I strive for open science and open source software in particular. I like contributing to software that I use regularly (that’s my way of thanking) and creating my own software. Here you can find some highlights of my contributions on GitHub. For an exhaustive list, you can check out my GitHub profile.
Created software
pyts: GitHub, Documentation
I created and maintain pyts, which is a Python package dedicated to time series classification. It aims to make time series classification easily accessible by providing preprocessing tools, dataset loading utilities and implementations of many algorithms.
Other software
scikit-learn: GitHub, Documentation
-
PR #14999: I changed the use of the random seeds in sklearn.ensemble.HistGradientBoostingClassifier and sklearn.ensemble.HistGradientBoostingRegressor, so that the same seed is used when warm starting is used (this way, the training and validation sets are always identical, which prevents data leakage).
-
PR #14516: I changed the default behavior for early stopping in sklearn.ensemble.HistGradientBoostingClassifier and sklearn.ensemble.HistGradientBoostingRegressor, which is now ‘auto’: by default, early stopping is enabled if there are at least 10,000 samples.
-
PR #14012: I added a
warm_startparameter to sklearn.ensemble.HistGradientBoostingClassifier and sklearn.ensemble.HistGradientBoostingRegressor, allowing users to reuse the solution of the previous call to fit and add more estimators to the ensemble. -
Typo fixes: PR #16738, PR #16379, PR #16106, PR #15280, PR #14245, PR #14130, PR #14073, PR #14072, PR #13932.
tslearn: GitHub - Documentation
-
PR #216: I added a contributing guide in the documentation to help new contributors.
-
PR #204: I updated a couple of examples in the documentation about dynamic time warping.
-
PR #104: I updated the functions to compute dynamic time warping and the constraint regions, replacing the Cython functions with numba functions.
numpy: GitHub - Documentation
scipy: GitHub - Documentation
-
PR #10435: I fixed a bug which prevented the use of the Yeo-Johnson transformation for integer data type.
-
Typo fixes: PR #9570.