What if Charles Darwin Built a Neural Network?

An alternative way to “train” a neural network with evolution

At the moment, backpropagation is nearly the only way for neural network training. It computes the gradient of the loss function with respect to the weights of the network and consequently updates the weights by backpropagating the error across the network layers using the chain rule.

Recently, the community has been trying to go beyond backpropagation, as, maestro Geoffrey Hinton himself, who popularized the …

backpropagation evolution genetic-algorithm gradient-descent network neural network neural networks