sun et al.

Visual attention has proven to be effective in improving the performance of person re-identification. Most existing methods apply visual attention heuristically by learning an additional attention map to re-weight the feature maps for person re-identification. However, this kind of methods inevitably increase the model complexity and inference time. In this paper, we propose to incorporate the attention learning as additional objectives in a person ReID network without changing the original structure, thus maintain the same inference time and model size. Two kinds of attentions have been considered to make the learned feature maps being aware of the person and related body parts respectively. Globally, a holistic attention branch (HAB) makes the feature maps obtained by backbone focus on persons so as to alleviate the influence of background. Locally, a partial attention branch (PAB) makes the extracted features be decoupled into several groups and be separately responsible for different body parts (i.e., keypoints), thus increasing the robustness to pose variation and partial occlusion. These two kinds of attentions are universal and can be incorporated into existing ReID networks. We have tested its performance on two typical networks (TriNet and Bag of Tricks) and observed significant performance improvement on five widely used datasets.

Han Wang, Muqing Cao, Hao Jiang et al.

This paper presents a robust computationally efficient real-time collision avoidance algorithm for Unmanned Aerial Vehicle (UAV), namely Memory-based Wall Following-Artificial Potential Field (MWF-APF) method. The new algorithm switches between Wall-Following Method (WFM) and Artificial Potential Field method (APF) with improved situation awareness capability. Historical trajectory is taken into account to avoid repetitive wrong decision. Furthermore, it can be effectively applied to platform with low computing capability. As an example, a quad-rotor equipped with limited number of Time-of-Flight (TOF) rangefinders is adopted to validate the effectiveness and efficiency of this algorithm. Both software simulation and physical flight test have been conducted to demonstrate the capability of the MWF-APF method in complex scenarios.

Han Wang, Juncheng Li, Maopeng Ran et al.

Loop closure detection plays an important role in reducing localization drift in Simultaneous Localization And Mapping (SLAM). It aims to find repetitive scenes from historical data to reset localization. To tackle the loop closure problem, existing methods often leverage on the matching of visual features, which achieve good accuracy but require high computational resources. However, feature point based methods ignore the patterns of image, i.e., the shape of the objects as well as the distribution of objects in an image. It is believed that this information is usually unique for a scene and can be utilized to improve the performance of traditional loop closure detection methods. In this paper we leverage and compress the information into a binary image to accelerate an existing fast loop closure detection method via binary content. The proposed method can greatly reduce the computational cost without sacrificing recall rate. It consists of three parts: binary content construction, fast image retrieval and precise loop closure detection. No offline training is required. Our method is compared with the state-of-the-art loop closure detection methods and the results show that it outperforms the traditional methods at both recall rate and speed.

Handuo Zhang, Karunasekera Hasith, Han Wang

Conventional SLAM algorithms takes a strong assumption of scene motionlessness, which limits the application in real environments. This paper tries to tackle the challenging visual SLAM issue of moving objects in dynamic environments. We present GMC, grid-based motion clustering approach, a lightweight dynamic object filtering method that is free from high-power and expensive processors. GMC encapsulates motion consistency as the statistical likelihood of detected key points within a certain region. Using this method can we provide real-time and robust correspondence algorithm that can differentiate dynamic objects with static backgrounds. We evaluate our system in public TUM dataset. To compare with the state-of-the-art methods, our system can provide more accurate results by detecting dynamic objects.

Yuan Hong, Han Wang, Shangyu Xie et al.

Energy has been increasingly generated or collected by different entities on the power grid (e.g., universities, hospitals and householdes) via solar panels, wind turbines or local generators in the past decade. With local energy, such electricity consumers can be considered as "microgrids" which can simulataneously generate and consume energy. Some microgrids may have excessive energy that can be shared to other power consumers on the grid. To this end, all the entities have to share their local private information (e.g., their local demand, local supply and power quality data) to each other or a third-party to find and implement the optimal energy sharing solution. However, such process is constrained by privacy concerns raised by the microgrids. In this paper, we propose a privacy preserving scheme for all the microgrids which can securely implement their energy sharing against both semi-honest and colluding adversaries. The proposed approach includes two secure communication protocols that can ensure quantified privacy leakage and handle collusions.

Habib Ammari, Han Wang

This paper presents premier and innovative time-domain multi-scale method for shape identification in electro-sensing using pulse-type signals. The method is based on transform-invariant shape descriptors computed from filtered polarization tensors at multi-scales. The proposed algorithm enjoys a remarkable noise robustness even with far-field measurements at very limited angle of view. It opens a door for pulsed imaging using echolocation and induction data.

Han Wang

We have initiated the study of topology of the space of coverings on grid domains. The space has the following constraint: while all the covering agents can move freely (we allow overlapping) on the domain, their union must cover the whole domain. A minimal number $N$ of the covering agents is required for a successful covering of the domain. In this paper, we demonstrate beautiful topological structures of this space on grid domains in 2D with $N+1$ coverings, the topology of the space has the homotopy type of $1$ dimensional complex, regardless of the domain shape. We also present the Euler characteristic formula which connects the topology of the space with that of the domain itself.