t-SNE parameters

[ch3yu]

Hi,

Thank you for the interesting work. The idea of disentangling semantic-related and semantic irrelevant contexts is innovative.

I am trying to recreate the results presented in the paper, and I am wondering what are the parameters used for the t-SNE dimension reduction technique. Similar testing accuracies for both seen and unseen samples were achieved, but I wasn't able to get the sensible t-SNE visualization of z_x^r.

Thanks