The goal of this project is to develop efficient algorithms for mapping the neural connections (the connectome) of a rat cortex. This is a challenging task, as generating even 1/1000th of a connectome requires almost 2PetaBytes of image data. I will be working to develop fast computer vision algorithms to detect synapses in neural scans and find the “edges” between the neurons in the connectome. Previous attempts to graph sections of the brain have focused on pixel level accuracy in the image segmentation. For this project, the goal is accuracy at the neuron graph level, and so I will work on designing algorithms efficient enough to process large amounts of data while still achieving map-level accuracy.

My past research has been focused mainly on development of machine learning algorithms to be used to control robotic agents trained from human rewards. I have also worked in natural language processing algorithm development. I believe my background in machine learning will give me a good foundation to tackle the difficult problem of neural mapping.