Lecture
This lecture delves into machine learning models for neuroscience, exploring how these models bridge limitations in brain function understanding by linking perception-action coupling at the circuit level. The instructor explains task-driven modeling, showcasing how models are created to understand why neurons fire in response to stimuli. The lecture also covers core object recognition, focusing on convolutional neural networks trained on ImageNet, which are the best models for the ventral pathway in humans and monkeys. The instructor demonstrates how these models can explain neural activity in the primate brain, particularly in the IT area. Additionally, the lecture discusses the application of machine learning in understanding proprioception, showcasing how candidate models are created and tested to explain direction-selective neurons in the brain.