A01Learning perceptual representations in biological and artificial neural networks

 How do biological neuronal networks learn is a fundamental question in Neuroscience as well in brain-inspired implementations of artificial intelligence. We plan to address this problem by examining how visual cortical networks learn to process the mid-level visual feature represented by texture stimuli. Our hypothesis is that the cortical encoding of textures is not innate as it might be for luminance or contrast but it emerges through visual experience as a consequence of the importance of textures for the animal’s survival in the environment. Learning occurs via an unsupervised learning (RL) process which relies on the unique statistical properties of textures.