Author name cluster

Jun Fang

Possible papers associated with this exact author name in Arrow. This page groups case-insensitive exact name matches and is not a full identity disambiguation profile.

6 papers

2 author rows

AAAI Conference 2026 Conference Paper

FANoise: Singular Value-Adaptive Noise Modulation for Robust Multimodal Representation Learning

Jiaoyang Li
Jun Fang
Tianhao Gao
Xiaohui Zhang
Zhiyuan Liu
Chao Liu
Pengzhang Liu
Qixia Jiang

Representation learning is fundamental to modern machine learning, powering applications such as text retrieval and multimodal understanding. However, learning robust and generalizable representations remains challenging. While prior work has demonstrated that active noise injection, a form of data augmentation, can enhance encoding performance, most existing methods rely on heuristic or static noise, overlooking the dynamic nature of feature distributions during training. In this work, we systematically study the role of noise in representation learning from both gradient-based and feature distribution perspectives, using InfoNCE loss as a representative example. Focusing on multimodal representation learning, we propose FANoise, a novel feature-adaptive noise injection strategy. By leveraging the dynamics of contrastive learning, FANoise effectively mitigates the negative impacts of noise while preserving its benefits. Under this theoretically grounded framework, comprehensive experiments demonstrate that FANoise consistently improves overall performance on multimodal tasks across various base VLM models.

PDF Details DOI

TCS Journal 2025 Journal Article

Algorithms for Shortest Path Tour Problem

Yucen Gao
Zhuoran Li
Jingyu He
Jun Fang
Hui Gao
Xiaofeng Gao
Guihai Chen

Carpooling route planning becomes an important problem with the growth of low-carbon traffic systems. When each passenger has multiple potential pick-up/drop-off locations, the problem will be more challenging. In the paper, we discussed a simplified carpooling route planning problem, namely the Shortest Path Tour Problem (SPTP), whose aim is to find a single-origin single-destination shortest path through an ordered sequence of disjoint node subsets. We propose Stage Dijkstra and Global Dijkstra algorithms to find the optimal shortest path, with the time complexity of O ( l ( n + m ) log ⁡ n ) and O ( l ( n + m ) log ⁡ ( l n ) ) respectively, where l represents the number of node subsets. To the best of our knowledge, O ( l ( n + m ) log ⁡ n ) is the best time complexity of the exact algorithms for SPTP. Besides, the Stage Dijkstra and Global Dijkstra algorithms both have the linear space complexity, which is highly suitable for resource-constrained environments. Experiments conducted on large-scale road networks and synthetic datasets demonstrate the effectiveness and efficiency of our proposed algorithms in terms of running time and memory consumption.

Details DOI

NeurIPS Conference 2025 Conference Paper

Salient Concept-Aware Generative Data Augmentation

Tianchen Zhao
Xuanbai Chen
Zhihua Li
Jun Fang
Dongsheng An
Xiang Xu
Zhuowen Tu
Yifan Xing

Recent generative data augmentation methods conditioned on both image and text prompts struggle to balance between fidelity and diversity, as it is challenging to preserve essential image details while aligning with varied text prompts. This challenge arises because representations in the synthesis process often become entangled with non-essential input image attributes such as environmental contexts, creating conflicts with text prompts intended to modify these elements. To address this, we propose a personalized image generation framework that uses a salient concept-aware image embedding model to reduce the influence of irrelevant visual details during the synthesis process, thereby maintaining intuitive alignment between image and text inputs. By generating images that better preserve class-discriminative features with additional controlled variations, our framework effectively enhances the diversity of training datasets and thereby improves the robustness of downstream models. Our approach demonstrates superior performance across eight fine-grained vision datasets, outperforming state-of-the-art augmentation methods with averaged classification accuracy improvements by 0. 73\% and 6. 5\% under conventional and long-tail settings, respectively.

PDF Details

ICLR Conference 2024 Conference Paper

Threshold-Consistent Margin Loss for Open-World Deep Metric Learning

Qin Zhang
Linghan Xu
Jun Fang
Qingming Tang
Ying Nian Wu
Joseph Tighe
Yifan Xing

Existing losses used in deep metric learning (DML) for image retrieval often lead to highly non-uniform intra-class and inter-class representation structures across test classes and data distributions. When combined with the common practice of using a fixed threshold to declare a match, this gives rise to significant performance variations in terms of false accept rate (FAR) and false reject rate (FRR) across test classes and data distributions. We define this issue in DML as threshold inconsistency. In real-world applications, such inconsistency often complicates the threshold selection process when deploying large-scale image retrieval systems. To measure this inconsistency, we propose a novel variance-based metric called Operating-Point-Inconsistency-Score (OPIS) that quantifies the variance in the operating characteristics across classes. Using the OPIS metric, we find that achieving high accuracy levels in a DML model does not automatically guarantee threshold consistency. In fact, our investigation reveals a Pareto frontier in the high-accuracy regime, where existing methods to improve accuracy often lead to degradation in threshold consistency. To address this trade-off, we introduce the Threshold-Consistent Margin (TCM) loss, a simple yet effective regularization technique that promotes uniformity in representation structures across classes by selectively penalizing hard sample pairs. Large-scale experiments demonstrate TCM's effectiveness in enhancing threshold consistency while preserving accuracy, simplifying the threshold selection process in practical DML settings.

Details

NeurIPS Conference 2022 Conference Paper

An In-depth Study of Stochastic Backpropagation

Jun Fang
Mingze Xu
Hao Chen
Bing Shuai
Zhuowen Tu
Joseph Tighe

In this paper, we provide an in-depth study of Stochastic Backpropagation (SBP) when training deep neural networks for standard image classification and object detection tasks. During backward propagation, SBP calculates gradients by using only a subset of feature maps to save GPU memory and computational cost. We interpret SBP as an efficient way to implement stochastic gradient decent by performing backpropagation dropout, which leads to significant memory saving and training run-time reduction, with a minimal impact on the overall model accuracy. We offer best practices to apply SBP for training image recognition models, which can be adopted in learning a wide range of deep neural networks. Experiments on image classification and object detection show that SBP can save up to 40% of GPU memory with less than 1% accuracy degradation. Code is available at: https: //github. com/amazon-research/stochastic-backpropagation

PDF Details

IS Journal 2022 Journal Article

Deep Fusion for Travel Time Estimation Based on Road Network Topology

Fuyong Sun
Ruipeng Gao
Weiwei Xing
Yaoxue Zhang
Wei Lu
Jun Fang
Shui Liu

With the wide application of vehicular location-based services, precise estimation of the travel time plays a crucial role in intelligent transportation systems, such as driving navigation, traffic monitoring, and route planning. Recent methods have made significant progress on public datasets, but are not satisfied for current ride-hailing platforms with complex road network topology and dynamic traffic fluctuation. In this article, we propose an end-to-end Deep Fusion framework for Travel Time Estimation, which exploits multisource heterogeneous traffic information within an encoder–decoder architecture. Specifically, we explore a relational fusion network to learn the relationship of road link segments, and employ an attention mechanism to capture efficient correlations among spatial and temporal features. Extensive experiments have been conducted on two large-scale real-world traffic datasets collected by DiDi Corporation (DiDi) platform, and the results have demonstrated our effectiveness compared with the state of the art.

Details DOI