river
river copied to clipboard
online-ml
🌊 Online machine learning in Python
- stream-learn has a nice `StreamGenerator` class that can generate different kinds of concept drift, see [here](https://stream-learn.readthedocs.io/en/latest/streams_guide.html). Also see [their paper](https://arxiv.org/pdf/2001.11077.pdf), which describes each type of drift. It would be...
- Context in [this](https://github.com/online-ml/river/discussions/706#discussioncomment-1349077) discussion - [Kaggle page](https://www.kaggle.com/katehighnam/beth-dataset) - The URL of the dataset is [https://data.hpc.imperial.ac.uk/resolve/?doi=9422&file=4&access=](https://data.hpc.imperial.ac.uk/resolve/?doi=9422&file=4&access=)
- [Extreme F-Measure Maximization using Sparse Probability Estimates](http://proceedings.mlr.press/v48/jasinska16.pdf) - [Probabilistic Label Trees for Extreme Multi-label Classification](https://arxiv.org/pdf/2009.11218.pdf) ### Filter trees Filter trees are a way to reduce multi-class classification to binary...
A few things: - Ideally, we want to remove the time_series.evaluate submodule in favor of the evaluate module - We need an example where we tune the parameters of SNARIMAX...
scikit-learn [added](https://scikit-learn.org/stable/auto_examples/release_highlights/plot_release_highlights_0_24_0.html#new-poisson-splitting-criterion-for-decisiontreeregressor) a new splitting criterion for trees that supposedly works better for count data. Let's explore and see if this works for us.
```python import datetime as dt from river import compose from river import datasets from river import feature_extraction from river import linear_model from river import metrics from river import preprocessing from...
- We now have an `active` module for online active learning. - Right now it contains a single strategy based on entropy. - The goal is now to explore different...
[https://github.com/online-ml/river/issues/460](https://github.com/online-ml/river/issues/460)