bioRxiv Preprint Server
July 29, 2019
Nikita Agarwal, Aditi Kathpalia, Nithin Nagaraj
preprint
A novel measure called Network Causal Activity, based on Compression-Complexity Causality, was used to analyze electrocorticographic signals from the lateral cortex of four monkeys. Network Causal Activity was consistently higher in the awake state compared with the anaesthetized state, suggesting it may serve as a quantitative indicator of consciousness.
Heliyon
February 1, 2019
Mohit Virmani, Nithin Nagaraj
A new measure of brain network complexity, ΦC, is proposed that bridges Integrated Information Theory (IIT) and the Perturbational Complexity Index (PCI). ΦC uses a perturbation-based compression-complexity approach, is mathematically well bounded, has negligible current state dependence, and avoids combinatorial explosion because its atomic bipartitions scale linearly with the number of network nodes. Computations show ΦC produces a hierarchy similar to IIT's Φ for several multiple-node networks and reveals a rich interplay between differentiation, integration, and entropy. ΦC is a promising heuristic for characterizing network complexity and may contribute to building a measure of consciousness, with potential applications to neurophysiological data.
arXiv Preprint Archive
August 23, 2016
Mohit Virmani, Nithin Nagaraj
A new measure called Φ^C bridges Integrated Information Theory (IIT) and the Perturbational Complexity Index (PCI) by using lossless data compression to quantify integrated information in brain networks. Unlike IIT's Φ, which is computationally expensive and dependent on current state, Φ^C is mathematically well bounded, has negligible state dependence, and scales linearly with network nodes, avoiding combinatorial explosion. Computer simulations show Φ^C produces similar hierarchies to Φ across multiple-node networks and reveals interactions between differentiation, integration, and entropy. It offers a faster heuristic for measuring integrated information—and thus a potential proxy for consciousness—in larger networks like the human brain, enabling tests of brain complexity predictions on real neural data.