lkpy
lkpy copied to clipboard
lenskit
Python recommendation toolkit
Need to add Numba 0.56 to the support matrix. Dependencies: - lenskit/csr#35
The manual is largely organized by feature. We should provide pointers to make it easy to find how to do specific things, such as measure the popularity (#316) of a...
It would be useful to add an intra-list similarity metric for computing diversity to LensKit. As with #316, this metric will need additional data (the item vectors). Other metrics can...
It would be useful to add popularity metrics to LensKit, based on measures in the existing literature. Candidates: - popularity rank - popularity (# of ratings) - log popularity -...
neither pip or conda work to install the lenskit-implicit wrappers .... or at least I couldn't get them to work (wanted to compare BPR and implicit ALS)
We should split our parallel processing code out of LensKit and into a separate package, so LensKit doesn't carry around as much non-recommendation code.
It works well to use scikit-optimize to tune hyperparameters for LensKit algorithms. Now that Michael has it working, we need to add it to LensKit. - [ ] Dcoumentation on...
It would be nice to have mean similarity, instead of sum of similarities, as a scoring option for item-item, to de-emphasize popular items in implicit item-item.
Right now the item-item CF implementation combines both item-item k-NN, and computing the k-NN matrix from rating matrix rows (collaborative filtering). There is no reason these have to be tied...
We should add a bridge to [LightFM](https://github.com/lyst/lightfm).