Ambuj K. Singh Research Group

University of California, Santa Barbara

How can learned graph representations be measured and designed to improve downstream tasks and robustness? The group develops geometry- and topology-based methods and graph-based models for drug discovery and chemical informatics. Researchers analyze social networks combining structure with attributes like sentiment and trust to model team decision dynamics and trust evolution. The group builds methods to integrate multimodal biological and imaging data to find discriminative sub-networks that predict global clinical or brain network states. They design scalable algorithms for querying and modeling dynamic composite networks to detect unusual patterns and enable efficient sampling.