Joel Zylberberg, Eric Shea-Brown
While recent experiments with relatively large neural populations show significant beyond-pairwise, or {\it higher-order} correlations (HOC), the impact of HOC on the network's ability to encode information is poorly understood. We investigate how the biophysical properties of neurons in networks shape HOC, and how HOC affect population coding. Specifically, we show that input nonlinearities similar to those observed in physiology experiments are equivalent to beyond-pairwise interactions in spin-glass-type statistical models. We then discuss one such model with parameterized pairwise- and higher-order interactions, revealing conditions under which beyond-pairwise interactions increase the mutual information between a given stimulus type and the population responses. For jointly Gaussian stimuli, coding performance is improved by shaping output HOC via input nonlinearities when neural firing rates are constrained to be sufficiently low. For natural image stimuli, performance improves for a broader range of firing rates. Our work suggests surprising connections between single-neuron biophysics, population activity statistics, and normative theories of population coding.
View original:
http://arxiv.org/abs/1212.3549
No comments:
Post a Comment