Episodic Memory and Information Recognition Using Solitonic Neural Networks Based on Photorefractive Plasticity

Neuromorphic models are proving capable of performing complex machine learning tasks, overcoming the structural limitations imposed by software algorithms and electronic architectures. Recently, both supervised unsupervised learnings were obtained in photonic neurons means spatial-soliton-waveguide X-junctions. This paper investigates behavior networks based on these solitonic neurons, which tasks such as bit-to-bit information memorization recognition. By exploiting photorefractive nonlinearity if it a biological neuroplasticity, network modifies adapts to incoming signals, memorizing recognizing them (photorefractive plasticity). The processing storage result plastic modification interconnections. Theoretical description numerical simulation reported applied 4-bit information.