Code Vectorization in custom layer
4 vues (au cours des 30 derniers jours)
Afficher commentaires plus anciens
Hi, we are designing a custom layer where we need to calculate the back-derivative from a 4D matrix
Here is a simple way using for loop to implement it
X = zeros(2,2,2,2);
X([1 5 7 10 12 14 16]) = rand(7,1);
kernelsize=5;
A=cell(2,1);
A{1}=rand(2,5);
A{2}=rand(2,5);
f=cell(2,1);
f{1}=rand(2,1);
f{2}=rand(2,1);
k = find(X);
[ii, jj, kk, ll] = ind2sub( size(X), k);
Z=zeros(size(X));
dLdW=zeros(2,5,2);
for j=1:kernelsize
for i=1:length(k)
Z(k(i))=X(k(i))*dot(A{jj(i)}(:,j),f{jj(i)});
end
sol=sum(Z,2);
dLdW(:,j,:)=sum(sol,4);
Z=zeros(size(X));
end
Is there a way to not use for loop? Because I want to use GPU to train it.
0 commentaires
Réponse acceptée
Joss Knight
le 15 Avr 2018
Adotf = cellfun(@(aa,ff)ff.'*aa, A, f, 'UniformOutput', false);
Adotf = cat(1, Adotf{:});
Z = X(k).*Adotf(jj,:);
j = repmat(1:kernelsize, numel(ii), 1);
ii = repmat(ii, 1, kernelsize);
kk = repmat(kk, 1, kernelsize);
dLdW = accumarray([ii(:), j(:), kk(:)], Z(:), [size(X,1) kernelsize, size(X,3)]);
Are all the A matrices and f vectors the same size? Because if so you shouldn't use a cell array, you should concatenate in dim 3 and use pagefun instead of cellfun (if you're using gpuArray).
A = cat(3, A{:});
f = cat(2, f{:});
f = shiftdim(f, -1);
Adotf = pagefun(@mtimes, f, A);
Adotf = permute(Adotf, [3 2 1]);
Z = X(k).*Adotf(jj,:);
j = repmat(1:kernelsize, numel(ii), 1);
ii = repmat(ii, 1, kernelsize);
kk = repmat(kk, 1, kernelsize);
dLdW = accumarray([ii(:), j(:), kk(:)], Z(:), [size(X,1) kernelsize, size(X,3)]);
Plus de réponses (0)
Voir également
Catégories
En savoir plus sur Matrices and Arrays dans Help Center et File Exchange
Community Treasure Hunt
Find the treasures in MATLAB Central and discover how the community can help you!
Start Hunting!