Searcharxiv⌕ Search

arXiv subjects

Tong Wei

Publications and source records attributed to Tong Wei.

At least 55 records · Page 3Linked to original sources
arXiv2024

Continuous Contrastive Learning for Long-Tailed Semi-Supervised Recognition

Long-tailed semi-supervised learning poses a significant challenge in training models with limited labeled data exhibiting a long-tailed label distribution. Current state-of-the-art LTSSL approaches heavily rely on high-quality pseudo-labels for large-scale unlabeled data. However, these methods often neglect the impact of representations learned by the neural network and struggle with real-world unlabeled data, which typically follows a different distribution than labeled data. This paper introduces a novel probabilistic framework that unifies various recent proposals in long-tail learning. Our framework derives the class-balanced contrastive loss through Gaussian kernel density estimation. We introduce a continuous contrastive learning method, CCL, extending our framework to unlabeled data using reliable and smoothed pseudo-labels. By progressively estimating the underlying label distribution and optimizing its alignment with model predictions, we tackle the diverse distribution of unlabeled data in real-world scenarios. Extensive experiments across multiple datasets with varying unlabeled data distributions demonstrate that CCL consistently outperforms prior state-of-the-art methods, achieving over 4% improvement on the ImageNet-127 dataset. Our source code is available at https://github.com/zhouzihao11/CCL

cs.LG↗
arXiv2024

Vision-Language Models are Strong Noisy Label Detectors

Recent research on fine-tuning vision-language models has demonstrated impressive performance in various downstream tasks. However, the challenge of obtaining accurately labeled data in real-world applications poses a significant obstacle during the fine-tuning process. To address this challenge, this paper presents a Denoising Fine-Tuning framework, called DeFT, for adapting vision-language models. DeFT utilizes the robust alignment of textual and visual features pre-trained on millions of auxiliary image-text pairs to sieve out noisy labels. The proposed framework establishes a noisy label detector by learning positive and negative textual prompts for each class. The positive prompt seeks to reveal distinctive features of the class, while the negative prompt serves as a learnable threshold for separating clean and noisy samples. We employ parameter-efficient fine-tuning for the adaptation of a pre-trained visual encoder to promote its alignment with the learned textual prompts. As a general framework, DeFT can seamlessly fine-tune many pre-trained models to downstream tasks by utilizing carefully selected clean samples. Experimental results on seven synthetic and real-world noisy datasets validate the effectiveness of DeFT in both noisy label detection and image classification.

cs.LG↗
arXiv2024

Efficient and Long-Tailed Generalization for Pre-trained Vision-Language Model

Pre-trained vision-language models like CLIP have shown powerful zero-shot inference ability via image-text matching and prove to be strong few-shot learners in various downstream tasks. However, in real-world scenarios, adapting CLIP to downstream tasks may encounter the following challenges: 1) data may exhibit long-tailed data distributions and might not have abundant samples for all the classes; 2) There might be emerging tasks with new classes that contain no samples at all. To overcome them, we propose a novel framework to achieve efficient and long-tailed generalization, which can be termed as Candle. During the training process, we propose compensating logit-adjusted loss to encourage large margins of prototypes and alleviate imbalance both within the base classes and between the base and new classes. For efficient adaptation, we treat the CLIP model as a black box and leverage the extracted features to obtain visual and textual prototypes for prediction. To make full use of multi-modal information, we also propose cross-modal attention to enrich the features from both modalities. For effective generalization, we introduce virtual prototypes for new classes to make up for their lack of training images. Candle achieves state-of-the-art performance over extensive experiments on 11 diverse datasets while substantially reducing the training time, demonstrating the superiority of our approach. The source code is available at https://github.com/shijxcs/Candle.

