arXiv subjects

Jingyao Wang

Publications and source records attributed to Jingyao Wang.

51 records · Page 3Linked to original sources

arXiv2024

Hacking Task Confounder in Meta-Learning

Meta-learning enables rapid generalization to new tasks by learning knowledge from various tasks. It is intuitively assumed that as the training progresses, a model will acquire richer knowledge, leading to better generalization performance. However, our experiments reveal an unexpected result: there is negative knowledge transfer between tasks, affecting generalization performance. To explain this phenomenon, we conduct Structural Causal Models (SCMs) for causal analysis. Our investigation uncovers the presence of spurious correlations between task-specific causal factors and labels in meta-learning. Furthermore, the confounding factors differ across different batches. We refer to these confounding factors as "Task Confounders". Based on these findings, we propose a plug-and-play Meta-learning Causal Representation Learner (MetaCRL) to eliminate task confounders. It encodes decoupled generating factors from multiple tasks and utilizes an invariant-based bi-level optimization mechanism to ensure their causality for meta-learning. Extensive experiments on various benchmark datasets demonstrate that our work achieves state-of-the-art (SOTA) performance.

cs.LG↗

arXiv2024

Meta-Auxiliary Learning for Micro-Expression Recognition

Micro-expressions (MEs) are involuntary movements revealing people's hidden feelings, which has attracted numerous interests for its objectivity in emotion detection. However, despite its wide applications in various scenarios, micro-expression recognition (MER) remains a challenging problem in real life due to three reasons, including (i) data-level: lack of data and imbalanced classes, (ii) feature-level: subtle, rapid changing, and complex features of MEs, and (iii) decision-making-level: impact of individual differences. To address these issues, we propose a dual-branch meta-auxiliary learning method, called LightmanNet, for fast and robust micro-expression recognition. Specifically, LightmanNet learns general MER knowledge from limited data through a dual-branch bi-level optimization process: (i) In the first level, it obtains task-specific MER knowledge by learning in two branches, where the first branch is for learning MER features via primary MER tasks, while the other branch is for guiding the model obtain discriminative features via auxiliary tasks, i.e., image alignment between micro-expressions and macro-expressions since their resemblance in both spatial and temporal behavioral patterns. The two branches of learning jointly constrain the model of learning meaningful task-specific MER knowledge while avoiding learning noise or superficial connections between MEs and emotions that may damage its generalization ability. (ii) In the second level, LightmanNet further refines the learned task-specific knowledge, improving model generalization and efficiency. Extensive experiments on various benchmark datasets demonstrate the superior robustness and efficiency of LightmanNet.

cs.CV↗

arXiv2023

Unleash Model Potential: Bootstrapped Meta Self-supervised Learning

The long-term goal of machine learning is to learn general visual representations from a small amount of data without supervision, mimicking three advantages of human cognition: i) no need for labels, ii) robustness to data scarcity, and iii) learning from experience. Self-supervised learning and meta-learning are two promising techniques to achieve this goal, but they both only partially capture the advantages and fail to address all the problems. Self-supervised learning struggles to overcome the drawbacks of data scarcity, while ignoring prior knowledge that can facilitate learning and generalization. Meta-learning relies on supervised information and suffers from a bottleneck of insufficient learning. To address these issues, we propose a novel Bootstrapped Meta Self-Supervised Learning (BMSSL) framework that aims to simulate the human learning process. We first analyze the close relationship between meta-learning and self-supervised learning. Based on this insight, we reconstruct tasks to leverage the strengths of both paradigms, achieving advantages i and ii. Moreover, we employ a bi-level optimization framework that alternates between solving specific tasks with a learned ability (first level) and improving this ability (second level), attaining advantage iii. To fully harness its power, we introduce a bootstrapped target based on meta-gradient to make the model its own teacher. We validate the effectiveness of our approach with comprehensive theoretical and empirical study.

