Negatively correlated firing: the functional meaning of lateral inhibition within cortical columns

Abstract

Lateral inhibition is a well documented aspect of neural architecture in the main sensory systems. Existing accounts of lateral inhibition focus on its role in sharpening distinctions between inputs that are closely related. However, these accounts fail to explain the functional role of inhibition in cortical columns, such as those in V1, where neurons have similar response properties. In this paper, we outline a model of position tracking using cortical columns of integrate-and-fire neurons which respond optimally to a particular location, to show that negatively-correlated firing patterns arise from lateral inhibition in cortical columns, and that this provides a clear benefit for population coding in terms of stability, accuracy, estimation time and neural resources.

A novel approach to detect hot-spots in large-scale multivariate data

“…it is easier to detect the changes in spatio-temporal patterns if the system is negatively correlated. This also supports one of our long-term working hypotheses that brain neural networks are negatively correlated [ 10 , 11 ] and which we have recently confirmed using multi-array recording approaches in sensory systems [ 12 ].…”

“…It has been applied successfully to both artificially generated data and real biological data from multi-array electrophysiology, gene array and fMRI experiments. The algorithm is particularly effective in identifying changes in spatio-temporal patterns where variable activity is negatively correlated such as we have found recently using multi-array recordings from brain sensory networks [ 11 , 12 ].…”

Enhancing the Resolution of a Sensor Via Negative Correlation: A Biologically Inspired Approach

“…If à ¼ const, as was proposed in Ref. [7], then the ISIs generated by model (1) are strongly negatively correlated.…”

“…When the measurements are statistically independent the uncertainty (i.e., standard deviation) of the population mean scales as 1= ffiffiffiffi N p , or, equivalently, as T À0:5 ob where T ob ¼ Nt is the total observation time of the measurements and t is the sampling period. However, it has been noted that if negative correlations between the individual measurements exist, then the uncertainty can be reduced as 1=N where > 0:5 [1]. This suggests that, potentially, the resolution of a measurement can be improved by the introduction of negative correlation.…”

Suprathreshold stochastic resonance in neural processing tuned by correlation

“…Our results are in broad agreement with previous theoretical studies on correlation in neural systems, which have demonstrated that even small positive correlation results in reduced information transmission [43], to the extent that population coding is redundant beyond units of at most a few hundred neurons [45,46], while negative correlation results in increased information transmission in the context of traditional SR [19]. The benefits of negative correlation exist not only for presynaptic firing, but also extend to neurons within the model itself, which may occur due to mutual inhibitory connections [47].…”