skglm
skglm copied to clipboard
scikit-learn-contrib
Fast and modular sklearn replacement for generalized linear models
## Context of the PR To try and account for the weights for samples, I have tried to modify the Quadratic class, since it was an issue during MCPRegression where...
## Context of the PR Tries to add an ElasticNet like regularization for SparseLogisticRegression ## Contributions of the PR Added an L1_plus_L2 penalty ### Checks before merging PR - [...
A quick proof-of-concept of a function that checks if the combination `(solver, datafit, penalty)` is supported. Currently we have some edge cases where one can pass `ProxNewton` solver with `L0_5`...
The goal of this PR is to write a CD (BCD) solver when `n_samples >> n_features`. Such configurations are solved much faster by pre-computing a Gram matrix XtX and updating...
A follow up of #228 is to update the equations of the [positive Group Lasso penalty](https://contrib.scikit-learn.org/skglm/tutorials/prox_nn_group_lasso.html#prox-nn-group-lasso). The equations are already available on an overleaf, cf. https://github.com/scikit-learn-contrib/skglm/pull/228#issuecomment-2035387421, and just need to...
more than 5 minutes on my machine Longer tests are: ``` ============================= slowest 20 durations ============================= 21.93s call skglm/tests/test_estimators.py::test_CoxEstimator[True-True] 19.72s call skglm/tests/test_estimators.py::test_check_estimator[Lasso] 15.58s call skglm/tests/test_datafits.py::test_cox[True] 14.25s call skglm/tests/test_estimators.py::test_equivalence_cox_SLOPE_cox_L1[True-True] 13.78s call...
## Description of the feature In multiple issues, it seems that practitioners had problems selecting the regularization parameter `alpha`, and a lot of them do not know the value of...
Sklearn ElasticNet (and its Lasso which inherits from it) exposes a sparse_coef_ attribute after fitting : https://github.com/scikit-learn/scikit-learn/blob/main/sklearn/linear_model/_coordinate_descent.py#L1122 we don't @Badr-MOUFAD suggested we implement one for compatibility reasons ; we never...
Currently we only support L1 in logreg: https://contrib.scikit-learn.org/skglm/generated/skglm.SparseLogisticRegression.html We could introduce a second regularization parameter corresponding to a squared L2 regularisation, like ElasticNet is to Lasso @PascalCarrivain would you give...
## Description of the feature Sklearn 'fit' functions have an option to pass 'sample_weights' array to assign importance to samples and modify the cost function accordingly. **Additional context** Wanted to...
Metadata
Owner
Metadata
Fast and modular sklearn replacement for generalized linear models