cs.LG↗

arXiv2023

Text-based Person Search without Parallel Image-Text Data

Text-based person search (TBPS) aims to retrieve the images of the target person from a large image gallery based on a given natural language description. Existing methods are dominated by training models with parallel image-text pairs, which are very costly to collect. In this paper, we make the first attempt to explore TBPS without parallel image-text data ($μ$-TBPS), in which only non-parallel images and texts, or even image-only data, can be adopted. Towards this end, we propose a two-stage framework, generation-then-retrieval (GTR), to first generate the corresponding pseudo text for each image and then perform the retrieval in a supervised manner. In the generation stage, we propose a fine-grained image captioning strategy to obtain an enriched description of the person image, which firstly utilizes a set of instruction prompts to activate the off-the-shelf pretrained vision-language model to capture and generate fine-grained person attributes, and then converts the extracted attributes into a textual description via the finetuned large language model or the hand-crafted template. In the retrieval stage, considering the noise interference of the generated texts for training model, we develop a confidence score-based training scheme by enabling more reliable texts to contribute more during the training. Experimental results on multiple TBPS benchmarks (i.e., CUHK-PEDES, ICFG-PEDES and RSTPReid) show that the proposed GTR can achieve a promising performance without relying on parallel image-text data.

cs.CV↗

arXiv2023

Towards the Sparseness of Projection Head in Self-Supervised Learning

In recent years, self-supervised learning (SSL) has emerged as a promising approach for extracting valuable representations from unlabeled data. One successful SSL method is contrastive learning, which aims to bring positive examples closer while pushing negative examples apart. Many current contrastive learning approaches utilize a parameterized projection head. Through a combination of empirical analysis and theoretical investigation, we provide insights into the internal mechanisms of the projection head and its relationship with the phenomenon of dimensional collapse. Our findings demonstrate that the projection head enhances the quality of representations by performing contrastive loss in a projected subspace. Therefore, we propose an assumption that only a subset of features is necessary when minimizing the contrastive loss of a mini-batch of data. Theoretical analysis further suggests that a sparse projection head can enhance generalization, leading us to introduce SparseHead - a regularization term that effectively constrains the sparsity of the projection head, and can be seamlessly integrated with any self-supervised learning (SSL) approaches. Our experimental results validate the effectiveness of SparseHead, demonstrating its ability to improve the performance of existing contrastive methods.

cs.LG↗

arXiv2023

CSSL-RHA: Contrastive Self-Supervised Learning for Robust Handwriting Authentication

Handwriting authentication is a valuable tool used in various fields, such as fraud prevention and cultural heritage protection. However, it remains a challenging task due to the complex features, severe damage, and lack of supervision. In this paper, we propose a novel Contrastive Self-Supervised Learning framework for Robust Handwriting Authentication (CSSL-RHA) to address these issues. It can dynamically learn complex yet important features and accurately predict writer identities. Specifically, to remove the negative effects of imperfections and redundancy, we design an information-theoretic filter for pre-processing and propose a novel adaptive matching scheme to represent images as patches of local regions dominated by more important features. Through online optimization at inference time, the most informative patch embeddings are identified as the "most important" elements. Furthermore, we employ contrastive self-supervised training with a momentum-based paradigm to learn more general statistical structures of handwritten data without supervision. We conduct extensive experiments on five benchmark datasets and our manually annotated dataset EN-HA, which demonstrate the superiority of our CSSL-RHA compared to baselines. Additionally, we show that our proposed model can still effectively achieve authentication even under abnormal circumstances, such as data falsification and corruption.

cs.CV↗

arXiv2023

Efficient Image-Text Retrieval via Keyword-Guided Pre-Screening

Under the flourishing development in performance, current image-text retrieval methods suffer from $N$-related time complexity, which hinders their application in practice. Targeting at efficiency improvement, this paper presents a simple and effective keyword-guided pre-screening framework for the image-text retrieval. Specifically, we convert the image and text data into the keywords and perform the keyword matching across modalities to exclude a large number of irrelevant gallery samples prior to the retrieval network. For the keyword prediction, we transfer it into a multi-label classification problem and propose a multi-task learning scheme by appending the multi-label classifiers to the image-text retrieval network to achieve a lightweight and high-performance keyword prediction. For the keyword matching, we introduce the inverted index in the search engine and create a win-win situation on both time and space complexities for the pre-screening. Extensive experiments on two widely-used datasets, i.e., Flickr30K and MS-COCO, verify the effectiveness of the proposed framework. The proposed framework equipped with only two embedding layers achieves $O(1)$ querying time complexity, while improving the retrieval efficiency and keeping its performance, when applied prior to the common image-text retrieval methods. Our code will be released.

