Learning Compact Discriminant Representation via Low-Rank Bilinear Pooling

判别式 联营 人工智能 线性判别分析 双线性插值 模式识别(心理学) 计算机科学 代表(政治) 秩(图论) 数学 机器学习 计算机视觉 组合数学 政治学 法学 政治
作者
Kun Song,Hao Li,Gong Cheng,Junwei Han,Feiping Nie,Bin Gu,Fakhri O. Karray
出处
期刊:IEEE Transactions on Pattern Analysis and Machine Intelligence [IEEE Computer Society]
卷期号:47 (12): 10914-10931 被引量:7
标识
DOI:10.1109/tpami.2025.3601355
摘要

In this paper, we explain the mechanism of bilinear pooling as a module of hard sample generation, and find that bilinear pooling significantly expands variances of the first-order vectors when it produces discriminative bilinear features. In conjunction with the extremely high dimensionality of the obtained bilinear features, those variances lead to overfitting in subsequent learning models. To solve this issue, we construct a bi-level optimization problem, where the high-level problem is the supervised classification loss, and the low-level problem is the principal component analysis (PCA). Then, we find that PCA on bilinear features is equivalent to spectral clustering, which allows us to mathematically prove that the first $\log _{2}(C)$log2(C) principal components can support the discriminant information of $C$C classes. By removing the rest principal components, the dimensionality and variances are simultaneously reduced. To the best of our knowledge, this is the first work providing a lower bound for dimension reduction for bilinear pooling. However, the PCA projection matrix $\mathbf{L}$L is prone to overfitting due to having many parameters. To address this issue, we propose a rank-$k$k general bilinear projection (RK-GBP) that decomposes $\mathbf{L}$L into two small matrices $\mathbf{U}$U and $\mathbf{V}$V, whose learnable parameters are smaller. Different from traditional bilinear projections used in factorized bilinear pooling (FBiP), our RK-GBP can preserve the orthogonality of columns in $\mathbf{L}$L by constraining the orthogonality of columns in $\mathbf{U}$U and $\mathbf{V}$V. For computational efficiency, we relax the PCA in the low-level task into a dictionary learning problem, obtaining the rank-$k$k orthogonal factorization bilinear pooling (RK-OFBP). The RK-OFBP can be considered as a general form of current factorization bilinear pooling methods (e.g., Hadamard product-based ones). Finally, we evaluate our approach on fine-grained images and large-scale datasets, demonstrating that our proposed method not only produces extremely low-dimensional features but also outperforms other methods in classification tasks. For example, our RK-OFBP can employ 32-dimensional vectors to achieve comparable results to B-CNN (Lin, 2015) (dimension: 512*512) for the 200-class classification task.
最长约 10秒,即可获得该文献文件

科研通智能强力驱动
Strongly Powered by AbleSci AI
科研通是完全免费的文献互助平台,具备全网最快的应助速度,最高的求助完成率。 对每一个文献求助,科研通都将尽心尽力,给求助人一个满意的交代。
实时播报
光风霁月完成签到 ,获得积分10
刚刚
momo完成签到,获得积分10
1秒前
坚强孤容发布了新的文献求助10
1秒前
1秒前
misscloud关注了科研通微信公众号
2秒前
orixero的应助被三羊开泰采纳,获得10
3秒前
Doro完成签到,获得积分10
3秒前
3秒前
5EN完成签到,获得积分10
4秒前
4秒前
2001完成签到,获得积分10
5秒前
kunkun发布了新的文献求助10
5秒前
6秒前
沫茶发布了新的文献求助10
6秒前
阿南发布了新的文献求助10
7秒前
青青发布了新的文献求助10
9秒前
木木发布了新的文献求助10
10秒前
10秒前
顾矜的应助被ZhuJing采纳,获得30
10秒前
詹慧子发布了新的文献求助10
10秒前
10秒前
玛卡巴卡完成签到,获得积分10
11秒前
12秒前
张tuanzi完成签到,获得积分20
13秒前
13秒前
F衣衫褴褛完成签到,获得积分10
13秒前
14秒前
科研通AI6.4的应助被无辜访彤采纳,获得10
14秒前
14秒前
wyp完成签到 ,获得积分10
15秒前
梅梅超勇敢完成签到,获得积分10
15秒前
16秒前
张tuanzi发布了新的文献求助10
16秒前
玛卡巴卡发布了新的文献求助10
17秒前
希希完成签到 ,获得积分10
17秒前
英姑的应助被戴帽子采纳,获得10
17秒前
18秒前
18秒前
18秒前
Owen的应助被X1采纳,获得10
19秒前
高分求助中
(应助此贴封号)【重要!!请各用户(尤其是新用户)详细阅读】【科研通的精品贴汇总】 10000
拉伸高分子熔体中缠结行为的模拟和小角中子散射研究 1700
The World Atlas of Wine 9th edition 1000
奈特基础组织学:配套组织病理学(第4版)(Netter's Essential Histology With Correlated Histopathology) 800
八旗制度の研究 600
高分子基功能界面水蒸发材料与应用研究 500
外场增强的水汽发电研究 500
热门求助领域 (近24小时)
化学 材料科学 医学 生物 纳米技术 计算机科学 工程类 有机化学 化学工程 生物化学 物理 复合材料 内科学 催化作用 光电子学 人工智能 细胞生物学 无机化学 物理化学 数学
热门帖子
关注 科研通微信公众号,转发送积分 8186898
求助须知:如何正确求助?哪些是违规求助? 9575232
关于积分的说明 21840305
捐赠科研通 7784667
什么是DOI,文献DOI怎么找? 3434891
关于科研通互助平台的介绍 2559294
邀请新用户注册赠送积分活动 2481741