Knowledge-Based Systems

Photonic crystal fiber (PCF) inverse design remains challenging because candidate geometries must satisfy coupled optical targets under expensive electromagnetic simulation. Existing pipelines improve surrogate prediction or one-shot parameter recommendation, but they do not accumulate reusable design knowledge across iterative trials. We formulate PCF inverse design as a memory-policy learning p…

Deep Echo State Networks (Deep ESNs) leverage hierarchical representations to capture multiple temporal scales efficiently. However, we show that the sequential application of randomized nonlinear mappings accumulates collinear features, deteriorates the condition number of the readout, and amplifies estimator variance. The standard closed-form solution of Deep ESNs based on ridge regression fail…