Optimox

@eduardocarvp something like this (from https://www.kaggle.com/davidedwards1/tabularmarch21-dae-starter) : ``` class SwapNoiseMasker(object): def __init__(self, probas): self.probas = torch.from_numpy(np.array(probas)) def apply(self, X): should_swap = torch.bernoulli(self.probas.to(X.device) * torch.ones((X.shape)).to(X.device)) corrupted_X = torch.where(should_swap == 1, X[torch.randperm(X.shape[0])],...

Thanks for your PR @plamenpasliev. The CI is broken for building the gpu so we'll need to fix this before hand. I did not have a close look at your...

Yes don't worry about the error regarding the gpu-docker-build, I'll fix it and you'll just have to rebase on develop. The test for your notebook seem to fail as well,...

Hello @askhan18, Basically the helper functions would facilitate #244. The very basic would just ask for the categorical column id and would generate the input for tabnet : `cat_dims`, `cat_emb_dims`...

You can always access the masks from attention (with the explain method) and try to cluster these with UMAP or TSNE but I'm not sure that's what you want ?

you would need to access the results before final mapping : https://github.com/dreamquark-ai/tabnet/blob/4fa545da50796f0d16f49d0cb476d5a30c2a27c1/pytorch_tabnet/tab_network.py#L480 But to be honest I think it would be better to use VAE if what you want to...

@ensonario do you have any plots to share ?

Hello @ensonario, - Yes each attention layer tends to 'specialize' in looking at specific features and totally ignoring the others. However, as you can see each mask for a same...

I'm not familiar with ONNX, but it would be quite easy to save the network as traced script (from pytorch jit: https://pytorch.org/docs/stable/jit.html), which could be used for inference without the...

@luigisaetta I think trying to ONNXify (whatever this is called) the entire class `TabNetClassifier` is doomed to fail - I'll be very surprised if you manage to export everything with...