D-VAE: A Variational Autoencoder for Directed Acyclic Graphs
This addresses the need for generative models for DAGs, which are crucial in machine learning applications like neural networks and Bayesian networks, representing an incremental advancement with a novel encoding method.
The paper tackles the problem of generating directed acyclic graphs (DAGs) by proposing D-VAE, a variational autoencoder that encodes DAGs using an asynchronous message passing scheme, and demonstrates its effectiveness in neural architecture search and Bayesian network structure learning by generating novel, valid DAGs and a smooth latent space for optimization.
Graph structured data are abundant in the real world. Among different graph types, directed acyclic graphs (DAGs) are of particular interest to machine learning researchers, as many machine learning models are realized as computations on DAGs, including neural networks and Bayesian networks. In this paper, we study deep generative models for DAGs, and propose a novel DAG variational autoencoder (D-VAE). To encode DAGs into the latent space, we leverage graph neural networks. We propose an asynchronous message passing scheme that allows encoding the computations on DAGs, rather than using existing simultaneous message passing schemes to encode local graph structures. We demonstrate the effectiveness of our proposed DVAE through two tasks: neural architecture search and Bayesian network structure learning. Experiments show that our model not only generates novel and valid DAGs, but also produces a smooth latent space that facilitates searching for DAGs with better performance through Bayesian optimization.