Correspondence between neuroevolution and gradient descent
This work connects neuroevolution and gradient descent, two typically distinct neural-network training methods, providing a theoretical bridge for researchers in machine learning.
The paper demonstrates analytically that neuroevolution, in the limit of small mutations, is equivalent to gradient descent on the loss function with Gaussian white noise, and this correspondence holds for finite mutations in shallow and deep networks through numerical simulations.
We show analytically that training a neural network by conditioned stochastic mutation or neuroevolution of its weights is equivalent, in the limit of small mutations, to gradient descent on the loss function in the presence of Gaussian white noise. Averaged over independent realizations of the learning process, neuroevolution is equivalent to gradient descent on the loss function. We use numerical simulation to show that this correspondence can be observed for finite mutations,for shallow and deep neural networks. Our results provide a connection between two families of neural-network training methods that are usually considered to be fundamentally different.