Should HOT be a modified instance of KNN? Cos I feel the only line that should be modified is [the similarity score computation](https://github.com/mangaki/zero/blob/master/zero/knn.py#L58) and we can use [SideInformation](https://github.com/mangaki/zero/blob/master/zero/side.py) or similar to load embeddings (like [tags in LASSO](https://github.com/mangaki/zero/blob/master/zero/lasso.py#L26)).
Should HOT be a modified instance of KNN?
Cos I feel the only line that should be modified is the similarity score computation and we can use SideInformation or similar to load embeddings (like tags in LASSO).