DeepCCA

DeepCCA copied to clipboard

Model unstable when using transformation

2

Hello @Michaelvll, @arminarj I am using the DeepCCA objective in a variational autoencoder setting. I use U, D, Vh = torch.linalg.svd(Tval) so i can getthe U and Vh together with...

hilmar05

Calculation of trace norm when using all singular values

10

The order of operations when using all singular values is currently sqrt(trace(T'T)) - note this is not the case when using topK. Equation (10) in Andrew's original paper due to...

jameschapman19

Hi @Michaelvll I want to use this approach and code for image-text clustering. could you help me please?

fatimaghafouri

using all singular values

2

adding @jameschapman19 suggestions to the loss, plus zero checking for torch.sqrt(0) nan values. Dear @Michaelvll please double-check it before merging it. Armin

arminarj

Hi, thanks for sharing the DCCA implementation which I found really helpful! Just want to report an issue about moving data tensor to device, e.g. lines like [this one](https://github.com/Michaelvll/DeepCCA/blob/a9ca478bc455e11e87a4a8d5e4071895e3edaf5c/main.py#L55), that...

syang11

RuntimeError: symeig_cuda: The algorithm failed to converge because the input matrix is ill-conditioned or has too many repeated eigenvalues (error code: 1807).

1

Hi, I used the dcca loss to train my dnn and got the following error message: 46 [D1, V1] = torch.symeig(SigmaHat11, eigenvectors=True) 47 [D2, V2] = torch.symeig(SigmaHat22, eigenvectors=True) 48 #...

guomanshan

Hi, I just want to use deepcca to build a multi class classification model. Our model consists three component, that is, embedding layer, CCA layer and the fully connected layer...

zqudm