<ul><li>The field of hypothesis generation in neuroscience aims to reduce costs by narrowing down interventional studies needed to study various phenomena.</li><li>Existing machine learning methods can generate scientific hypotheses from complex datasets, but they often assume static causal relationships over time, limiting their applicability to systems with dynamic, state-dependent behavior.</li><li>A novel method is proposed to model dynamic causal graphs as a conditionally weighted superposition of static graphs, allowing the detection of complex, time-varying interactions beyond linear limitations.</li><li>The method improves f1-scores of predicted dynamic causal patterns significantly over baselines in experiments and demonstrates the ability to uncover relationships in real brain data linked to specific behavioral states.</li></ul>

Generating Hypotheses of Dynamic Causal Graphs in Neuroscience: Leveraging Generative Factor Models of Observed Time Series

Discover more