Policy Message Passing: A New Algorithm for Probabilistic Graph Inference
This addresses the need for more robust and effective graph inference methods in machine learning, particularly for handling noisy edges and complex reasoning tasks, representing a novel approach rather than an incremental improvement.
The paper tackles the problem of graph-structured neural network inference by introducing the Policy Message Passing algorithm, which reformulates information aggregation as stochastic sequential processes, and shows it consistently outperforms state-of-the-art models by a significant margin on multiple complex graph reasoning and prediction tasks.
A general graph-structured neural network architecture operates on graphs through two core components: (1) complex enough message functions; (2) a fixed information aggregation process. In this paper, we present the Policy Message Passing algorithm, which takes a probabilistic perspective and reformulates the whole information aggregation as stochastic sequential processes. The algorithm works on a much larger search space, utilizes reasoning history to perform inference, and is robust to noisy edges. We apply our algorithm to multiple complex graph reasoning and prediction tasks and show that our algorithm consistently outperforms state-of-the-art graph-structured models by a significant margin.