mlr3proba
Package website: release | dev
Probabilistic Supervised Learning for mlr3.
What is mlr3proba ?
mlr3proba is a machine learning toolkit for making probabilistic predictions within the mlr3 ecosystem. It currently supports the following tasks:
- Probabilistic supervised regression - Supervised regression with a predictive distribution as the return type.
- Predictive survival analysis - Survival analysis where individual predictive hazards can be queried. This is equivalent to probabilistic supervised regression with censored observations.
- Unconditional distribution estimation, where the distribution is returned. Sub-cases are density estimation and unconditional survival estimation.
Key features of mlr3proba are
- A unified fit/predict model interface to any probabilistic predictive model (frequentist, Bayesian, or other)
- Pipeline/model composition
- Task reduction strategies
- Domain-agnostic evaluation workflows using task specific algorithmic performance measures.
mlr3proba makes use of the distr6 probability distribution interface as its probabilistic predictive return type.
Feature Overview
The current mlr3proba release focuses on survival analysis, and contains:
- Task frameworks for survival analysis (
TaskSurv
) - A comprehensive selection of 17 predictive survival learners
- A comprehensive selection of 21 performance measures for predictive survival learners, with respect to prognostic index (continuous rank) prediction, and probabilistic (distribution) prediction
- PipeOps integrated with mlr3pipelines, for basic pipeline building, and reduction/composition strategies using linear predictors and baseline hazards.
Roadmap
The vision of mlr3proba is to provide comprehensive machine learning functionality to the mlr3 ecosystem for continuous probabilistic return types.
The lifecycle of the survival task and features are considered
maturing
and any major changes are unlikely.
The density and probabilistic supervised regression tasks are currently in the early stages of development. Task frameworks have been drawn up, but may not be stable; learners need to be interfaced, and contributions are very welcome (see issues).
Installation
Install the last release from CRAN:
install.packages("mlr3proba")
Install the development version from GitHub:
remotes::install_github("mlr-org/mlr3proba")
Survival Analysis
Survival Learners
Learners are located either in mlr3proba, the mlr3learners repository, or the mlr3learners organisation. See here for instructions in how to install learners from the mlr3learners organisation.
ID | Learner | Package |
---|---|---|
surv.akritas | Akritas Conditional Non-Parametric Estimator | mlr3learners.proba |
surv.blackboost | Gradient Boosting with Regression Trees | mboost |
surv.coxboost | Cox Model with Likelihood Based Boosting | CoxBoost |
surv.coxph | Cox Proportional Hazards | survival |
surv.coxtime | Non-Linear Cox Neural Network | pycox |
surv.cvcoxboost | Cox Model with Cross-Validation Likelihood Based Boosting | CoxBoost |
surv.cvglmnet | Cross-Validated GLM with Elastic Net Regularization | glmnet |
surv.deephit | Discerete Deep Ranking Neural Network | pycox |
surv.deepsurv | Deep Cox Proportional Hazards | pycox |
surv.dnn | Deep Neural Network with Pseudo Values | mlr3learners.proba |
surv.flexible | Flexible Parametric Spline Models | flexsurv |
surv.gamboost | Gradient Boosting for Additive Models | mboost |
surv.gbm | Generalized Boosting Regression Modeling | gbm |
surv.glmboost | Gradient Boosting with Component-wise Linear Models | mboost |
surv.glmnet | GLM with Elastic Net Regularization | glmnet |
surv.kaplan | Kaplan-Meier Estimator | survival |
surv.loghaz | Logistic Hazard Neural Network | pycox |
surv.mboost | Gradient Boosting for Generalized Additive Models | mboost |
surv.nelson | Nelson-Aalen Estimator | survival |
surv.parametric | Fully Parametric Survival Models | survival |
surv.penalized | L1 and L2 Penalized Estimation in GLMs | penalized |
surv.pchazard | Piecewise Constant Hazard Neural Network | pycox |
surv.randomForestSRC | RandomForestSRC Survival Forest | randomForestSRC |
surv.ranger | Ranger Survival Forest | ranger |
