Optimal Transport for Comparative Analysis of Neural Representations

I will develop novel tools to compare internal representations in AI systems and human brains at the level of individual units, spatial organization, and processing hierarchies, and to identify which stimuli evoke shared versus divergent representations – enabling fine-grained, interpretable benchmarks of internal alignment beyond output behavior.

Biography

What can brains teach us about how AI actually learns — and vice versa? Meenakshi Khosla is an Assistant Professor of Cognitive Science at UC San Diego, where she leads the NeuroML research group. Holding a PhD from Cornell, she previously worked as a postdoc at MIT’s McGovern Institute. Her research develops large-scale computational tools to compare internal representations across biological and artificial neural networks, bridging neuroscience and AI to benchmark how closely machines align with the human mind.