cs.CV↗
arXiv2024

Long-Tail Learning with Foundation Model: Heavy Fine-Tuning Hurts

The fine-tuning paradigm in addressing long-tail learning tasks has sparked significant interest since the emergence of foundation models. Nonetheless, how fine-tuning impacts performance in long-tail learning was not explicitly quantified. In this paper, we disclose that heavy fine-tuning may even lead to non-negligible performance deterioration on tail classes, and lightweight fine-tuning is more effective. The reason is attributed to inconsistent class conditions caused by heavy fine-tuning. With the observation above, we develop a low-complexity and accurate long-tail learning algorithms LIFT with the goal of facilitating fast prediction and compact models by adaptive lightweight fine-tuning. Experiments clearly verify that both the training time and the learned parameters are significantly reduced with more accurate predictive performance compared with state-of-the-art approaches. The implementation code is available at https://github.com/shijxcs/LIFT.

cs.CV↗
arXiv2024

Erasing the Bias: Fine-Tuning Foundation Models for Semi-Supervised Learning

Semi-supervised learning (SSL) has witnessed remarkable progress, resulting in the emergence of numerous method variations. However, practitioners often encounter challenges when attempting to deploy these methods due to their subpar performance. In this paper, we present a novel SSL approach named FineSSL that significantly addresses this limitation by adapting pre-trained foundation models. We identify the aggregated biases and cognitive deviation problems inherent in foundation models, and propose a simple yet effective solution by imposing balanced margin softmax and decoupled label smoothing. Through extensive experiments, we demonstrate that FineSSL sets a new state of the art for SSL on multiple benchmark datasets, reduces the training cost by over six times, and can seamlessly integrate various fine-tuning and modern SSL algorithms. The source code is available at https://github.com/Gank0078/FineSSL.

cs.LG↗
arXiv2024

Fire in SRRN: Next-Gen 3D Temperature Field Reconstruction Technology

In aerospace and energy engineering, accurate 3D combustion field temperature measurement is critical. The resolution of traditional methods based on algebraic iteration is limited by the initial voxel division. This study introduces a novel method for reconstructing three-dimensional temperature fields using the Spatial Radiation Representation Network (SRRN). This method utilizes the flame thermal radiation characteristics and differentiable rendering in graphics, and combines it with a multi-layer perceptron to achieve a functional representation of the flame temperature field. The effectiveness of SRRN is evaluated through simulated temperature field reconstruction experiments with different levels of complexity. The maximum root mean square error is 10.17, which proves the robustness of the algorithm to Gaussian noise and salt-and-pepper noise. We conducted a butane flame temperature field reconstruction experiment, and the maximum relative error between the reconstruction result and the thermocouple measurement value was 4.86%, confirming that the algorithm can achieve accurate reconstruction.

eess.IV↗
arXiv2024

QR Decomposition of Dual Matrices and its Application to Traveling Wave Identification in the Brain

Matrix decompositions in dual number representations have played an important role in fields such as kinematics and computer graphics in recent years. In this paper, we present a QR decomposition algorithm for dual number matrices, specifically geared towards its application in traveling wave identification, utilizing the concept of proper orthogonal decomposition. When dealing with large-scale problems, we present explicit solutions for the QR, thin QR, and randomized QR decompositions of dual number matrices, along with their respective algorithms with column pivoting. The QR decomposition of dual matrices is an accurate first-order perturbation, with the Q-factor satisfying rigorous perturbation bounds, leading to enhanced orthogonality. In numerical experiments, we discuss the suitability of different QR algorithms when confronted with various large-scale dual matrices, providing their respective domains of applicability. Subsequently, we employed the QR decomposition of dual matrices to compute the DMPGI, thereby attaining results of higher precision. Moreover, we apply the QR decomposition in the context of traveling wave identification, employing the notion of proper orthogonal decomposition to perform a validation analysis of large-scale functional magnetic resonance imaging (fMRI) data for brain functional circuits. Our approach significantly improves the identification of two types of wave signals compared to previous research, providing empirical evidence for cognitive neuroscience theories.

math.NA↗
arXiv2024

Bridging the Gap: Learning Pace Synchronization for Open-World Semi-Supervised Learning

