reservoir computing

Reservoir computing is a neural network framework where a fixed, randomly connected reservoir of nodes transforms input signals into a high-dimensional state space, with only the readout layer trained. Recent research highlighted on WindowsForum.com demonstrates a biohybrid implementation using living rat cortical cultures as the reservoir, trained via closed-loop electrical stimulation and FORCE learning to generate target waveforms. This work, published in PNAS, represents a step toward brain-machine interfaces and biohybrid computing, showing how reservoir computing can bridge artificial and biological systems.