Hyperbolical Embedding of Protein Sequences

This is my attempt to apply recent hierarchical data embedding methods to protein sequence labels (roles, attributes, classes, ...). There are a couple of references in one of the jupyter notebooks if anyone would like to dive deeper into the topic. Unfortunately this attempt was unsuccessful, both the training and validation losses increase forever the longer one trains. Further investigation is required to find the issues.