arXiv 2103.027702021-03-04cs.CVcs.LG

SVMax: A Feature Embedding Regularizer

University of Maryland, College Park（马里兰大学帕克分校）

Ahmed Taha, Alex Hanson, Abhinav Shrivastava, Larry Davis

英文摘要：

A neural network regularizer (e.g., weight decay) boosts performance by explicitly penalizing the complexity of a network. In this paper, we penalize inferior network activations -- feature embeddings -- which in turn regularize the network's weights implicitly. We propose singular value maximization (SVMax) to learn a more uniform feature embedding. The SVMax regularizer supports both supervised and unsupervised learning. Our formulation mitigates model collapse and enables larger learning rates. We evaluate the SVMax regularizer using both retrieval and generative adversarial networks. We leverage a synthetic mixture of Gaussians dataset to evaluate SVMax in an unsupervised setting. For retrieval networks, SVMax achieves significant improvement margins across various ranking losses. Code available at https://bit.ly/3jNkgDt

机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。

阅读原文

阅读 PDF ↗

arXiv 页面 ↗HTML 原文 ↗

本文内容

英文摘要相关论文

机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。

继续探索

相关论文

Homology-constrained vector quantization entropy regularizer

This paper describes an entropy regularization term for vector quantization (VQ) based on the analysis of persistent homology of the VQ embeddings. Higher embedding entropy positively correlates with higher codebook utilization, mitigating overfit towards the identity and codebook collapse in VQ-based autoencoders [1]. We show that homology-constrained regularization is an effective way to increase entropy of the VQ process (approximated to input entropy) while preserving the approximated topology in the quantized latent space, averaged over mini batches. This work further explores some patter…

2022-11-25cs.LG

MotionJEPA：通过在潜在空间中捕捉视觉变化来防止时间特征坍缩

MotionJEPA: Preventing Temporal Feature Collapse by Capturing Visual Changes in Latent Space

University of Oxford（牛津大学）；vivo Tech Research GmbH（vivo科技研究有限公司）；Bielefeld University（比勒费尔德大学）；Slater Labs（Slater实验室）；Cold Spring Harbor Laboratory（冷泉港实验室）；Brown University（布朗大学）；AMI Labs（AMI实验室）；vivo BlueImage Lab, vivo Mobile Communication Co., Ltd., China（vivo蓝心影像实验室，维沃移动通信有限公司）

MotionJEPA通过引入DISReg正则化器，在潜在空间中预测时间差分图像嵌入，平衡静态与动态特征学习，防止时间特征坍缩，从而提升下游规划成功率。

2026-09-22cs.CV

UniJEPA：面向任务无关视觉世界建模的统一联合嵌入预测架构

UniJEPA: A Unified Joint-Embedding Predictive Architecture for Task-Agnostic Visual World Modeling

UniJEPA是统一JEPAs，共享潜在空间联合学习图像级与视频级预测，抗坍塌，经后训练可零样本规划，在多基准上性能相当或更优，规划速度更快。

2026-08-10cs.CV

Boosting Standard Classification Architectures Through a Ranking Regularizer

University of Maryland, College Park（马里兰大学帕克分校）；Honda Research Institute, USA（本田研究院（美国））

We employ triplet loss as a feature embedding regularizer to boost classification performance. Standard architectures, like ResNet and Inception, are extended to support both losses with minimal hyper-parameter tuning. This promotes generality while fine-tuning pretrained networks. Triplet loss is a powerful surrogate for recently proposed embedding regularizers. Yet, it is avoided due to large batch-size requirement and high computational cost. Through our experiments, we re-assess these assumptions. During inference, our network supports both classification and embedding tasks without any co…

2020-03-02cs.CV

Metric Learning With HORDE: High-Order Regularizer for Deep Embeddings

Université Paris Seine（巴黎塞纳大学）；UCP（塞吉-蓬图瓦兹大学）；ENSEA（国立高等电子及应用学院）；CNRS（法国国家科学研究中心）；École des Ponts（巴黎路桥学院）；UPE（巴黎东大学）；Pôle Judiciaire de la Gendarmerie Nationale（国家宪兵司法中心）

Learning an effective similarity measure between image representations is key to the success of recent advances in visual search tasks (e.g. verification or zero-shot learning). Although the metric learning part is well addressed, this metric is usually computed over the average of the extracted deep features. This representation is then trained to be discriminative. However, these deep features tend to be scattered across the feature space. Consequently, the representations are not robust to outliers, object occlusions, background variations, etc. In this paper, we tackle this scattering prob…

2019-08-07cs.CV

Improving GAN with neighbors embedding and gradient matching

ST Electronics - SUTD Cyber Security Laboratory（新科电子-新加坡科技设计大学网络安全实验室）；Singapore University of Technology and Design（新加坡科技设计大学）

We propose two new techniques for training Generative Adversarial Networks (GANs). Our objectives are to alleviate mode collapse in GAN and improve the quality of the generated samples. First, we propose neighbor embedding, a manifold learning-based regularization to explicitly retain local structures of latent samples in the generated samples. This prevents generator from producing nearly identical data samples from different latent samples, and reduces mode collapse. We propose an inverse t-SNE regularizer to achieve this. Second, we propose a new technique, gradient matching, to align the d…

2018-11-04cs.CV

更多相关论文收起更多论文

因果变分深度嵌入：一类针对混杂图像的可干预生成器

A Causal Bootstrap for Diverse Image Generation under Unobserved Confounding

AWS AI Lab（亚马逊云科技AI实验室）

提出CauVaDE，基于规范增广SCM将未观测混杂因子分解为离散潜变量，通过熵正则化追踪可行干预分布族，在图像数据上生成多样化干预样本并提升FID。

2026-10-27cs.LG

ASMR: Learning Attribute-Based Person Search with Adaptive Semantic Margin Regularizer

