Machine Learning: Data Analysis Project Presentation

  • Ph.D. Student
  • Joint Ph.D. Program in Neural Computation/Machine Learning
  • Machine Learning Department
Project Presentations

Strategies for resolving neural redundancy

Millions of neurons in the brain drive the activity of hundreds of muscles, meaning many different neural activity patterns could generate the same movement. How does the brain resolve this neural redundancy? To address this question, we leveraged a brain-computer interface paradigm which allowed us to define precisely which neural activity patterns were behaviorally equivalent. We trained Rhesus monkeys to move computer cursors by modulating neural activity in primary motor cortex. We then predicted the distribution of behaviorally equivalent activity under different hypotheses by finding solutions to various optimization problems. Minimal energy principles, inspired by work on muscular redundancy, did not accurately predict this activity. Instead, we found that behaviorally equivalent activity was well predicted by a strategy in which the animal selects any activity pattern from a fixed repertoire of patterns that produce the desired movement. Our findings suggest that neural redundancy is resolved differently than muscular redundancy.

DAP Committee:
Steve Chase, Byron Yu, Rob Kass

For More Information, Please Contact: