Synergistic Information in Simulated Neural Networks: Effects of Connectivity and Comparison to Cortical Data

The brain is a vast neural network made up of approximately 100 billion neurons that continuously exchange information. This flow of information can be quantized through its probability. When multiple neurons send signals to a common target, they share a mutual information that can be decomposed into unique, redundant, and synergistic components. Using simulated neural networks, it was investigated how changes in network connectivity influence synergistic information relative to both the receiving neuron and to the total mutual information. These results are then compared with findings from Sherrill et al. (2021), who demonstrated in organotypic cortical cultures that synergistic information is enhanced downstream of recurrent information flow. Building on this, computational modeling enables a more precise examination of the parameters driving this effect, offering insights that are difficult to obtain in organic experiments.