Exploring Photonic Oscillators for Oscillatory Neural Networks

Oscillatory Neural Networks (ONNs) have emerged as a promising computational paradigm utilizing phase-based encoding being particularly attractive for tasks such as pattern recognition. Typically, ONNs have been widely investigated using a Kuramoto sinusoidal oscillator model. In contrast, this work investigates the ONN performance using a photonic oscillator described by the Yamada model for pattern retrieval tasks while comparing it against the Kuramoto model.We obtained the phase transition curve of two coupled Yamada oscillators, which exhibit two distinct states, in-phase and out-of-phase. These states are crucial for performing associative memory tasks. Additionally, we performed simulations on two different datasets under different
scenarios. The results indicate that the Yamada model achieves similarly high accuracies in pattern retrieval as the Kuramoto model, hence affirming the effectiveness of photonic laser neuron for ONN computing.