Computational Neuroscience

Computational neuroscience offers a new way to understand the brain by leveraging advances in computing, artificial intelligence, applied mathematics, statistics and biophysics. The lab's research in computational neuroscience focuses on:

  1. Predictive coding and Bayesian brain models

  2. Models for perceptual and social decision making based on partially observable Markov decision processes (POMDPs)

  3. Hierarchical recurrent neural networks implementing the above models

Brain-Computer Interfaces

Brain-computer interfaces (BCIs) are devices that connect brains directly to computers. BCIs can both record from the brain ("read" the brain) and stimulate the brain ("write" the brain). BCIs be used to help those who have impairments due to injury or neurological conditions. For example, by recording electrical signals from the brain, a BCI (1) can decode these signals into intention, such as the intention to move, and (2) actuate this intention by performing some output, such as moving a prosthetic arm, or turning on a light. Our lab works closely with neuroscientists, neurosurgeons and patients to explore novel methods to improve BCI technology. Current research projects include:

  1. Brain co-processors: This project utilizes artificial intelligence to adaptively deliver stimulation and compute control signals as a function of the brain's ongoing neural activity and external sensory signals.

  2. Naturalistic BCIs: This project uses deep learning methods to improve decoding of brain signals in naturalistic settings.

  3. Decoding pain and mood: This project seeks to identify neural biomarkers for pain and mood.

  4. Modeling electrical stimulation: This project models the effects of therapeutic electrical stimulation.

Artificial Intelligence

The lab focuses on neurally-inspired artificial intelligence (AI). Research projects include:

  1. Computer vision using dynamic hierarchical predictive coding

  2. Hypernetworks for multimodal video recognition

  3. Planning as inference and reinforcement learning

  4. Using AI and online human interactions to understand moral decision making

  5. Using AI to analyze the Indus script and art