Supervised CAusal Disentanglement

Supervised causal disentanglement aims to separate underlying factors influencing observed data, enabling a deeper understanding of complex systems and improved prediction accuracy. Current research focuses on developing self-supervised and weakly supervised methods to reduce reliance on expensive labeled datasets, employing techniques like masked structural causal models and contrastive regularization within latent variable models. These advancements are proving valuable in diverse applications, including hate speech detection, video editing, and human mobility prediction, by allowing for more robust and interpretable models that can handle noisy or incomplete data. The resulting disentangled representations offer improved generalization and facilitate targeted interventions or manipulations within the modeled system.

Papers

October 9, 2024

Zero-Shot Generalization of Vision-Based RL Without Data Augmentation
Sumeet Batra, Gaurav S. Sukhatme
Zero Shot Data Augmentation Associative Memory Vision Based Reinforcement Learning Supervised CAusal Disentanglement

April 17, 2024

Cross-Platform Hate Speech Detection with Weakly Supervised Causal Disentanglement
Paras Sheth, Tharindu Kumarage, Raha Moraffah, Aman Chadha, Huan Liu
Hate Speech Hate Speech Detection Content Moderation Causal Disentanglement Supervised CAusal Disentanglement

November 11, 2023

SCADI: Self-supervised Causal Disentanglement in Latent Variable Models
Heejeong Nam
Causal Relationship Structural Causal Model Latent Variable Model Causal Disentanglement Supervised CAusal Disentanglement

January 16, 2023

DPE: Disentanglement of Pose and Expression for General Video Portrait Editing
Youxin Pang, Yong Zhang, Weize Quan, Yanbo Fan, Xiaodong Cun, Ying Shan, Dong-ming Yan
Human Pose Disentanglement Capability Facial Motion Human Expression Pose Generation Talking Face Generation Portrait Editing Video Portrait Supervised CAusal Disentanglement

November 17, 2022

Predicting Human Mobility via Self-supervised Disentanglement Learning
Qiang Gao, Jinyu Hong, Xovee Xu, Ping Kuang, Fan Zhou, Goce Trajcevski
Human Mobility Trajectory Data Disentanglement Learning Supervised CAusal Disentanglement

Supervised CAusal Disentanglement

Papers

Zero-Shot Generalization of Vision-Based RL Without Data Augmentation

Cross-Platform Hate Speech Detection with Weakly Supervised Causal Disentanglement

SCADI: Self-supervised Causal Disentanglement in Latent Variable Models

DPE: Disentanglement of Pose and Expression for General Video Portrait Editing

Predicting Human Mobility via Self-supervised Disentanglement Learning