Saved in:
Bibliographic Details
Main Authors: Yang, Hanzhe, Wu, Youlong, Wen, Dingzhu, Zhou, Yong, Shi, Yuanming
Format: Preprint
Published: 2024
Subjects:
Online Access:https://arxiv.org/abs/2412.08222
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866915246881374208
author Yang, Hanzhe
Wu, Youlong
Wen, Dingzhu
Zhou, Yong
Shi, Yuanming
author_facet Yang, Hanzhe
Wu, Youlong
Wen, Dingzhu
Zhou, Yong
Shi, Yuanming
contents The Information Bottleneck (IB) principle has emerged as a promising approach for enhancing the generalization, robustness, and interpretability of deep neural networks, demonstrating efficacy across image segmentation, document clustering, and semantic communication. Among IB implementations, the IB Lagrangian method, employing Lagrangian multipliers, is widely adopted. While numerous methods for the optimizations of IB Lagrangian based on variational bounds and neural estimators are feasible, their performance is highly dependent on the quality of their design, which is inherently prone to errors. To address this limitation, we introduce Structured IB, a framework for investigating potential structured features. By incorporating auxiliary encoders to extract missing informative features, we generate more informative representations. Our experiments demonstrate superior prediction accuracy and task-relevant information preservation compared to the original IB Lagrangian method, even with reduced network size.
format Preprint
id arxiv_https___arxiv_org_abs_2412_08222
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Structured IB: Improving Information Bottleneck with Structured Feature Learning
Yang, Hanzhe
Wu, Youlong
Wen, Dingzhu
Zhou, Yong
Shi, Yuanming
Information Theory
Machine Learning
The Information Bottleneck (IB) principle has emerged as a promising approach for enhancing the generalization, robustness, and interpretability of deep neural networks, demonstrating efficacy across image segmentation, document clustering, and semantic communication. Among IB implementations, the IB Lagrangian method, employing Lagrangian multipliers, is widely adopted. While numerous methods for the optimizations of IB Lagrangian based on variational bounds and neural estimators are feasible, their performance is highly dependent on the quality of their design, which is inherently prone to errors. To address this limitation, we introduce Structured IB, a framework for investigating potential structured features. By incorporating auxiliary encoders to extract missing informative features, we generate more informative representations. Our experiments demonstrate superior prediction accuracy and task-relevant information preservation compared to the original IB Lagrangian method, even with reduced network size.
title Structured IB: Improving Information Bottleneck with Structured Feature Learning
topic Information Theory
Machine Learning
url https://arxiv.org/abs/2412.08222