Stochastically Structured Reservoir Computers for Financial and Economic System Identification

Abstract

This paper introduces a methodology for identifying and simulating financial and economic systems using stochastically structured reservoir computers (SSRCs). The framework combines structure-preserving embeddings with graph-informed coupling matrices to model inter-agent dynamics while enhancing interpretability. A constrained optimization scheme guarantees compliance with both stochastic and structural constraints. Two empirical case studies, a nonlinear stochastic dynamic model and regional inflation network dynamics, demonstrate the effectiveness of the approach in capturing complex nonlinear patterns and enabling interpretable predictive analysis under uncertainty.