Computing That Serves

Colloquium: Structured Probabilistic Models for Understanding Neural Circuitry


Thursday, September 24, 2015 - 11:00am


Eric Jonas


David Wingate

Structured Probabilistic Models for Understanding Neural Circuitry 
Thursday, September 24, 2015
11:00 am 1170 TMCB

Recent experimental work has started generating dense schematics of neural systems -- "connectomics" -- with every cell and every synapse being measured and traced. However, extracting biological insight from these new data is a massive challenge.  It has long been assumed that discovering neuron types and their relation to microcircuitry is crucial to understanding neural function. Here we developed a nonparametric Bayesian technique that identifies neuron types and microcircuitry patterns in connectomics data. It combines the information traditionally used by biologists in a principled and probabilistically-coherent manner, including including connectivity, cell body location and the spatial distribution of synapses. We show that the approach recovers known neuron types in the retina and enables predictions of connectivity, better than simpler algorithms. It also can reveal interesting structure in the nervous system of C. elegans and a historic microprocessor. Ultimately we seek to use structured probabilistic models to understand a much wider range of neural and biological processes. 

Joint work with Konrad Kording. 


Eric Jonas is currently a postdoctoral scientist in computer science at UC Berkeley working with Ben Recht on compressive sensing and inverse problem theory for improving biological measurement. He earned his SB in Electrical Engineering and Computer Science, his SB in Neuroscience, his M. Eng. in Electrical Engineering, and his PhD in Neuroscience, all from MIT. In 2011 he founded Prior Knowledge, a predictive database company which was acquired by in 2012, where he ran machine learning through 2013.