Showing 1–1 of 1 results for author: Wren, A J

Learning Discrete Directed Acyclic Graphs via Backpropagation

Authors: Andrew J. Wren, Pasquale Minervini, Luca Franceschi, Valentina Zantedeschi

Abstract: Recently continuous relaxations have been proposed in order to learn Directed Acyclic Graphs (DAGs) from data by backpropagation, instead of using combinatorial optimization. However, a number of techniques for fully discrete backpropagation could instead be applied. In this paper, we explore that direction and propose DAG-DB, a framework for learning DAGs by Discrete Backpropagation. Based on the… ▽ More Recently continuous relaxations have been proposed in order to learn Directed Acyclic Graphs (DAGs) from data by backpropagation, instead of using combinatorial optimization. However, a number of techniques for fully discrete backpropagation could instead be applied. In this paper, we explore that direction and propose DAG-DB, a framework for learning DAGs by Discrete Backpropagation. Based on the architecture of Implicit Maximum Likelihood Estimation [I-MLE, arXiv:2106.01798], DAG-DB adopts a probabilistic approach to the problem, sampling binary adjacency matrices from an implicit probability distribution. DAG-DB learns a parameter for the distribution from the loss incurred by each sample, performing competitively using either of two fully discrete backpropagation techniques, namely I-MLE and Straight-Through Estimation. △ Less

Submitted 27 October, 2022; originally announced October 2022.

Comments: 15 pages, 2 figures, 7 tables. Accepted for NeurIPS 2022 workshops on: Causal Machine Learning for Real-World Impact; and Neuro Causal and Symbolic AI