CDRM: Causal disentangled representation learning for missing data

Mingjie Chen, Hongcheng Wang, Ruxin Wang, Yuzhong Peng, Hao Zhang

Abstract: Highlights•A causal disentangled representation learning framework CDRM for missing data is proposed.•The causal relationships of missing data are recovered by capturing the feature interactions with edge embeddings.•Causal relationships of the data are incorporated as constraints to learn disentangled representations.•Counterfactual outputs can be generated by manipulating the intervening causal variables.