Hybrid Causal Identification and Causal Mechanism Clustering

2025-07-29

AI Summary

Real-world observational data often originate from heterogeneous multi-source environments, where causal relationships exhibit mechanism-level heterogeneity; however, mainstream additive noise models (ANMs) assume a single homogeneous causal mechanism and thus fail to capture such complexity. To address this, we propose the Mixture Conditional Variational Inference (MCVI) model—a novel, identifiable mixture ANM that jointly integrates Gaussian mixture models with neural networks. MCVI explicitly clusters distinct causal mechanisms while simultaneously identifying causal directions. It optimizes the evidence lower bound via a mixture conditional variational autoencoder, synergizing neural nonlinearity for flexible function approximation and Gaussian mixture structure for interpretable mechanism representation. Extensive experiments on synthetic and real-world benchmarks demonstrate that MCVI significantly outperforms state-of-the-art methods: it achieves accurate heterogeneous causal direction identification and uncovers the underlying distributional structure of latent causal mechanisms. MCVI establishes a new, interpretable, and scalable paradigm for heterogeneous causal discovery.

Technology Category

• Machine Learning: Causal Learning
• Reasoning under Uncertainty: Causality
• Search and Optimization: Mixed Discrete/Continuous Search

Application Category

• User Modeling, Personalization and Recommendation: Explainable and interpretable methods for personalization
• Graph Algorithms and Modeling for the Web: Algorithms and analysis for heterogeneous, signed, attributed, multi-relational, temporal, higher-order, and annotated Web-related graphs
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Bivariate causal direction identification is a fundamental and vital problem in the causal inference field. Among binary causal methods, most methods based on additive noise only use one single causal mechanism to construct a causal model. In the real world, observations are always collected in different environments with heterogeneous causal relationships. Therefore, on observation data, this paper proposes a Mixture Conditional Variational Causal Inference model (MCVCI) to infer heterogeneous causality. Specifically, according to the identifiability of the Hybrid Additive Noise Model (HANM), MCVCI combines the superior fitting capabilities of the Gaussian mixture model and the neural network and elegantly uses the likelihoods obtained from the probabilistic bounds of the mixture conditional variational auto-encoder as causal decision criteria. Moreover, we model the casual heterogeneity into cluster numbers and propose the Mixture Conditional Variational Causal Clustering (MCVCC) method, which can reveal causal mechanism expression. Compared with state-of-the-art methods, the comprehensive best performance demonstrates the effectiveness of the methods proposed in this paper on several simulated and real data.

Problem

• Identify heterogeneous causal relationships in observed data
• Cluster causal mechanisms to reveal expression patterns
• Improve accuracy in bivariate causal direction identification

Innovation

• Hybrid Additive Noise Model identifiability
• Mixture Conditional Variational Auto-encoder likelihoods
• Causal mechanism clustering into numbers