Project: Our perception of the world seems a coherent whole, yet it is built out of the activities of many thousands or even millions of neurons, and similarly for our memories, thoughts, and actions. It seems difficult to understand the emergence of behavioral and phenomenal coherence unless the underlying neural activity also is coherent. But if the neurons are not independent, how do we describe their collective activity? In this work, The PI takes several approaches to this problem. In each case the PI is reaching for a theoretical framework with a generality that transcends the details of particular systems, but in each case the work is grounded by intimate collaboration with several experimental groups. It is now conventional to assert that the emergence of new, larger data sets in the study of biological systems creates a need for new and more efficient methods of data management and analysis. Thus, it is widely expected that the BRAIN initiative, which will provide substantial resources to advance our ability to record simultaneously from many neurons, will also have a significant "computational" component related to the storage, handling, and analysis of the huge bodies of data that will be generated. It is much less widely appreciated that making sense of the collective behavior in systems with many interacting degrees of freedom requires theory, not just analysis. The goal of this work is to develop such a theoretical framework.
The PI's collaborators are using multi-electrode arrays to monitor activity in populations of ganglion cells in the vertebrate retina, as well as optical methods to monitor activity in mammalian cortical and hippocampal circuits, and in the whole brain of the nematode C elegans. The PI will use maximum entropy ideas, which have had some success in such problems, to build models of the joint distribution of activity across the hundreds of neurons in each of these experiments. The entropy of this distribution determines the capacity of the network to carry information, and the geometry of the distribution captures intuition about network function. Beyond exploring the distribution of activity, the PI will construct a "thermodynamics" for these networks that will allow to place real networks in a phase diagram of possible networks. Finally, the PI will turn to dynamical aspects of network behavior, focusing on the encoding of predictive information. The PI's research is well integrated with teaching at the graduate and undergraduate level, and there will be a special effort to create a "current topics" course that brings issues at the interface of physics and biology to the attention of a broader group of students. In addition, the PI is engaged is a series of public outreach efforts, which will culminate in a book for a general audience.
