Add new models
MKaptein opened this issue · comments
MKaptein commented
The list of models we export is actually larger than in mentioned in the current version. We need to update the docs and "re-push" the package.
Robin van Emden commented
Classification | Regression | |
---|---|---|
Linear |
|
|
SVM |
|
|
Tree |
|
|
Random Forest |
|
|
Boosting |
|
|
MKaptein commented
In the midst of adding this, but needs an update of the toolchain to support lightgbm
(which is also the package name in the bundle).
MKaptein commented
Ran tests for:
package | model | New | Tested (01052020) | Note |
---|---|---|---|---|
lightgbm | LGBMClassifier(rf booster only) | yes | ok | |
lightgbm | LGBMRegressor(rf booster only) | yes | ok | |
lightning | AdaGradClassifier | yes | no | Skip for now |
lightning | AdaGradRegressor | yes | no | Skip for now |
lightning | CDClassifier | yes | no | Skip for now |
lightning | CDRegressor | yes | no | Skip for now |
lightning | FistaClassifier | yes | no | Skip for now |
lightning | FistaRegressor | yes | no | Skip for now |
lightning | KernelSVC | yes | no | Skip for now |
lightning | SAGAClassifier | yes | no | Skip for now |
lightning | SAGARegressor | yes | no | Skip for now |
lightning | SAGClassifier | yes | no | Skip for now |
lightning | SAGRegressor | yes | no | Skip for now |
lightning | SDCAClassifier | yes | no | Skip for now |
lightning | SDCARegressor | yes | no | Skip for now |
sklearn | ARDRegression | no | ok | |
sklearn | BayesianRidge | yes | ok | |
sklearn | DecisionTreeClassifier | no | ok | |
sklearn | DecisionTreeRegressor | no | ok | |
sklearn | ElasticNet | no | ok | |
sklearn | ElasticNetCV | no | ok | |
sklearn | ExtraTreeClassifier | no | ok | |
sklearn | ExtraTreeRegresso | no | ok | |
sklearn | ExtraTreesClassifier | no | ok | |
sklearn | ExtraTreesRegressor | no | ok | |
sklearn | HuberRegressor | no | ok | |
sklearn | Lars | no | ok | |
sklearn | LarsCV | no | ok | |
sklearn | Lasso | no | ok | |
sklearn | LassoCV | no | ok | |
sklearn | LassoLars | no | ok | |
sklearn | LassoLarsCV | no | ok | |
sklearn | LassoLarsIC | no | ok | |
sklearn | LinearRegression | no | ok | |
sklearn | LinearSVC | no | ok | |
sklearn | LinearSVR | no | ok | |
sklearn | LogisticRegression | yes | ok | |
sklearn | LogisticRegressionCV | yes | ok | |
sklearn | NuSVC | no | ok | |
sklearn | NuSVR | no | ok | |
sklearn | OrthogonalMatchingPursuit | no | ok | |
sklearn | OrthogonalMatchingPursuitCV | no | ok | |
sklearn | PassiveAggressiveClassifier | no | ok | |
sklearn | PassiveAggressiveRegressor | no | ok | |
sklearn | Perceptron | yes | ok | |
sklearn | RandomForestClassifier | no | ok | |
sklearn | RandomForestRegressor | no | ok | |
sklearn | RANSACRegressor | no | ok | |
sklearn | Ridge | no | ok | |
sklearn | RidgeClassifier | yes | ok | |
sklearn | RidgeClassifierCV | yes | ok | |
sklearn | RidgeCV | no | ok | |
sklearn | SGDClassifier | no | ok | |
sklearn | SGDRegressor | no | ok | |
sklearn | SVC | no | ok | |
sklearn | SVR | no | ok | |
sklearn | TheilSenRegressor | no | ok | |
statsmodels | [Gaussian] Process Regression Using Maximum Likelihood-based Estimation (ProcessMLE) | yes | not ok | Skip for now |
statsmodels | Generalized Least Squares (GLS) | ok | ||
statsmodels | Generalized Least Squares with AR Errors (GLSAR) | yes | ok | |
statsmodels | Ordinary Least Squares (OLS) | ok | ||
statsmodels | Quantile Regression (QuantReg) | yes | ok | |
statsmodels | Weighted Least Squares (WLS) | ok | ||
xgboost | XGBClassifier(gbtree/gblinear booster only) | no | ok | |
xgboost | XGBRegressor(gbtree/gblinear booster only) | no | ok | |
xgboost | XGBRFClassifier(binary only, multiclass is not supported yet) | no | ok | |
xgboost | XGBRFRegressor | no | ok |
So, added a few models (including lightgbm) and xgboost. Not added the lightning ones (although in test_all_models.py
these are specified already and they should be simple to add by:
- Checking if _model_fitted() and _precit() work for lightning models
- Mapping the lightning package to the sklearn toolchain (in main.py)
- Running the tests.
I am closing this for now.
MKaptein commented
package | model | Tested (05052020) | Note |
---|---|---|---|
lightgbm | LGBMClassifier | ok | |
lightgbm | LGBMRegressor | ok | |
sklearn | ARDRegression | ok | |
sklearn | BayesianRidge | ok | |
sklearn | DecisionTreeClassifier | ok | |
sklearn | DecisionTreeRegressor | ok | |
sklearn | ElasticNet | ok | |
sklearn | ElasticNetCV | ok | |
sklearn | ExtraTreeClassifier | ok | |
sklearn | ExtraTreeRegresso | ok | |
sklearn | ExtraTreesClassifier | ok | |
sklearn | ExtraTreesRegressor | ok | |
sklearn | HuberRegressor | ok | |
sklearn | Lars | ok | |
sklearn | LarsCV | ok | |
sklearn | Lasso | ok | |
sklearn | LassoCV | ok | |
sklearn | LassoLars | ok | |
sklearn | LassoLarsCV | ok | |
sklearn | LassoLarsIC | ok | |
sklearn | LinearRegression | ok | |
sklearn | LinearSVC | ok | |
sklearn | LinearSVR | ok | |
sklearn | LogisticRegression | ok | |
sklearn | LogisticRegressionCV | ok | |
sklearn | NuSVC | ok | |
sklearn | NuSVR | ok | |
sklearn | OrthogonalMatchingPursuit | ok | |
sklearn | OrthogonalMatchingPursuitCV | ok | |
sklearn | PassiveAggressiveClassifier | ok | |
sklearn | PassiveAggressiveRegressor | ok | |
sklearn | Perceptron | ok | |
sklearn | RandomForestClassifier | ok | |
sklearn | RandomForestRegressor | ok | |
sklearn | RANSACRegressor | ok | |
sklearn | Ridge | ok | |
sklearn | RidgeClassifier | ok | |
sklearn | RidgeClassifierCV | ok | |
sklearn | RidgeCV | ok | |
sklearn | SGDClassifier | ok | |
sklearn | SGDRegressor | ok | |
sklearn | SVC | ok | |
sklearn | SVR | ok | |
sklearn | TheilSenRegressor | ok | |
statsmodels | Generalized Least Squares (GLS) | ok | |
statsmodels | Generalized Least Squares with AR Errors (GLSAR) | ok | |
statsmodels | Ordinary Least Squares (OLS) | ok | |
statsmodels | Quantile Regression (QuantReg) | ok | |
statsmodels | Weighted Least Squares (WLS) | ok | |
xgboost | XGBClassifier | ok | |
xgboost | XGBRegressor | ok | |
xgboost | XGBRFClassifier | ok | Binary only |
xgboost | XGBRFRegressor | ok |