Coordinated Disentanglement with Iterative Mode Discovery Under Hidden Correlations

2026/07/19 by Rong Hu, Ling Chen
#cs.AI

paper · pdf

Abstract

Disentangled representation learning is a powerful paradigm for robust attribute prediction. While recent methods address attribute correlations, hidden correlations remain underexplored, where data under the value of a certain attribute exhibit underlying modes correlated with other attributes. To preserve mode information and achieve disentanglement, we jointly discover modes and enforce mode-based conditional independence. Yet, the interdependency between these two modules may lead to error amplification under naive iterations. We propose Coordinated Disentanglement with Iterative mode Discovery (CoDID), an end-to-end framework featuring a dynamic architecture that adapts to evolving number of modes, and a coordination mechanism that mitigates error amplification via meta-optimization. Empirical results demonstrate the state-of-the-art performance on diverse tasks.

Citations

• PyTorch: An Imperative Style, High-Performance Deep Learning Library
• Comparing clusterings—an information based distance
• Monte Carlo sampling methods using Markov chains and their applications

Related

• Disentanglement and Generalization Under Correlation Shifts
• On Disentangled Representations Learned From Correlated Data
• Conditional Mutual Information for Disentangled Representations in Reinforcement Learning
• Self-Distilled Disentangled Learning for Counterfactual Prediction
• Leveraging Relational Information for Learning Weakly Disentangled Representations