Over the last decades, the combination of quantum computing and machine learning has opened many possibilities, for example enhancing machine learning algorithms through quantum platforms. However, one of the current challenges consists in combining the linear unitary evolution of closed quantum systems with the nonlinearity required by neural networks, which are currently the most widely used and versatile machine learning algorithms. This issue can now be addressed by a novel photonic tool, the quantum memristor,1 which displays a nonlinear behavior, while preserving quantum coherence, through a weak controlled interaction of its input state with the environment. Here, we show how its operation can be extended to deal with higher frequency modulations of the input and, possibly, with a simplification in its scheme. This method can prove useful for the future implementation of memristor-based quantum neural networks.
|