Computational Neuroscience InitiativeBRB 251 and via Zoom
Adam Calhoun
Princeton Neuroscience Institute
Measuring behavior to identify computations in the brain
Animals making decisions must take into account what their external environment is telling them as well as their internal needs. I have developed a new method to quantify behavior by identifying novel internal states, and have applied it to a dynamic social interaction (Drosophila courtship). Using this method, I show how the animal integrates sensory information in distinct ways depending on internal needs, then identify the role of neurons in either integrating incoming information or directing the production of motor outputs, roles that were previously hidden. Additionally, I have developed a new method to train deep neural networks on purely behavioral data in order to create a 1-1 mapping between model neurons and classes of biological neurons. This allows me to identify the neural-level representation the animal is using to process stimuli during natural behavior, how the state of these neurons guide behavior, and provides a framework to decipher the role of individual neurons in complex behaviors that rely on population neural activity.
My results reveal how animals compose behavior from previously unidentified internal states, a necessary step for quantitative descriptions of animal behavior that link environmental cues, internal needs, neuronal activity, and motor outputs.