DCCD: Reducing Neural Network Redundancy via Distillation

计算机科学 冗余(工程) 人工智能 蒸馏 过度拟合 机器学习 人工神经网络 模式识别(心理学) 操作系统 有机化学 化学
作者
Liu Yuang,Jun Chen,Yong Liu
出处
期刊:IEEE transactions on neural networks and learning systems [Institute of Electrical and Electronics Engineers]
卷期号:35 (7): 10006-10017 被引量:5
标识
DOI:10.1109/tnnls.2023.3238337
摘要

Deep neural models have achieved remarkable performance on various supervised and unsupervised learning tasks, but it is a challenge to deploy these large-size networks on resource-limited devices. As a representative type of model compression and acceleration methods, knowledge distillation (KD) solves this problem by transferring knowledge from heavy teachers to lightweight students. However, most distillation methods focus on imitating the responses of teacher networks but ignore the information redundancy of student networks. In this article, we propose a novel distillation framework difference-based channel contrastive distillation (DCCD), which introduces channel contrastive knowledge and dynamic difference knowledge into student networks for redundancy reduction. At the feature level, we construct an efficient contrastive objective that broadens student networks' feature expression space and preserves richer information in the feature extraction stage. At the final output level, more detailed knowledge is extracted from teacher networks by making a difference between multiview augmented responses of the same instance. We enhance student networks to be more sensitive to minor dynamic changes. With the improvement of two aspects of DCCD, the student network gains contrastive and difference knowledge and reduces its overfitting and redundancy. Finally, we achieve surprising results that the student approaches and even outperforms the teacher in test accuracy on CIFAR-100. We reduce the top-1 error to 28.16% on ImageNet classification and 24.15% for cross-model transfer with ResNet-18. Empirical experiments and ablation studies on popular datasets show that our proposed method can achieve state-of-the-art accuracy compared with other distillation methods.
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
leo发布了新的文献求助10
刚刚
ibigbird完成签到,获得积分10
1秒前
赵彬旭发布了新的文献求助10
1秒前
亨利完成签到,获得积分10
1秒前
2秒前
2秒前
以后完成签到,获得积分10
2秒前
2秒前
2秒前
luohan完成签到,获得积分10
2秒前
斯文败类的应助被善良鸡翅采纳,获得10
3秒前
zhao完成签到,获得积分10
5秒前
5秒前
龟龟酱完成签到,获得积分10
5秒前
sxy完成签到,获得积分10
5秒前
黄饱饱完成签到,获得积分10
5秒前
6秒前
7秒前
燕翱翔完成签到,获得积分20
7秒前
7秒前
顺利柚子完成签到,获得积分10
7秒前
Eric的应助被达不溜人采纳,获得10
8秒前
碎觉觉发布了新的文献求助30
8秒前
DDD完成签到,获得积分10
8秒前
yiii发布了新的文献求助10
9秒前
9秒前
xcgh的应助被YANG采纳,获得10
10秒前
10秒前
10秒前
深情安青的应助被Drcai采纳,获得10
10秒前
香蕉觅云的应助被苑小胖采纳,获得10
11秒前
平安完成签到,获得积分10
11秒前
小泥点子完成签到 ,获得积分10
11秒前
m0405完成签到,获得积分10
11秒前
DDD发布了新的文献求助10
11秒前
11秒前
ally发布了新的文献求助10
11秒前
SMQH完成签到,获得积分10
13秒前
123发布了新的文献求助10
13秒前
13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 8187612
求助须知:如何正确求助?哪些是违规求助? 9575734
关于积分的说明 21842594
捐赠科研通 7785476
什么是DOI,文献DOI怎么找? 3435073
关于科研通互助平台的介绍 2559478
邀请新用户注册赠送积分活动 2481952