Convolutional Learning on Directed Acyclic Graphs

arXiv:2405.03056v1 Announce Type: new
Abstract: We develop a novel convolutional architecture tailored for learning from data defined over directed acyclic graphs (DAGs). DAGs can be used to model causal relationships among variables, but their nilpotent adjacency matrices pose unique challenges towards developing DAG signal processing and machine learning tools. To address this limitation, we harness recent advances offering alternative definitions of causal shifts and convolutions for signals on DAGs. We develop a novel convolutional graph neural network that integrates learnable …

abstract architecture arxiv causal challenges convolutional cs.lg dag data graphs learning tools machine machine learning novel processing relationships signal tools type unique variables