lkpy
lkpy copied to clipboard
lenskit
Python recommendation toolkit
When LensKit is working in parallel (e.g. `batch.recommend`), and the OOM killer takes out a worker, the parent LensKit process will (sometimes) hang instead of terminating. We should detect this...
We need to support new user ratings across LensKit, at least for every algorithm where it makes sense. - [x] Bias - [x] Item-Item - [x] UnratedItemsCandidateSelector - [x] Popular...
Add bridging support to [Spotlight](https://github.com/maciejkula/spotlight).
Add PyTorch implementations of key matrix factorization algorithms.
I'm experiencing parallel lockups running the LensKit tests in WSL2 (Debian, Anaconda with Python 3.8).
Even though there are [questions about the usefulness](https://dl.acm.org/doi/10.1145/3383313.3412488) of Neural CF's MLP, now that we have the TensorFlow models, it shouldn't be hard to add this for comparison purposes.
It would be useful to be able to do sigmoid transformations for rating predictions. Two possible pieces: - Build this in to some algorithms, like the TensorFlow `BiasedMF`. - Provide...
We need the ability to return *clamped* rating predictions: they are restricted to be in the range of allowable ratings. I think the easiest way to do this is likely...
We should implement a maximum marginal relevance recommender, in part to make it available, and in part so that an out-of-the-box LensKit component violates the assumption that filtering recommendation lists...
We should support the `partial_fit` API to update recommender models when it makes sense - [ ] ALS (#60) - [ ] Bias models - [ ] Popularity model -...