In open-world semi-supervised learning, a machine learning model is tasked with uncovering novel categories from unlabeled data while maintaining performance on seen categories from labeled data. The central challenge is the substantial learning gap between seen and novel categories, as the model learns the former faster due to accurate supervisory information. Moreover, capturing the semantics of unlabeled novel category samples is also challenging due to the missing label information. To address the above issues, we introduce 1) the adaptive synchronizing marginal loss which imposes class-specific negative margins to alleviate the model bias towards seen classes, and 2) the pseudo-label contrastive clustering which exploits pseudo-labels predicted by the model to group unlabeled data from the same category together in the output space. Extensive experiments on benchmark datasets demonstrate that previous approaches may significantly hinder novel class learning, whereas our method strikingly balances the learning pace between seen and novel classes, achieving a remarkable 3% average accuracy increase on the ImageNet dataset. Importantly, we find that fine-tuning the self-supervised pre-trained model significantly boosts the performance, which is overlooked in prior literature. Our code is available at https://github.com/yebo0216best/LPS-main.

cs.LG↗
arXiv2023

EAT: Towards Long-Tailed Out-of-Distribution Detection

Despite recent advancements in out-of-distribution (OOD) detection, most current studies assume a class-balanced in-distribution training dataset, which is rarely the case in real-world scenarios. This paper addresses the challenging task of long-tailed OOD detection, where the in-distribution data follows a long-tailed class distribution. The main difficulty lies in distinguishing OOD data from samples belonging to the tail classes, as the ability of a classifier to detect OOD instances is not strongly correlated with its accuracy on the in-distribution classes. To overcome this issue, we propose two simple ideas: (1) Expanding the in-distribution class space by introducing multiple abstention classes. This approach allows us to build a detector with clear decision boundaries by training on OOD data using virtual labels. (2) Augmenting the context-limited tail classes by overlaying images onto the context-rich OOD data. This technique encourages the model to pay more attention to the discriminative features of the tail classes. We provide a clue for separating in-distribution and OOD data by analyzing gradient noise. Through extensive experiments, we demonstrate that our method outperforms the current state-of-the-art on various benchmark datasets. Moreover, our method can be used as an add-on for existing long-tail learning approaches, significantly enhancing their OOD detection performance. Code is available at: https://github.com/Stomach-ache/Long-Tailed-OOD-Detection .

cs.LG↗
arXiv2023

How Re-sampling Helps for Long-Tail Learning?

Long-tail learning has received significant attention in recent years due to the challenge it poses with extremely imbalanced datasets. In these datasets, only a few classes (known as the head classes) have an adequate number of training samples, while the rest of the classes (known as the tail classes) are infrequent in the training data. Re-sampling is a classical and widely used approach for addressing class imbalance issues. Unfortunately, recent studies claim that re-sampling brings negligible performance improvements in modern long-tail learning tasks. This paper aims to investigate this phenomenon systematically. Our research shows that re-sampling can considerably improve generalization when the training images do not contain semantically irrelevant contexts. In other scenarios, however, it can learn unexpected spurious correlations between irrelevant contexts and target labels. We design experiments on two homogeneous datasets, one containing irrelevant context and the other not, to confirm our findings. To prevent the learning of spurious correlations, we propose a new context shift augmentation module that generates diverse training images for the tail class by maintaining a context bank extracted from the head-class images. Experiments demonstrate that our proposed module can boost the generalization and outperform other approaches, including class-balanced re-sampling, decoupled classifier re-training, and data augmentation methods. The source code is available at https://www.lamda.nju.edu.cn/code_CSA.ashx.

cs.CV↗
arXiv2023

Adaptive Reordering Sampler with Neurally Guided MAGSAC

We propose a new sampler for robust estimators that always selects the sample with the highest probability of consisting only of inliers. After every unsuccessful iteration, the inlier probabilities are updated in a principled way via a Bayesian approach. The probabilities obtained by the deep network are used as prior (so-called neural guidance) inside the sampler. Moreover, we introduce a new loss that exploits, in a geometrically justifiable manner, the orientation and scale that can be estimated for any type of feature, e.g., SIFT or SuperPoint, to estimate two-view geometry. The new loss helps to learn higher-order information about the underlying scene geometry. Benefiting from the new sampler and the proposed loss, we combine the neural guidance with the state-of-the-art MAGSAC++. Adaptive Reordering Sampler with Neurally Guided MAGSAC (ARS-MAGSAC) is superior to the state-of-the-art in terms of accuracy and run-time on the PhotoTourism and KITTI datasets for essential and fundamental matrix estimation. The code and trained models are available at https://github.com/weitong8591/ars_magsac.

