Optimizing OOD Detection in Molecular Graphs: A Novel Approach with Diffusion Models

arXiv:2404.15625v1 Announce Type: new
Abstract: The open-world test dataset is often mixed with out-of-distribution (OOD) samples, where the deployed models will struggle to make accurate predictions. Traditional detection methods need to trade off OOD detection and in-distribution (ID) classification performance since they share the same representation learning model. In this work, we propose to detect OOD molecules by adopting an auxiliary diffusion model-based framework, which compares similarities between input molecules and reconstructed graphs. Due to the generative bias towards reconstructing …

abstract arxiv classification cs.lg dataset detection detection methods diffusion diffusion models distribution graphs mixed novel open-world performance predictions representation representation learning samples struggle test trade type will world