Rate of Convergence of Polynomial Networks to Gaussian Processes

Adam Klukowski

We examine one-hidden-layer neural networks with random weights. It is well-known that in the limit of infinitely many neurons they simplify to Gaussian processes. For networks wit…