jku-vds-lab/paradime

Implement low-dimensional transform for t-SNE

Closed this issue · 2 comments

See title. Check diagonal of relations matrix and transform with t-distribution.

The low-dimensional transform uses the alpha parameter, which can be made learnable. Think of a way how to allow learnable parameters either through the transform, or more likely inside the ParametricDR class.

Learnable parameters are possible by default. Users can simply set alpha=model.alpha, where model.alpha is a registered PyTorch parameter of the model.