CHANGES.md

version 0.24.5

• Update LazyDeepMTS: No more LazyMTS class, instead, you can use LazyDeepMTS with n_layers=1
• Specify forecasting horizon in LazyDeepMTS (see updated docs and examples/lazy_mts_horizon.py)
• New class ClassicalMTS for classsical models (for now VAR and VECM adapted from statsmodels) in multivariate time series forecasting
• partial_fit for CustomClassifier and CustomRegressor

version 0.23.1

• Copula simulation for time series residuals in classes MTS and DeepMTS
  • based on copulas of in-sample residuals: vine-tll (default), vine-bb1, vine-bb6, vine-bb7, vine-bb8, vine-clayton, vine-frank, vine-gaussian, vine-gumbel, vine-indep, vine-joe, vine-student
  • scp-vine-tll (default), scp-vine-bb1, scp-vine-bb6, scp-vine-bb7, scp-vine-bb8, scp-vine-clayton, scp-vine-frank, scp-vine-gaussian, scp-vine-gumbel, scp-vine-indep, scp-vine-joe, scp-vine-student
  • scp2-vine-tll, scp2-vine-bb1, scp2-vine-bb6, scp2-vine-bb7, scp2-vine-bb8, scp2-vine-clayton, scp2-vine-frank, scp2-vine-gaussian, scp2-vine-gumbel, scp2-vine-indep, scp2-vine-joe, scp2-vine-student
• cross_val_score: time series cross-validation for MTS and DeepMTS Technical:
• Do not scale sparse matrices before training
• Add MaxAbsScaler

version 0.22.7

• Implement new types of predictive simulation intervals (type_pis): independent bootstrap, block bootstrap, 2 variants of split conformal prediction in class MTS (see updated docs)
• Gaussian prediction intervals type_pi == "gaussian" in class MTS
• Implement Winkler score in LazyMTS and LazyDeepMTS for probabilistic forecasts
• Use conformalized Estimators in MTS (see examples/mts_conformal_not_sims.py)
• Include block_size for block bootstrapping methods for *MTS classes

version 0.20.6

Technical:

• Import all_estimators from sklearn.utils
• Use both sparse and sparse_output in OneHotEncoder (for compatibility with older versions of sklearn)

version 0.18.0

• Bayesian CustomRegressor
• Conformalized CustomRegressor (splitconformal and localconformal for now)
• See this example, this example, and this notebook

version 0.17.2

• self.n_classes_ = len(np.unique(y)) # for compatibility with sklearn

version 0.17.1

• preprocessing for all LazyDeep*

version 0.17.0

• Attribute estimators (a list of Estimator's as strings) for LazyClassifier, LazyRegressor, LazyDeepClassifier, LazyDeepRegressor, LazyMTS, and LazyDeepMTS
• New documentation for the package, using pdoc (not pdoc3)
• Remove external regressors xreg at inference time for MTS and DeepMTS
• New class Downloader: querying the R universe API for datasets (see https://thierrymoudiki.github.io/blog/2023/12/25/python/r/misc/mlsauce/runiverse-api2 for similar example in mlsauce)
• Add custom metric to Lazy*
• Rename Deep regressors and classifiers to Deep* in Lazy*
• Add attribute sort_by to Lazy* -- sort the data frame output by a given metric
• Add attribute classes_ to classifiers (ensure consistency with sklearn)

version 0.16.8

• Subsample response by using the number of rows, not only a percentage (see https://thierrymoudiki.github.io/blog/2024/01/22/python/nnetsauce-subsampling)
• Improve consistency with sklearn's v1.2, for OneHotEncoder

version 0.16.3

• add robust scaler
• relatively faster scaling in preprocessing
• Regression-based classifiers (see https://www.researchgate.net/publication/377227280_Regression-based_machine_learning_classifiers)
• DeepMTS (multivariate time series forecasting with deep quasi-random layers): see https://thierrymoudiki.github.io/blog/2024/01/15/python/quasirandomizednn/forecasting/DeepMTS
• AutoML for MTS (multivariate time series forecasting): see https://thierrymoudiki.github.io/blog/2023/10/29/python/quasirandomizednn/MTS-LazyPredict
• AutoML for DeepMTS (multivariate time series forecasting): see https://github.com/Techtonique/nnetsauce/blob/master/nnetsauce/demo/thierrymoudiki_20240106_LazyDeepMTS.ipynb
• Spaghetti plots for MTS and DeepMTS (multivariate time series forecasting): see https://thierrymoudiki.github.io/blog/2024/01/15/python/quasirandomizednn/forecasting/DeepMTS
• Subsample continuous and discrete responses

version 0.16.0

• actually implement deep Estimators in /deep (in addition to /lazypredict)
• include new multi-output regression-based classifiers (see https://thierrymoudiki.github.io/blog/2021/09/26/python/quasirandomizednn/classification-using-regression for more details)
• use proper names for Estimators in /lazypredict and /deep
• expose SubSampler (stratified subsampling) to the external API

version 0.15.0

• lazy predict for classification and regression (see https://thierrymoudiki.github.io/blog/2023/10/22/python/quasirandomizednn/nnetsauce-lazy-predict-preview)
• lazy predict for multivariate time series (see https://thierrymoudiki.github.io/blog/2023/10/29/python/quasirandomizednn/MTS-LazyPredict)
• lazy predict for deep classifiers and regressors (see this example for classification and this example for regression)

version 0.14.0

• update and align as much as possible with R version
• colored graphics for class MTS

version 0.13.0

• Fix error in nodes' simulation (base.py)
• Use residuals and KDE for predictive simulations
• plot method for MTS objects

version 0.12.1

• Begin residuals simulation

version 0.11.5

• Avoid division by zero in scaling

version 0.11.4

• less dependencies in setup

version 0.11.0

• Implement RandomBagRegressor
• Use of a DataFrame in MTS

version 0.10.0

• rename attributes with underscore
• add more examples to documentation

version 0.9.6

• Fix numbers' simulations

version 0.9.4

• Remove memoize from Simulator

version 0.9.2

• loosen the range of Python packages versions

version 0.9.0

• Add Poisson and Laplace regressions to GLMRegressor
• Remove smoothing weights from MTS

version 0.8.0

• Use C++ for simulation
• Fix R Engine problem

version 0.7.0

• RandomBag classifier cythonized

version 0.6.0

• Documentation with MkDocs
• Cython-ready

version 0.5.0

• contains a refactorized code for the Base class, and for many other utilities.
• makes use of randtoolbox for a faster, more scalable generation of quasi-random numbers.
• contains a (work in progress) implementation of most algorithms on GPUs, using JAX. Most of the nnetsauce's changes related to GPUs are currently made on potentially time consuming operations such as matrices multiplications and matrices inversions.

version 0.4.0

• (Work in progress) documentation in /docs
• MultitaskClassifier
• Rename Mtask to Multitask
• Rename Ridge2ClassifierMtask to Ridge2MultitaskClassifier

version 0.3.3

• Use "return_std" only in predict for MTS object

version 0.3.2

• Fix for potential error "Sample weights must be 1D array or scalar"

version 0.3.1

• One-hot encoding not cached (caused errs on multitask ridge2 classifier)
• Rename ridge to ridge2 (2 shrinkage params compared to ridge)

version 0.3.0

• Implement ridge2 (regressor and classifier)
• Upper bound on Adaboost error
• Test Time series split

version 0.2.0

• Add AdaBoost classifier
• Add RandomBag classifier (bagging)
• Add multinomial logit Ridge classifier
• Remove dependency to package sobol_seq(not used)

version 0.1.0

• Initial version