cs.CV↗
arXiv2023

Generalized Differentiable RANSAC

We propose $\nabla$-RANSAC, a generalized differentiable RANSAC that allows learning the entire randomized robust estimation pipeline. The proposed approach enables the use of relaxation techniques for estimating the gradients in the sampling distribution, which are then propagated through a differentiable solver. The trainable quality function marginalizes over the scores from all the models estimated within $\nabla$-RANSAC to guide the network learning accurate and useful inlier probabilities or to train feature detection and matching networks. Our method directly maximizes the probability of drawing a good hypothesis, allowing us to learn better sampling distributions. We test $\nabla$-RANSAC on various real-world scenarios on fundamental and essential matrix estimation, and 3D point cloud registration, outdoors and indoors, with handcrafted and learning-based features. It is superior to the state-of-the-art in terms of accuracy while running at a similar speed to its less accurate alternatives. The code and trained models are available at https://github.com/weitong8591/differentiable_ransac.

cs.CV↗
arXiv2023

Scene-Aware Feature Matching

Current feature matching methods focus on point-level matching, pursuing better representation learning of individual features, but lacking further understanding of the scene. This results in significant performance degradation when handling challenging scenes such as scenes with large viewpoint and illumination changes. To tackle this problem, we propose a novel model named SAM, which applies attentional grouping to guide Scene-Aware feature Matching. SAM handles multi-level features, i.e., image tokens and group tokens, with attention layers, and groups the image tokens with the proposed token grouping module. Our model can be trained by ground-truth matches only and produce reasonable grouping results. With the sense-aware grouping guidance, SAM is not only more accurate and robust but also more interpretable than conventional feature matching models. Sufficient experiments on various applications, including homography estimation, pose estimation, and image matching, demonstrate that our model achieves state-of-the-art performance.

cs.CV↗
arXiv2023

Singular Value Decomposition of Dual Matrices and its Application to Traveling Wave Identification in the Brain

Matrix factorizations in dual number algebra, a hypercomplex system, have been applied to kinematics, mechanisms, and other fields recently. We develop an approach to identify spatiotemporal patterns in the brain such as traveling waves using the singular value decomposition of dual matrices in this paper. Theoretically, we propose the compact dual singular value decomposition (CDSVD) of dual complex matrices with explicit expressions as well as a necessary and sufficient condition for its existence. Furthermore, based on the CDSVD, we report on the optimal solution to the best rank-$k$ approximation under a newly defined quasi-metric in the dual complex number system. The CDSVD is also related to the dual Moore-Penrose generalized inverse. Numerically, comparisons with other available algorithms are conducted, which indicate less computational costs of our proposed CDSVD. In addition, the infinitesimal part of the CDSVD can identify the true rank of the original matrix from the noise-added matrix, but the classical SVD cannot. Next, we employ experiments on simulated time-series data and a road monitoring video to demonstrate the beneficial effect of the infinitesimal parts of dual matrices in spatiotemporal pattern identification. Finally, we apply this approach to the large-scale brain fMRI data, identify three kinds of traveling waves, and further validate the consistency between our analytical results and the current knowledge of cerebral cortex function.