POSTECH（浦项科技大学）；Graduate School of AI POSTECH（浦项科技大学人工智能研究生院）

Attribute-based person search is the task of finding person images that are best matched with a set of text attributes given as query. The main challenge of this task is the large modality gap between attributes and images. To reduce the gap, we present a new loss for learning cross-modal embeddings in the context of attribute-based person search. We regard a set of attributes as a category of people sharing the same traits. In a joint embedding space of the two modalities, our loss pulls images close to their person categories for modality alignment. More importantly, it pushes apart a pair o…

2021-08-10cs.AIcs.CVcs.LG

UR-JEPA：均匀可整流性作为联合嵌入预测架构的正则化器

UR-JEPA: Uniform Rectifiability as a Regularizer for Joint-Embedding Predictive Architectures

Spatiolyx LLC（Spatiolyx公司）

提出UR-JEPA，通过高斯核平滑的Carleson型平方函数实现均匀n-可整流测度正则化，防止表示坍塌，在多个数据集上达到与LeJEPA相当的峰值精度但具有更低的种子方差。

2026-05-31cs.AIcs.CVcs.LG

条件噪声是LoRA微调的自由正则化器：组织病理学中基于扩散的伪影检测无需病理编码器

Conditioning noise is a free regularizer for LoRA fine-tuning: no pathology encoder required for diffusion-based artifact detection in histopathology

University of Piraeus（比雷埃夫斯大学）

本研究证明在LoRA微调中引入随机高斯条件噪声可提升组织病理学伪影检测的干净/伪影分离度，无需额外编码器，且效果经多轮消融和预注册实验验证。

2026-09-16cs.CVeess.IV

在深度神经网络中嵌入水印

Embedding Watermarks into Deep Neural Networks

KDDI Research, Inc.（KDDI研究公司）；Saitama Research, Inc.（埼玉研究公司）；National Institute of Informatics（日本国立信息学研究所）

深度神经网络最近取得了重大进展。共享这些深度神经网络的训练模型在研究或开发深度神经网络系统的快速进展中是非常重要的。同时，有必要保护共享训练模型的权利。为此，我们建议使用数字水印技术来保护知识产权或检测训练模型的知识产权侵权。首先，我们提出了一个新的问题：将水印嵌入深度神经网络。我们还定义了要求，嵌入的情况下，和攻击类型的水印深度神经网络。其次，我们提出了一个通用的框架嵌入水印到模型参数使用参数正则化。我们的方法不会损害嵌入水印的网络的性能。最后，我们进行了全面的实验，以揭示水印的潜力，深神经网络作为这个新问题的基础。我们证明了我们的框架可以在从头开始训练网络、微调和提取的情况下嵌入水印，而不会损害深度神经网络的性能。嵌入的水印即使在微调或参数修剪后也不会消失;即使删除65%的参数被修剪，水印也完全保留。本研究的实现是：https://github.com/yu4u/dnn-watermark

2017-04-20cs.CV

双头胜一：双退化表示用于盲超分辨率

Two Heads Better than One: Dual Degradation Representation for Blind Super-Resolution

本文提出双退化表示方法，通过双分支网络预测模糊和噪声嵌入，提升盲超分辨率任务的性能。

2025-11-21cs.CV

通过来自人脸的一致性知识蒸馏进行眼周嵌入学习

Periocular Embedding Learning with Consistent Knowledge Distillation from Face

Northeastern University（东北大学）；AiV Research, AiV Co.（AiV公司AiV研究院）；Yonsei University（延世大学）；Institute for Basic Science（基础科学研究院）；Samsung Electronics Co., Ltd.（三星电子株式会社）

针对眼周生物特征判别性不足的问题，提出一致性知识蒸馏方法，从人脸网络向眼周网络迁移判别信息，在预测层和特征层施加一致性约束，实现野外环境下鲁棒的眼周识别，在标准基准上取得最优结果。

2024-01-28cs.CV

GREN: Graph-Regularized Embedding Network for Weakly-Supervised Disease Localization in X-ray Images

HwaMei Hospital, University of Chinese Academy of Sciences（中国科学院大学华美医院）；Ningbo Institute of Life and Health Industry, University of Chinese Academy of Sciences（中国科学院大学宁波生命与健康产业研究院）；The University of Hong Kong（香港大学）；Institute of Automation, Chinese Academy of Sciences（中国科学院自动化研究所）；Beihang University（北京航空航天大学）

Locating diseases in chest X-ray images with few careful annotations saves large human effort. Recent works approached this task with innovative weakly-supervised algorithms such as multi-instance learning (MIL) and class activation maps (CAM), however, these methods often yield inaccurate or incomplete regions. One of the reasons is the neglection of the pathological implications hidden in the relationship across anatomical regions within each image and the relationship across images. In this paper, we argue that the cross-region and cross-image relationship, as contextual and compensating in…

2022-08-05cs.AIcs.CV

基于自集成与条件嵌入对齐的动作识别无监督域自适应

Unsupervised Domain Adaptation for Action Recognition via Self-Ensembling and Conditional Embedding Alignment

University of Maryland, Baltimore County（马里兰大学巴尔的摩郡分校）；Indian Institute of Technology Jodhpur（焦特布尔印度理工学院）；Amazon（亚马逊公司）

本文提出μDAR架构，通过一致性正则化增强、时间集成伪标签生成和条件分布对齐三个模块的联合优化，解决可穿戴动作识别中标注稀缺和域偏差问题，在四个基准数据集上平均宏观F1分数提升约4-12%。

2024-10-23cs.AIcs.CV