An Information Criterion for Controlled Disentanglement of Multimodal Data

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Wang, Chenyu, Gupta, Sharut, Zhang, Xinyi, Tonekaboni, Sana, Jegelka, Stefanie, Jaakkola, Tommi, Uhler, Caroline
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866915201108934656
author Wang, Chenyu
Gupta, Sharut
Zhang, Xinyi
Tonekaboni, Sana
Jegelka, Stefanie
Jaakkola, Tommi
Uhler, Caroline
author_facet Wang, Chenyu
Gupta, Sharut
Zhang, Xinyi
Tonekaboni, Sana
Jegelka, Stefanie
Jaakkola, Tommi
Uhler, Caroline
contents Multimodal representation learning seeks to relate and decompose information inherent in multiple modalities. By disentangling modality-specific information from information that is shared across modalities, we can improve interpretability and robustness and enable downstream tasks such as the generation of counterfactual outcomes. Separating the two types of information is challenging since they are often deeply entangled in many real-world applications. We propose Disentangled Self-Supervised Learning (DisentangledSSL), a novel self-supervised approach for learning disentangled representations. We present a comprehensive analysis of the optimality of each disentangled representation, particularly focusing on the scenario not covered in prior work where the so-called Minimum Necessary Information (MNI) point is not attainable. We demonstrate that DisentangledSSL successfully learns shared and modality-specific features on multiple synthetic and real-world datasets and consistently outperforms baselines on various downstream tasks, including prediction tasks for vision-language data, as well as molecule-phenotype retrieval tasks for biological data. The code is available at https://github.com/uhlerlab/DisentangledSSL.
format Preprint
id arxiv_https___arxiv_org_abs_2410_23996
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle An Information Criterion for Controlled Disentanglement of Multimodal Data
Wang, Chenyu
Gupta, Sharut
Zhang, Xinyi
Tonekaboni, Sana
Jegelka, Stefanie
Jaakkola, Tommi
Uhler, Caroline
Machine Learning
Artificial Intelligence
Information Theory
Multimodal representation learning seeks to relate and decompose information inherent in multiple modalities. By disentangling modality-specific information from information that is shared across modalities, we can improve interpretability and robustness and enable downstream tasks such as the generation of counterfactual outcomes. Separating the two types of information is challenging since they are often deeply entangled in many real-world applications. We propose Disentangled Self-Supervised Learning (DisentangledSSL), a novel self-supervised approach for learning disentangled representations. We present a comprehensive analysis of the optimality of each disentangled representation, particularly focusing on the scenario not covered in prior work where the so-called Minimum Necessary Information (MNI) point is not attainable. We demonstrate that DisentangledSSL successfully learns shared and modality-specific features on multiple synthetic and real-world datasets and consistently outperforms baselines on various downstream tasks, including prediction tasks for vision-language data, as well as molecule-phenotype retrieval tasks for biological data. The code is available at https://github.com/uhlerlab/DisentangledSSL.
title An Information Criterion for Controlled Disentanglement of Multimodal Data
topic Machine Learning
Artificial Intelligence
Information Theory
url https://arxiv.org/abs/2410.23996