Deep Learning for Cognitive Neuroscience

Abstract

Neural network models can now recognise images, understand text, translate languages, and play many human games at human or superhuman levels. Deep learning allows us to scale up from principles and circuit models to end-to-end trainable models capable of performing complex tasks.

Questions this source addresses

• How can deep neural network models serve as testable, end-to-end trainable implementations of cognitive theories?
• How do the computational requirements that different tasks place on neural networks parallel cognitive load in human learners?
• In what ways can insights from deep learning models inform the design of instructional and learning technology systems?
• What does scaling from principles and circuit models to complex end-to-end trainable systems reveal about how the brain manages information processing demands?