Sacha Morin
Sacha Morin
A few additions to a very useful module!
We have this in `scripts/run_real_example.py` but it includes a lot of boilerplate to reproduce all the tables from the paper. I suggest we design a simpler example (e.g., with only...
Users should be able to specify which parameters to start optimizing from. StepMix can already be provided with parameters (see examples in `datasets.py`), but we need to make sure they...
We should document common issues when estimating models in a specific doc section and suggest solutions. A good example is when running into division by 0 errors while fitting a...
Current categorical model one-hot encodes integer data every time we call the E or M step in the optimisation loop. We could probably obtain a meaningful speedup if we cached...
Running UMAP score method on train split before running on test split leads to obviously wrong test embeddings. (This was hot fixed for the paper by scoring test before train...
A lot of pairs have 0 distance, but aren't exactly equal. Need to remove duplicates given a small tolerance.