(yet another) Zygote: mutating arrays is not supported

[rakeshvar]

@Dhruva2 , One of the threads that @ChrisRackauckas is talking about is this one:

Gradient of Gradient in Zygote

I am trying to find the Hessian of a quadratic form. I know the answer. But if I apply gradient on the gradient, I get an error. Can Zygote do the differentiation only if the output is a scalar? But then how can it work for Neural Networks, where each layer can return a very big tensor? Here is an example… > n = 3 > A = reshape(0:(n^2-1), n, n) .% (n+1) 3×3 Array{Int64,2}: 0 3 2 1 0 3 2 1 0 > H = 2*A'*A 3×3 Array{Int64,2}: 10 4 6 4 20 12 6 12 26 > x1 = collect(0:(n-1)) …

I had a much more minimal working example there, and I ask the same question. (I do not even recollect what I was trying to do. May be I was just curious.) Chris gives a very comprehensive answer there. Do check it out.