river icon indicating copy to clipboard operation
river copied to clipboard

🌊 Online machine learning in Python

Results 123 river issues
Sort by recently updated
recently updated
newest added

We've implemented `ensemble.AdaBoostClassifier`, but there's other variants to explore. Some interesting papers: - [Online Gradient Boosting](https://arxiv.org/pdf/1506.04820.pdf) - [Online Bagging and Boosting](https://ti.arc.nasa.gov/m/profile/oza/files/ozru01a.pdf) - [Optimal and Adaptive Algorithms for Online Boosting](http://proceedings.mlr.press/v37/beygelzimer15.pdf) See...

Needs researching

we now have a mutate method. But it’s not clear what parameters and attributes are mutable or not. I suggest we put this in the docs. An idea is to...

Documentation

- See [this](https://arxiv.org/pdf/1205.2618.pdf) paper. I'm not an expert on this but it should fit nicely into the `reco` module. - There’s an implementation in https://github.com/benfred/implicit

New feature

As of now there is a `SDFT` class in the `feature_extraction` module which computes an exact discrete Fourier transform. This isn't in the public API yet because it isn't a...

New feature

scikit-learn added `[PolynomialCountSketch](https://scikit-learn.org/stable/modules/generated/sklearn.kernel_approximation.PolynomialCountSketch.html#sklearn.kernel_approximation.PolynomialCountSketch)` in version 0.24. Looks like it can be adapted online.

New feature

See [this](http://proceedings.mlr.press/v51/chen16c.pdf) paper and [this](https://github.com/cchangyou/Santa/blob/master/FNN_CNN/algorithms/Santa.m) Matlab implementation by the authors. Looks promising!

New feature

- https://pythonhosted.org/mord/ ## Loss functions As anyone who reads this might know, there's classification and regression. There's a special case of regression that some people call "ordinal regression". This involves...

New feature

- Context: https://github.com/online-ml/river/issues/1353 - Goal would be to implement this in Rust.

New feature

Implement RRCF proposed by Guha et al. at ICML 2016 [http://proceedings.mlr.press/v48/guha16.pdf](http://proceedings.mlr.press/v48/guha16.pdf).

New feature

See [RS-Forest: A Rapid Density Estimator for Streaming Anomaly Detection](https://ieeexplore.ieee.org/document/7023377). The authors say that they are more efficient than half-space trees, it might be worth a try!

New feature