Overview
Speaker: Siwei Wang, NITMB Affiliate Member and Assistant Professor in the Department of Neurobiology and Behavior at Stony brook University
Description: This workshop shows how math comes alive in neuroscience. It opens with a presentation of the speaker’s ongoing research on cuttlefish camouflage. The study elucidates how people detect animals hidden in complex natural scenes. It pairs human eye tracking with a deep neural network metric that scores how well a cuttlefish matches its background. The core computation is dimensionality reduction. A natural image lives in a very high-dimensional space. Dimensionality reduction projects that space down to the features that matter for detection. This shows how basic linear algebra concepts drive a real perceptual task. With the right low-dimensional representation, viewers recognize camouflage within reasonable times. Then they introduced claymation to let them experience this neural computation by hand. A clay cuttlefish keeps only what matters for blending in: color, shape, and contrast against the scene. Each group built several cuttlefish in different colors and a coral or rock reef scene, then devised a short story in which a cuttlefish moves through its habitat and makes different camouflage choices. Finally, each group presented their movie to the group.