Information-limiting correlations

By Rubén Moreno-Bote, Jeffrey M. Beck, Ingmar Kanitscheider, Xaq Pitkow, Peter E. Latham, Alexandre Pouget. 2014

530 citations recorded by the source corpus.

Computational strategies used by the brain strongly depend on the amount of information that can be stored in population activity, which in turn strongly depends on the pattern of noise correlations. In vivo, noise correlations tend to be positive and proportional to the similarity in tuning properties. Such correlations are thought to limit information, which has led to the suggestion that decorrelation increases information. In contrast, we found, analytically and numerically, that decorrelation does not imply an increase in information. Instead, the only information-limiting correlations are what we refer to as differential correlations: correlations proportional to the product of the derivatives of the tuning curves. Unfortunately, differential correlations are likely to be very small and buried under correlations that do not limit information, making them particularly difficult to detect. We found, however, that the effect of differential correlations on information can be detected with relatively simple decoders.

Subfields: Artificial Intelligence, Cognitive Neuroscience, Molecular Biology

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Cognitive Neuroscience

In Molecular Biology

Citations in the corpus

Re-evaluating circuit mechanisms underlying pattern separation (2019, positive): “Indeed, the precise statistical structure of correlated neural variability can have a strong impact on the overlap of distinct neural representations (Averbeck et al., 2006; Moreno-Bote et al., 2014).”
Neural population coding: combining insights from microscopic and mass signals (2015, methodology): “…can inﬂuence neural population coding in more complex way than described above: depending on the precise pattern of signal and noise correlations, and the choice of readout mechanisms, correlations can increase, decrease, or leave unaffected the information carried by a population [1,15,40,49–51].”
An Interference Model of Visual Working Memory (2017, methodology): “A more sophisticated alternative comes from modeling work by Moreno-Bote et al. (2014) who investigated the effect of correlated noise in biologically realistic models of neural networks.”
Correlations and Neuronal Population Information (2016, methodology): “Critically, although differential correlations cause information to saturate, their contribution to overall correlations may be negligible (Moreno-Bote et al. 2014, Kanitscheider et al. 2015b).”
A statistical framework for neuroimaging data analysis based on mutual information estimated via a gaussian copula (2016, background): “…et al., 2009; Reich et al., 2001], characterize different neurons [Sharpee, 2014] as well as quantify the effect of correlations between neurons [Ince et al., 2009, 2010; Moreno-Bote et al., 2014] and the importance of spike timing [Kayser et al., 2010; Nemenman et al., 2008; Panzeri et al., 2001].”
Towards the neural population doctrine. (2019, background): “In fact, a tremendous amount of research has led to the fundamental belief that correlations are linked to the information carrying capacity of the brain, which itself has widely been postulated to be directly related to the level of performance of an individual on a given task [13,14].”