cs.CV↗

arXiv2022

Spin triplet superconducting pairing in doped MoS$_2$

Searching for triplet superconductivity has been pursued intensively in a broad field of material science and quantum information for decades. Nevertheless, these novel states remain rare. Within a simplified effective three-orbital model, we reveal a spin triplet pairing in doped MoS$_2$ by employing both the finite temperature determinant quantum Monte Carlo approach and the ground state constrained-phase quantum Monte Carlo method. In a wide filling region of $\avg{n}=0.60-0.80$ around charge neutrality $\avg{n}=2/3$, the $f$-wave pairing dominates over other symmetries. The pairing susceptibility strongly increases as the on-site Coulomb interaction increases, and it is insensitive to spin-orbit coupling.

cond-mat.supr-con↗

arXiv2022

SSD-Faster Net: A Hybrid Network for Industrial Defect Inspection

The quality of industrial components is critical to the production of special equipment such as robots. Defect inspection of these components is an efficient way to ensure quality. In this paper, we propose a hybrid network, SSD-Faster Net, for industrial defect inspection of rails, insulators, commutators etc. SSD-Faster Net is a two-stage network, including SSD for quickly locating defective blocks, and an improved Faster R-CNN for defect segmentation. For the former, we propose a novel slice localization mechanism to help SSD scan quickly. The second stage is based on improved Faster R-CNN, using FPN, deformable kernel(DK) to enhance representation ability. It fuses multi-scale information, and self-adapts the receptive field. We also propose a novel loss function and use ROI Align to improve accuracy. Experiments show that our SSD-Faster Net achieves an average accuracy of 84.03%, which is 13.42% higher than the nearest competitor based on Faster R-CNN, 4.14% better than GAN-based methods, more than 10% higher than that of DNN-based detectors. And the computing speed is improved by nearly 7%, which proves its robustness and superior performance.

cs.CV↗

arXiv2022

UTD-Yolov5: A Real-time Underwater Targets Detection Method based on Attention Improved YOLOv5

As the treasure house of nature, the ocean contains abundant resources. But the coral reefs, which are crucial to the sustainable development of marine life, are facing a huge crisis because of the existence of COTS and other organisms. The protection of society through manual labor is limited and inefficient. The unpredictable nature of the marine environment also makes manual operations risky. The use of robots for underwater operations has become a trend. However, the underwater image acquisition has defects such as weak light, low resolution, and many interferences, while the existing target detection algorithms are not effective. Based on this, we propose an underwater target detection algorithm based on Attention Improved YOLOv5, called UTD-Yolov5. It can quickly and efficiently detect COTS, which in turn provides a prerequisite for complex underwater operations. We adjusted the original network architecture of YOLOv5 in multiple stages, including: replacing the original Backbone with a two-stage cascaded CSP (CSP2); introducing the visual channel attention mechanism module SE; designing random anchor box similarity calculation method etc. These operations enable UTD-Yolov5 to detect more flexibly and capture features more accurately. In order to make the network more efficient, we also propose optimization methods such as WBF and iterative refinement mechanism. This paper conducts a lot of experiments based on the CSIRO dataset [1]. The results show that the average accuracy of our UTD-Yolov5 reaches 78.54%, which is a great improvement compared to the baseline.

cs.CV↗

arXiv2020

Intermediate Phase in Interacting Dirac Fermions with Staggered Potential

By performing exact quantum Monte Carlo simulations of a model of interacting Dirac Fermions with staggered potential, we reveal a novel intermediate phase where the electronic correlations drive a band insulator metallic, and at a larger interaction, drive the metal to Mott insulator. We also show that the Mott insulating phase is antiferromagnetic. A complete phase diagram is achieved by studying the phase transitions at large staggered potential and interaction strengths, which shows that the intermediate state is robust and occupies a large part of the phase diagram and that it should be more feasible to be detected experimentally.

