The role of heterogeneity in reservoir computing dynamics
Date:
Poster presentation about heterogeneity in reservoir computing.
Abstract: Reservoir computers (RCs) are a type of artificial neural network with fixed weights that excel at time series forecasting tasks. Moreover, RCs provide a biologically plausible model for understanding information processing in neural and other biological networks. However, the role of heterogeneity in RC network connectivity—such as variations in neuron degree (number of connections)—remains poorly understood. Here, we refine a simple mean-field approximation of RC dynamics to explicitly account for degree heterogeneity. This refinement offers insight into how heterogeneity affects average neuron activation and the onset of oscillatory dynamics, including in networks with Erdős–Rényi-like and power-law degree distributions. Our analysis shows that heterogeneity shifts the region of hyperparameter space where non-saturated neuron activation occurs, as well as regions of good task performance, providing a framework for further understanding the role of heterogeneity in biological information processing networks.