river
river copied to clipboard
🌊 Online machine learning in Python
This PR implements a first version of the Text Compression Classifier algorithm, discussed on issue #1291.
[GloVe](https://dsnotes.com/post/fast-parallel-async-adagrad/) and [Skip-gram](https://aegis4048.github.io/optimize_computational_efficiency_of_skip-gram_with_negative_sampling) models should be doable online. - [Incremental Skip-gram Model with Negative Sampling](https://aclanthology.org/D17-1037.pdf)
- Reproducing [this](https://arxiv.org/pdf/2209.02009.pdf) paper would provide a lot of learnings. In particular, it would open the door to real-time stock trading via online optimization. - The authors use energy prices...
I noticed the Hyperparameter Tuning portion of RiverML is marked `ToDo`. With the exception of `SuccessiveHalving`, does anyone have any libraries that are easily compatible with River? I found `SpotPython`...
It would be nice to only have classes and functions proposed when auto-completing
- It would be called `PerOutputRegressor` - Much like https://scikit-learn.org/stable/modules/generated/sklearn.multioutput.MultiOutputRegressor.html
- [https://scikit-learn.org/stable/modules/generated/sklearn.naive_bayes.CategoricalNB.html](https://scikit-learn.org/stable/modules/generated/sklearn.naive_bayes.CategoricalNB.html) - [https://github.com/online-ml/river/discussions/774](https://github.com/online-ml/river/discussions/774)
- https://github.com/Nixtla/statsforecast#theta-family - https://www.e3s-conferences.org/articles/e3sconf/pdf/2019/10/e3sconf_pe2019_01004.pdf First question: can it be done online? Second question: can we reuse seasonality and trend logic we've already implemented for SNARIMAX?