math.NA↗
arXiv2023

Multi-IRS-Aided Doppler-Tolerant Wideband DFRC System

Intelligent reflecting surface (IRS) is recognized as an enabler of future dual-function radar-communications (DFRC) by improving spectral efficiency, coverage, parameter estimation, and interference suppression. Prior studies on IRS-aided DFRC focus either on narrowband processing, single-IRS deployment, static targets, non-clutter scenario, or on the under-utilized line-of-sight (LoS) and non-line-of-sight (NLoS) paths. In this paper, we address the aforementioned shortcomings by optimizing a wideband DFRC system comprising multiple IRSs and a dual-function base station that jointly processes the LoS and NLoS wideband multi-carrier signals to improve both the communications SINR and the radar SINR in the presence of a moving target and clutter. We formulate the transmit, {receive} and IRS beamformer design as the maximization of the worst-case radar signal-to-interference-plus-noise ratio (SINR) subject to transmit power and communications SINR. We tackle this nonconvex problem under the alternating optimization framework, where the subproblems are solved by a combination of Dinkelbach algorithm, consensus alternating direction method of multipliers, and Riemannian steepest decent. Our numerical experiments show that the proposed multi-IRS-aided wideband DFRC provides over $4$ dB radar SINR and $31.7$\% improvement in target detection over a single-IRS system.

cs.IT↗
arXiv2023

RIS-Aided Wideband Holographic DFRC

To enable non-line-of-sight (NLoS) sensing and communications, dual-function radar-communications (DFRC) systems have recently proposed employing reconfigurable intelligent surface (RIS) as a reflector in wireless media. However, in the dense environment and higher frequencies, severe propagation and attenuation losses are a hindrance for RIS-aided DFRC systems to utilize wideband processing. To this end, we propose equipping the transceivers with the reconfigurable holographic surface (RHS) that, different from RIS, is a metasurface with an embedded connected feed deployed at the transceiver for greater control of the radiation amplitude. This surface is crucial for designing compact low-cost wideband wireless systems, wherein ultra-massive antenna arrays are required to compensate for the losses incurred by severe attenuation and diffraction. We consider a novel wideband DFRC system equipped with an RHS at the transceiver and a RIS reflector in the channel. We jointly design the digital, holographic, and passive beamformers to maximize the radar signal-to-interference-plus-noise ratio (SINR) while ensuring the communications SINR among all users. The resulting nonconvex optimization problem involves maximin objective, constant modulus, and difference of convex constraints. We develop an alternating maximization method to decouple and iteratively solve these subproblems. Numerical experiments demonstrate that the proposed method achieves better radar performance than non-RIS, random-RHS, and randomly configured RIS-aided DFRC systems.

eess.SP↗
arXiv2022

A Survey on Extreme Multi-label Learning

Multi-label learning has attracted significant attention from both academic and industry field in recent decades. Although existing multi-label learning algorithms achieved good performance in various tasks, they implicitly assume the size of target label space is not huge, which can be restrictive for real-world scenarios. Moreover, it is infeasible to directly adapt them to extremely large label space because of the compute and memory overhead. Therefore, eXtreme Multi-label Learning (XML) is becoming an important task and many effective approaches are proposed. To fully understand XML, we conduct a survey study in this paper. We first clarify a formal definition for XML from the perspective of supervised learning. Then, based on different model architectures and challenges of the problem, we provide a thorough discussion of the advantages and disadvantages of each category of methods. For the benefit of conducting empirical studies, we collect abundant resources regarding XML, including code implementations, and useful tools. Lastly, we propose possible research directions in XML, such as new evaluation metrics, the tail label problem, and weakly supervised XML.

cs.LG↗
arXiv2022

Transfer and Share: Semi-Supervised Learning from Long-Tailed Data

Long-Tailed Semi-Supervised Learning (LTSSL) aims to learn from class-imbalanced data where only a few samples are annotated. Existing solutions typically require substantial cost to solve complex optimization problems, or class-balanced undersampling which can result in information loss. In this paper, we present the TRAS (TRAnsfer and Share) to effectively utilize long-tailed semi-supervised data. TRAS transforms the imbalanced pseudo-label distribution of a traditional SSL model via a delicate function to enhance the supervisory signals for minority classes. It then transfers the distribution to a target model such that the minority class will receive significant attention. Interestingly, TRAS shows that more balanced pseudo-label distribution can substantially benefit minority-class training, instead of seeking to generate accurate pseudo-labels as in previous works. To simplify the approach, TRAS merges the training of the traditional SSL model and the target model into a single procedure by sharing the feature extractor, where both classifiers help improve the representation learning. According to extensive experiments, TRAS delivers much higher accuracy than state-of-the-art methods in the entire set of classes as well as minority classes.

cs.LG↗