Invariance and Equivariance in Brains and Machines

Explore the intersection of neuroscience and artificial intelligence in this 53-minute lecture by UC Berkeley Professor Bruno Olshausen. Delve into the challenges of building machines that can perceive and act like humans and animals, and discover how neuroscience and AI research can inform each other. Learn about Olshausen's approach to creating invariant and equivariant representations in vision, rooted in animal behavior observations and informed by neurobiological mechanisms and mathematical principles. Examine the emerging neural circuit for factorization that can learn about shapes and their transformations from image data, as well as a model of the grid-cell system based on high-dimensional encodings of residue numbers. Gain insights into how these models provide efficient solutions to long-studied problems and their potential applications in neuromorphic hardware and forming hypotheses about visual and entorhinal cortex. Understand Olshausen's research focus on decoding the information processing strategies of the visual system for tasks like object recognition and scene analysis, and how this work aims to advance our understanding of the brain and develop new algorithms for image analysis based on biological vision systems.

Invariance and equivariance in brains and machines