Enhancing the Performance of Neural Networks Through Causal Discovery and Integration of Domain Knowledge

2023-11-29

arXiv.org

AI Summary

To address the challenge of effectively integrating causal prior knowledge into neural networks for improved predictive accuracy, this paper proposes Causal-Informed Neural Networks (CINNs). CINNs jointly couple continuous differentiable DAG learning—based on the NOTEARS framework—with neural architecture design: causal structures are strictly mapped to hierarchical network topologies via order-preserving encoding, while a node-type-aware multi-output supervision mechanism and a conflict-gradient projection module enable end-to-end co-optimization of causal structure discovery and parameter learning. A novel multi-task loss function explicitly encodes causal directionality and task synergy. Extensive experiments on multiple UCI benchmarks demonstrate significant improvements over state-of-the-art methods. Ablation studies confirm that causal-structure-guided knowledge integration yields interpretable, progressive performance gains.

Abstract

In this paper, we develop a generic methodology to encode hierarchical causality structure among observed variables into a neural network in order to improve its predictive performance. The proposed methodology, called causality-informed neural network (CINN), leverages three coherent steps to systematically map the structural causal knowledge into the layer-to-layer design of neural network while strictly preserving the orientation of every causal relationship. In the first step, CINN discovers causal relationships from observational data via directed acyclic graph (DAG) learning, where causal discovery is recast as a continuous optimization problem to avoid the combinatorial nature. In the second step, the discovered hierarchical causality structure among observed variables is systematically encoded into neural network through a dedicated architecture and customized loss function. By categorizing variables in the causal DAG as root, intermediate, and leaf nodes, the hierarchical causal DAG is translated into CINN with a one-to-one correspondence between nodes in the causal DAG and units in the CINN while maintaining the relative order among these nodes. Regarding the loss function, both intermediate and leaf nodes in the DAG graph are treated as target outputs during CINN training so as to drive co-learning of causal relationships among different types of nodes. As multiple loss components emerge in CINN, we leverage the projection of conflicting gradients to mitigate gradient interference among the multiple learning tasks. Computational experiments across a broad spectrum of UCI data sets demonstrate substantial advantages of CINN in predictive performance over other state-of-the-art methods. In addition, an ablation study underscores the value of integrating structural and quantitative causal knowledge in enhancing the neural network's predictive performance incrementally.

Problem

• Neural Network
• Prediction Accuracy
• Causal Relationships

Innovation

• Causal Integration
• Neural Networks Enhancement
• Quantified Causal Knowledge