cond-mat.str-el↗

arXiv2018

An Output-Feedback Control Approach to the $H_{\infty}$ Consensus Integrated with Transient Performance Improvement Problem

This paper considers the consensus performance improvement problem of networked general linear agents subject to external disturbances over Markovian randomly switching communication topologies. The consensus control laws can only use its local output information. Firstly, a class of full-order observer-based control protocols is proposed to solve this problem, which depends solely on the relative outputs of neighbours. Then, to eliminate the redundancy involved in the full-order observer, a class of reduced-order observer-based control protocols is designed. Algorithms to construct both protocols are presented, which guarantee that agents can reach consensus in the asymptotic mean square sense when they are not perturbed by disturbances, and that they have decent $H_{\infty}$ performance and transient performance when the disturbances exist. At the end of this manuscript, numerical simulations which apply both algorithms to four networked Raptor-90 helicopters are performed to verify the theoretical results.

math.OC↗

arXiv2018

Linear Quadratic Synchronization of Multi-Agent Systems: A Distributed Optimization Approach

The distributed optimal synchronization problem with linear quadratic cost is solved in this paper for multi-agent systems with an undirected communication topology. For the first time, the optimal synchronization problem is formulated as a distributed optimization problem with a linear quadratic cost functional that integrates quadratic synchronization errors and quadratic input signals subject to agent dynamics and synchronization constraints. By introducing auxiliary synchronization state variables and combining the distributed synchronization method with the alternating direction method of multiplier (ADMM), a new distributed control protocol is designed for solving the distributed optimization problem. With this construction, the optimal synchronization control problem is separated into several independent subproblems: a synchronization optimization, an input minimization and a dual optimization. These subproblems are then solved by distributed numerical algorithms based on the Lyapunov method and dynamic programming. Numerical examples for both homogeneous and heterogeneous multi-agent systems are given to demonstrate the effectiveness of the proposed method.

eess.SY↗

arXiv2017

Non-Concave Network Utility Maximization in Connectionless Networks: A Fully Distributed Traffic Allocation Algorithm

This paper considers the optimization-based traffic allocation problem among multiple end points in connectionless networks. The network utility function is modeled as a non-concave function, since it is the best description of the quality of service perceived by users with inelastic applications, such as video and audio streaming. However, the resulting non-convex optimization problem, is challenging and requires new analysis and solution techniques. To overcome these challenges, we first propose a hierarchy of problems whose optimal value converges to the optimal value of the non-convex optimization problem as the number of moments tends to infinity. From this hierarchy of problems, we obtain a convex relaxation of the original non-convex optimization problem by considering truncated moment sequences. For solving the convex relaxation, we propose a fully distributed iterative algorithm, which enables each node to adjust its date allocation/ rate adaption among any given set of next hops solely based on information from the neighboring nodes. Moreover, the proposed traffic allocation algorithm converges to the optimal value of the convex relaxation at a $O(1/K)$ rate, where $K$ is the iteration counter, with a bounded optimality. At the end of this paper, we perform numerical simulations to demonstrate the soundness of the developed algorithm.

math.OC↗

arXiv2016

Character of frustration on magnetic correlation in doped Hubbard model

The magnetic correlation in the Hubbard model on a two-dimensional anisotropic triangular lattice is studied by using the determinant quantum Monte Carlo method. Around half filling, it is found that the increasing frustration $t'/t$ could change the wave vector of maximum spin correlation along ($π,π$)$\rightarrow$($π,\frac{5π}{6}$)$\rightarrow$($\frac{5π}{6},\frac{5π}{6}$)$\rightarrow$ ($\frac{2π}{3},\frac{2π}{3}$), indicating the frustration's remarkable effect on the magnetism. In the studied filling region =1.0-1.3, the doping behaves like some kinds of {\it{frustration}}, which destroys the $(π,π)$ AFM correlation quickly and push the magnetic order to a wide range of the $(\frac{2π}{3},\frac{2π}{3})$ $120^{\circ}$ order when the $t'/t$ is large enough. Our non-perturbative calculations reveal a rich magnetic phase diagram over both the frustration and electron doping.

cond-mat.str-el↗