Reservoir Computing with Quantum Memristors

In this master’s thesis, a quantum photonic reservoir computing framework was developed by adapting Michele Spagnolo et al.'s photonic quantum memristor. The design we propose improves memory and nonlinear dynamics by incorporating an additional optical mode and employing dual-rail encoding to introduce the injection of previous outputs. Starting with single memristors, the system is scaled through spatial multiplexing with random masks and varying configurations for each unit. This approach enhances short-term memory, leading to better accuracy for nonlinear tasks such as forecasting the Lorenz system. Simulations indicate that the system achieves high accuracy with fewer training data points compared to traditional methods. Future work should address the linear scaling limitation of independent units by designing photonic systems that exploit the exponential growth of quantum internal degrees of freedom.