Emergence of a stochastic resonance in machine learning

Department of Physics, Arizona State University, Tempe, Arizona 85287, USA (Dated: November 21, 2022) Can noise be beneficial to machine-learning prediction of chaotic systems? Utilizing reservoir computers as a paradigm, we find that injecting noise to the training data can induce a stochastic resonance with significant benefits to both short-term prediction of the state variables and longterm prediction of the attractor of the system. A key to inducing the stochastic resonance is to include the amplitude of the noise in the set of hyperparameters for optimization. By so doing, the prediction accuracy, stability and horizon can be dramatically improved. The stochastic resonance phenomenon is demonstrated using two prototypical high-dimensional chaotic systems. The interplay between noise and nonlinear dynamics have revealed that, if the hyperparameters are often leads to surprising phenomena with potentially significant not optimized, noise in the training data can improve applications and thus has always been an active to certain extent the prediction performance.