Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 7-Day Trial for You or Your Team.

Learn More →

Neural correlations, population coding and computation

Neural correlations, population coding and computation Correlations among neurons can affect both the amount of information encoded in a population and strategies for decoding the population. These two issues — encoding and decoding — lead to complementary perspectives about the role of correlations. In the encoding perspective, the information encoded in a population of correlated neurons is compared with the information that would be encoded if the population were uncorrelated. In the decoding perspective, the amount of information lost if correlations are ignored when decoding is measured. Note that the decoding perspective is much more subtle than the encoding perspective — it asks whether a potentially suboptimal strategy, ignoring correlations, really is suboptimal, and, if so, just how bad it is. If we knew only that neural responses were correlated, we would not know whether or not those correlations affected information encoding, nor would we know whether or not they affected decoding strategies. Furthermore, correlations can increase, decrease or not affect the amount of information encoded, just as they can affect or not affect the amount of information extracted using a decoder that ignores correlations. As a corollary to the previous point, the information present in neural responses, as well as the change in information due to attentional or learning-related factors, cannot be estimated by single neuron recordings. At the level of pairs of neurons, the measured effects of correlations on encoding and decoding have been small (in all but one study less than ∼10%) across many brain areas and species. Correlations can have a large effect at the population level even when they have a small effect at the level of pairs. Consequently, results obtained for pairs of neurons cannot be directly extrapolated to populations, a fact that is true for both encoding and decoding. http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Nature Reviews Neuroscience Springer Journals

Neural correlations, population coding and computation

Download PDF
9 pages

Loading next page...
 
/lp/springer-journals/neural-correlations-population-coding-and-computation-k0BU7YDLlj

References (86)

Publisher
Springer Journals
Copyright
Copyright © 2006 by Nature Publishing Group
Subject
Biomedicine; Biomedicine, general; Neurosciences; Behavioral Sciences; Biological Techniques; Neurobiology; Animal Genetics and Genomics
ISSN
1471-003X
eISSN
1471-0048
DOI
10.1038/nrn1888
Publisher site
See Article on Publisher Site

Abstract

Correlations among neurons can affect both the amount of information encoded in a population and strategies for decoding the population. These two issues — encoding and decoding — lead to complementary perspectives about the role of correlations. In the encoding perspective, the information encoded in a population of correlated neurons is compared with the information that would be encoded if the population were uncorrelated. In the decoding perspective, the amount of information lost if correlations are ignored when decoding is measured. Note that the decoding perspective is much more subtle than the encoding perspective — it asks whether a potentially suboptimal strategy, ignoring correlations, really is suboptimal, and, if so, just how bad it is. If we knew only that neural responses were correlated, we would not know whether or not those correlations affected information encoding, nor would we know whether or not they affected decoding strategies. Furthermore, correlations can increase, decrease or not affect the amount of information encoded, just as they can affect or not affect the amount of information extracted using a decoder that ignores correlations. As a corollary to the previous point, the information present in neural responses, as well as the change in information due to attentional or learning-related factors, cannot be estimated by single neuron recordings. At the level of pairs of neurons, the measured effects of correlations on encoding and decoding have been small (in all but one study less than ∼10%) across many brain areas and species. Correlations can have a large effect at the population level even when they have a small effect at the level of pairs. Consequently, results obtained for pairs of neurons cannot be directly extrapolated to populations, a fact that is true for both encoding and decoding.

Journal

Nature Reviews Neuroscience – Springer Journals

Published: May 1, 2006

There are no references for this article.