Probabilistic Models of Interactions between Neurons

How interactions between neurons give rise to behavior is one of the central questions of systems neuroscience. Here we present several probabilistic models of interactions between neurons and methods for estimating interactions from simultaneous multi-electrode recordings. Probabilistic, model-based approaches differ from many models of interactions between neurons in that they are highly phenomenological. Rather than modeling the detailed biophysics of synaptic transmission and dendritic integration, we ask the neural encoding problem---how well can we predict a neuron's spiking given the activity of other observed neurons. Here we apply this statistical modeling approach to data from a variety of brain areas including visual, motor, auditory, and sensory cortices as well as hippocampus. In each of these areas we examine the structure of interactions between neurons, the relationship between neural interactions and tuning properties, as well as, several sensory adaptation phenomena. Probabilistic models of interactions between neurons improve spike prediction accuracy, create novel interpretations of existing electrophysiological data, and may provide new insight into how the brain represents and processes information.