Source-linked AI summary
Prediction of spatio-temporal patterns of neural activity from pairwise correlations
Olivier Marre, Sami El Boustani, Yves Fregnac, Alain Destexhe
TL;DR
To predict spatio-temporal neural patterns, the paper combines a maximum-entropy model with a Markovian assumption to capture spatial and temporal pairwise correlations. It reports more accurate multi-bin predictions than Ising alternatives in simulated activity and an advantage over Ising for slow-wave-sleep recordings.
Problem
Existing maximum-entropy models focused on single-bin patterns, which were not suited to predicting temporal statistics of neural activity.
Source: p. 1Method
The authors use a maximum-entropy model with a Markovian assumption to incorporate spatial and temporal pairwise correlations.
Source: p. 1Results
For patterns of two or more bins, the Markov model has lower Jensen-Shannon divergence than Ising and independent models in Glauber data, and outperforms Ising for SWS recordings.
Source: pp. 3–4Takeaways & Limitations
The model predicts larger temporal patterns and can generate surrogate rasters that mimic the data’s spatial and temporal correlation structure.
Source: p. 5The approximation is stated to be valid in the weak temporal-correlation limit and assumes detailed balance for a stationary distribution.
Source: pp. 1–2
Abstract
from arXiv · showhide
We designed a model-based analysis to predict the occurrence of population patterns in distributed spiking activity. Using a maximum entropy principle with a Markovian assumption, we obtain a model that accounts for both spatial and temporal pairwise correlations among neurons. This model is tested on data generated with a Glauber spin-glass system and is shown to correctly predict the occurrence probabilities of spatio-temporal patterns significantly better than Ising models taking into account only pairwise correlations. This increase of predictability was also observed on experimental data recorded in parietal cortex during slow-wave sleep. This approach can also be used to generate surrogates that reproduce the spatial and temporal correlations of a given data set.