Royi
Royi
Did you verify results of the Auto Grad are like in the appendix of the paper?
[It seems it hasn't been](https://github.com/CPJKU/deep_lda/issues/1#issuecomment-392250186).
@zorrocai , Singular Values are the Eigen Values of $ {A}^{T} A $ or $ A {A}^{T} $ (Which share the same eigen values). Hence the connection between the two...
Well, currently what I do is looping over the 3rd dimension of a pre allocated area. I'm after a non allocating solution beside the explicit loop.
I mean I'm doing something like: ```julia for ii in 1:3 @views map!(mK, mO[:, :, ii], extend(mI[:, :, ii], StaticKernels.ExtensionReplicate())); end ``` It still allocates few KB's, but works good....
Indeed, It seems to be a difference since I didn't wrap it as a function but only as a local block using `begin`. I don't understand what you mean in...
Is there way to have `valid` mode? Namely it will work only on valid indices where the kernel is well defined?
Think of a simple 3x3 kernel (`k`) over an image of 100x100 (`mI`). The valid part of the kernel is the inner 98x98. Let's say I want to apply this...
I will try to give a concrete example, from the discussion at https://discourse.julialang.org/t/92691: ```julia using BenchmarkTools; using LoopVectorization; using StaticKernels; function SolveLaplace!( mT, numItr ) numRows, numCols = size(mT); for...
In this context it is OK to work in place (Gauss Seidel vs. Jacobi iterations). I see, so the point is that in the case of `StaticKernels.jl` the compiler sees...