Lb-pidsim-train

Beta release v0.2.1 of the **lb-pidsim-train** package! 🔥

What's new?
- Bug fixing
- _Ask-and-Tell_ Hopaas configuration fixed avoiding multi-telling loss value at the end of the training
- Unexpected behavior with the optimization of the Muon models understood and solved

Beta release v0.2.0 of the **lb-pidsim-train** package! 🔥

What's new?
- Enabled hyperparameters optimization through [hopaas_client](https://github.com/landerlini/hopaas_client)
- Added a third independent player (named _classifier_) to the two-players game of the GAN system to use its loss for optimization studies
- Bug fixing
- Preprocessing issues, in particular the mismatch between the order of transformers in the Scikit-Learn's `ColumnTransformer` class in direct and inverse transformation
- Fixed efficiency plots produced for the HTML training report

Beta release v0.1.3 of the **lb-pidsim-train** package! 🔥

What's new?
- Added new GAN algorithm: WGAN-ALP
- Bug fixing
- Weight-aware QuantileTransformer exported as a Scikit-Learn class
- Fixed issues with the custom [ColumnTransformer](https://scikit-learn.org/stable/modules/generated/sklearn.compose.ColumnTransformer.html), namely [LbColTransformer](https://github.com/mbarbetti/lb-pidsim-train/blob/main/lb_pidsim_train/preprocessing/LbColTransformer.py)
- Re-enabled models export through snakemake

Beta release v0.1.2 of the **lb-pidsim-train** package! 🔥

What's new?
- Enabled training with simulated samples
- Added a _weight-aware_ version of the [QuantileTransformer](https://scikit-learn.org/stable/modules/generated/sklearn.preprocessing.QuantileTransformer.html) as strategy for preprocessing
- Updated the information included within the HTML report
- Added hyperparameter `key : value`s
- Added a schematic representation of the discriminator and generator architectures
- Added 1D correlation plots in 8 different kinematic bins

Beta release v0.1.1 of the **lb-pidsim-train** package! 🔥

What's new?
- Added output variable to the Muon-detector parameterizations
- Added log-scale validation plots to the HTML report
- Upgraded the [QuantileTransformer](https://scikit-learn.org/stable/modules/generated/sklearn.preprocessing.QuantileTransformer.html) strategy for preprocessing
- Official validation plots moved to https://gitlab.cern.ch/mabarbet/lamarr-models-validation/
- Bug fixing

I am happy to announce the first beta release (v0.1.0) of the **lb-pidsim-train** package! 🎉

Available trainers
- [x] DataHandler
- [x] BaseTrainer
- [x] TensorTrainer
- [x] GanTrainer
- [ ] VaeTrainer
- [ ] NormFlowTrainer
- [x] ScikitTrainer
- [x] ScikitClassifier

Available metrics
- [x] Kullback-Leibler divergence
- [x] Jensen–Shannon divergence
- [x] Kolmogorov-Smirnov test
- [x] Chi-squared test

Available algorithms
- [x] Generative Adversarial Networks
- [x] GAN
- [x] BceGAN
- [ ] WGAN
- [x] WGAN-GP
- [x] CramerGAN
- [ ] Variational Autoencoder
- [ ] Normalizing Flows