search

hyperdimensional-computing / torchhd

Torchhd is a Python library for Hyperdimensional Computing and Vector Symbolic Architectures
https://torchhd.readthedocs.io
MIT License
229 stars 24 forks source link

intRVFL #98

Closed denkle closed 1 year ago

denkle commented 1 year ago

Created the first draft for intRVFL model that uses 'Abalone' and all other datasets from the collection that would eventually have to be included. I have left comments starting with "DK:" in multiple places to indicate my doubts. Hope we will figure them through the interaction.

mikeheddes commented 1 year ago

Great work! It will be very helpful to have an example of using the newly added datasets

denkle commented 1 year ago

@mikeheddes, I tried my best to implement your suggestions from the previous review. As usually, happy to continue iterating. It should be admitted that I have not paid detailed attention to the documentation. This would have to be double-checked once we settle down with the structure for the benchmark and its primitives

denkle commented 1 year ago

It looks great! Thank you for the amazing work :)

I will test the code locally this weekend and maybe make some minor refactoring changes and then merge it.

Sounds great!

mikeheddes commented 1 year ago

@denkle your code works like a charm! Very good work.

I decided to remove the classification module for now and just add the ridge regression function to the example. We can in a follow up PR try to extract the IntRVFL model to the models module which I think can also include the ridge regression function. I also made the EncodingDensityClipped a bit more general so it supports other VSA models and renamed it to just Density. Other than that I added some python typing and mostly made some small modifications.