surv.rpart | Rpart Survival Forest | rpart |
surv.svm | Regression, Ranking and Hybrid Support Vector Machines | survivalsvm |
surv.xgboost | Cox Model with Gradient Boosting Trees | xgboost |
Survival Measures
ID | Measure | Package |
---|---|---|
surv.calib_alpha | van Houwelingen’s Alpha Calibration | mlr3proba |
surv.calib_beta | van Houwelingen’s Beta Calibration | mlr3proba |
surv.chambless_auc | Chambless and Diao’s AUC | survAUC |
surv.graf | Integrated Graf Score | mlr3proba |
surv.hungAUC | Hung and Chiang’s AUC | survAUC |
surv.intlogloss | Integrated Log Loss | mlr3proba |
surv.logloss | Log Loss | mlr3proba |
surv.nagelk_r2 | Nagelkerke’s R2 | survAUC |
surv.oquigley_r2 | O’Quigley, Xu, and Stare’s R2 | survAUC |
surv.song_auc | Song and Zhou’s AUC | survAUC |
surv.song_tnr | Song and Zhou’s TNR | survAUC |
surv.song_tpr | Song and Zhou’s TPR | survAUC |
surv.uno_auc | Uno’s AUC | survAUC |
surv.uno_tnr | Uno’s TNR | survAUC |
surv.uno_tpr | Uno’s TPR | survAUC |
surv.xu_r2 | Xu and O’Quigley’s R2 | survAUC |
Density Estimation
Density Learners
Learners are located either in mlr3proba, the mlr3learners repository, or the mlr3learners organisation. See here for instructions in how to install learners from the mlr3learners organisation.
ID | Learner | Package |
---|---|---|
dens.hist | Univariate Histogram Density Estimator | graphics |
dens.kde | Univariate KDE for Different Kernels | distr6 |
dens.kdeKD | Nonparametric KDE Using Plug-in Method of Polansky and Baker | kerdiest |
dens.kdeKS | Nonparametric Gaussian KDE | ks |
dens.locfit | Nonparametric KDE Using Gaussian kernel | locfit |
dens.logspline | Logspline Method for Density Estimation | logspline |
dens.mixed | KDE Using Li and Racine Bandwidth Specification | np |
dens.nonpar | Nonparametric KDE Using Normal Optimal Smoothing Parameter | sm |
dens.pen | Density Estimation with a Penalized Mixture | pendensity |
dens.plug | Density Estimation with Iterative Plug-in Bandwidth Selection | plugdensity |
dens.spline | Density Estimation Using Smoothing Spline ANOVA | gss |
Density Measures
ID | Measure | Package |
---|---|---|
dens.logloss | Log Loss | mlr3proba |
Near-Future Plans
- Add
prob
predict type toTaskRegr
, and associated learners/measures - Allow
MeasureSurv
to return measures at multiple time-points simultaneously - Continue to add survival measures and learners
Bugs, Questions, Feedback
mlr3proba is a free and open source software project that encourages participation and feedback. If you have any issues, questions, suggestions or feedback, please do not hesitate to open an “issue” about it on the GitHub page!
In case of problems / bugs, it is often helpful if you provide a “minimum working example” that showcases the behaviour (but don’t worry about this if the bug is obvious).
Similar Projects
Predecessors to this package are previous instances of survival modelling in mlr. The skpro package in the python/scikit-learn ecosystem follows a similar interface for probabilistic supervised learning and is an architectural predecessor. Several packages exist which allow probabilistic predictive modelling with a Bayesian model specific general interface, such as rjags and stan. For implementation of a few survival models and measures, a central package is survival. There does not appear to be a package that provides an architectural framework for distribution/density estimation, see this list for a review of density estimation packages in R.
Acknowledgements
Several people contributed to the building of mlr3proba
. Firstly,
thanks to Michel Lang for writing mlr3survival
. Several learners and
measures implemented in mlr3proba
, as well as the prediction, task,
and measure surv objects, were written initially in mlr3survival
before being absorbed into mlr3proba
. Secondly thanks to Franz Kiraly
for major contributions towards the design of the proba-specific parts
of the package, including compositors and predict types. Also for
mathematical contributions towards the scoring rules implemented in the
package. Finally thanks to Bernd Bischl and the rest of the mlr core
team for building mlr3
and for many conversations about the design of
mlr3proba
.