DAG Learning from Zero-Inflated Count Data Using Continuous Optimization

TL;DR

This paper introduces ZICO, a novel continuous optimization method for learning network structures from zero-inflated count data, demonstrating improved speed and accuracy over existing methods, especially in gene regulatory network inference.

Contribution

The paper proposes ZICO, a new zero-inflated generalized linear model approach with a differentiable acyclicity constraint for efficient network structure learning.

Findings

ZICO outperforms existing algorithms in simulated data.

ZICO achieves comparable or better results in gene regulatory network inference.

ZICO is scalable to larger variable sets with practical runtimes.

Abstract

We address network structure learning from zero-inflated count data by casting each node as a zero-inflated generalized linear model and optimizing a smooth, score-based objective under a directed acyclic graph constraint. Our Zero-Inflated Continuous Optimization (ZICO) approach uses node-wise likelihoods with canonical links and enforces acyclicity through a differentiable surrogate constraint combined with sparsity regularization. ZICO achieves superior performance with faster runtimes on simulated data. It also performs comparably to or better than common algorithms for reverse engineering gene regulatory networks. ZICO is fully vectorized and mini-batched, enabling learning on larger variable sets with practical runtimes in a wide range of domains.