Changelog

Development Version

Version 0.4.0

(June 2022)

• DOC explain how to implement a winner-take-all model.

• FIX comply with most recent scikit-learn’s check_estimator.

• FIX avoid an indexing error in the hypergradient solver, when early stopping after different numbers of iterations for different batches.

Version 0.3.6

(April 2022)

• DOC improve documentation website, add estimator flowchart.

• TST improve test robustness.

• ENH add batching over targets in predict_weighted_kernel_ridge().

• ENH add solver="auto" in KernelRidge, which switches solver based on the presence of a separate alpha per target.

Version 0.3.5

(February 2022)

• MNT speed up examples on CPU, to build the doc faster on github actions.

• ENH add batching over targets in Ridge, KernelRidge, and WeightedKernelRidge.

• ENH add warnings to guide the user between using Ridge or KernelRidge.

• ENH add user-friendly errors when the number of samples is inconsistent.

• ENH raise ValueError if the indices in cross-validation exceed number of samples.

Version 0.3.4

(November 2021)

• FIX Ridge with n_samples < n_targets.

• FIX update of alphas when local_alpha=False in MultipleKernelRidgeCV.

• EXA refactor examples with new generate_multikernel_dataset() function.

• MNT add github actions for running tests, building and publishing the doc, and publishing to PyPI.

Version 0.3.3

(November 2021)

• FIX KernelRidge with n_samples < n_targets.

• FIX random search with single alpha in MultipleKernelRidgeCV.

Version 0.3.2

(November 2021)

• ENH add r2_score_split_svd() scoring function.

• ENH add correlation_score_split() scoring function.

• ENH add split parameter to the score method in WeightedKernelRidge, MultipleKernelRidgeCV, and GroupRidgeCV.

• ENH add force_cpu parameter in all estimators.

• FIX remove deprecation warnings for cupy v9.

• DOC mention that pytorch 1.9+ is preferred.

Version 0.3.1

(September 2021)

• MNT Rename BandedRidgeCV into GroupRidgeCV (both names are available).

• ENH improve robustness to noise in the cross-validation scores.

• ENH start the random search with equal weights in MultipleKernelRidgeCV and GroupRidgeCV.

• FIX remove deprecation warnings with pytorch 1.8.

• TST improve test coverage.

Version 0.3.0

(April 2021)

• ENH add fit_intercept parameter in Ridge, RidgeCV, and BandedRidgeCV.

• ENH add fit_intercept parameter in KernelRidge, KernelRidgeCV, solve_multiple_kernel_ridge_gradient_descent(), and solve_multiple_kernel_ridge_random_search().

• ENH add KernelCenterer.

• ENH allow change of backend midscript.

• ENH Add option to return selected alpha values in solve_multiple_kernel_ridge_random_search().

Version 0.2.0

(December 2020)

Version 0.1.0

(March 2020)