When are feedforward microcircuits well-modeled by maximum entropy methods?

Describing the collective activity of neural populations is a daunting task: the number of possible patterns grows exponentially with the number of cells, resulting in practically unlimited complexity. Recent empirical studies, however, find that the activity patterns of some, but not all, circuits are well described by maximum entropy models that incorporate only pairwise interactions. Why are such models successful in some cases but not others? Here, we study the ability of pairwise maximum entropy models to capture the activity patterns of feedforward circuits with different architectures and inputs. Responses to bimodal input signals shared by all circuit elements deviated substantially from maximum entropy predictions, while responses of unimodal inputs, regardless of connectivity, did not. Responses of circuits based on the measured synaptic input and dynamics of spike generation of retinal ganglion cells were well described by maximum entropy statistics across a broad range of ...

• Type: Journal
• Year: 2010
• Where: CORR
• Authors: Andrea K. Barreiro, Julijana Gjorgjieva, Fred Rieke, Eric Shea-Brown
• Keywords: Activity Patterns, CORR 2010, Education, Maximum Entropy, Maximum Entropy Models