-
-
Notifications
You must be signed in to change notification settings - Fork 20
/
README.Rmd
135 lines (101 loc) · 7.98 KB
/
README.Rmd
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
---
output: github_document
---
```{r, include = FALSE}
knitr::opts_chunk$set(
cache = FALSE,
collapse = TRUE,
comment = "#>"
)
```
# mlr3proba
Probabilistic Supervised Learning for **[mlr3](https://github.com/mlr-org/mlr3/)** ([website](https://mlr3proba.mlr-org.com/)).
<!-- badges: start -->
[](https://github.com/mlr-org/mlr3proba/actions/workflows/r-cmd-check.yml)
[](https://mlr-org.r-universe.dev/mlr3proba)
[](https://github.com/mlr-org/mlr3proba/discussions)
[](https://doi.org/10.1093/bioinformatics/btab039)
[](https://stackoverflow.com/questions/tagged/mlr3)
[](https://lmmisld-lmu-stats-slds.srv.mwn.de/mlr_invite/)
<!-- badges: end -->
## What is mlr3proba?
`mlr3proba` is a machine learning toolkit for making probabilistic predictions within the **[mlr3](https://github.com/mlr-org/mlr3)** ecosystem.
It currently supports the following tasks:
1. **Predictive survival analysis**: survival analysis where individual hazards and survival distributions can be queried.
2. **Unconditional distribution estimation**: main returned output is the distribution.
Sub-cases are density estimation and unconditional survival estimation.
3. **Probabilistic supervised regression**: Supervised regression with a predictive distribution as the return type.
The survival analysis part is considered in a mature state, the rest are in early stages of development.
## Feature Overview
Key features of `mlr3proba` focus on survival analysis and are:
- Task frameworks for survival analysis (`TaskSurv`)
- A comprehensive selection of predictive survival learners (mostly via
[mlr3extralearners](https://github.com/mlr-org/mlr3extralearners/))
- A unified `train`/`predict` model interface to any probabilistic predictive model (frequentist, Bayesian, Deep Learning, or other)
- Use of the **[distr6](https://github.com/alan-turing-institute/distr6)**
probability distribution interface as its probabilistic predictive return type
- A comprehensive selection of measures for evaluating the performance of survival learners, with respect to prognostic index (continuous rank) prediction, and probabilistic (distribution) prediction
- Basic ML pipeline building integrated with **[mlr3pipelines](https://github.com/mlr-org/mlr3pipelines)**
- Reduction/composition strategies using linear predictors and baseline hazards
## Installation
`mlr3proba` is not currently on CRAN.
Please follow one of the two following methods to install it:
### R-universe
```r
install.packages("mlr3proba", repos = "https://mlr-org.r-universe.dev")
```
Or for easier installation going forward:
1. Run `usethis::edit_r_environ()` then in the file that opened add or edit `options` to look something like:
```r
options(repos = c(
raphaels1 = "https://raphaels1.r-universe.dev",
mlrorg = "https://mlr-org.r-universe.dev", # add this line
CRAN = "https://cloud.r-project.org"
))
```
2. Save and close the file, restart your `R` session
3. Run `install.packages("mlr3proba")` as usual
### GitHub
Install the latest development version:
```r
remotes::install_github("mlr-org/mlr3proba")
```
## Learners
- [Core learners](https://mlr3proba.mlr-org.com/reference/index.html#survival-learners) are implemented in `mlr3proba` and include the Kaplan-Meier Estimator, the Cox Proportional Hazards model and the Survival Tree learner.
- In [mlr3extralearners](https://github.com/mlr-org/mlr3extralearners) we have interfaced several advanced ML survival learners.
Use the [interactive search table](https://mlr-org.com/learners.html) to search for the available survival learners and see the [learner status page](https://mlr3extralearners.mlr-org.com/articles/learner_status.html) for their live status.
## Measures
For density estimation and probabilistic regression only the **log-loss** is currently implemented.
For survival analysis, see full list [here](https://mlr3proba.mlr-org.com/reference/index.html#survival-measures).
Some commonly used measures are the following:
| ID | Measure | Package | Category | Prediction Type
| :--| :------ | :------ | :------ | :------- |
| [surv.dcalib](https://mlr3proba.mlr-org.com/reference/mlr_measures_surv.dcalib.html) | D-Calibration | [mlr3proba](https://CRAN.R-project.org/package=mlr3proba) | Calibration | distr
| [surv.cindex](https://mlr3proba.mlr-org.com/reference/mlr_measures_surv.cindex.html) | Concordance Index | [mlr3proba](https://CRAN.R-project.org/package=mlr3proba) | Discrimination | crank
| [surv.uno_auc](https://mlr3proba.mlr-org.com/reference/mlr_measures_surv.uno_auc.html) | Uno's AUC | [survAUC](https://CRAN.R-project.org/package=survAUC) | Discrimination | lp
| [surv.graf](https://mlr3proba.mlr-org.com/reference/mlr_measures_surv.graf.html) | Integrated Brier Score | [mlr3proba](https://CRAN.R-project.org/package=mlr3proba) | Scoring Rule | distr
| [surv.rcll](https://mlr3proba.mlr-org.com/reference/mlr_measures_surv.rcll.html) | Right-Censored Log loss | [mlr3proba](https://CRAN.R-project.org/package=mlr3proba) | Scoring Rule | distr
| [surv.intlogloss](https://mlr3proba.mlr-org.com/reference/mlr_measures_surv.intlogloss.html) | Integrated Log-Likelihood | [mlr3proba](https://CRAN.R-project.org/package=mlr3proba) | Scoring Rule | distr
## 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](https://github.com/mlr-org/mlr3proba/issues)\!
In case of problems / bugs, it is often helpful if you provide a "minimum working example" using [reprex](https://reprex.tidyverse.org/) that showcases the behavior.
## Similar Projects
Predecessors to this package are previous instances of survival modelling in **[mlr](https://github.com/mlr-org/mlr)**.
The **[skpro](https://github.com/alan-turing-institute/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](https://CRAN.R-project.org/package=rjags)** and **[stan](https://github.com/stan-dev/rstan)**.
For implementation of a few survival models and measures, a central package is **[survival](https://github.com/therneau/survival)**.
There does not appear to be a package that provides an architectural framework for distribution/density estimation, see **[this list](https://vita.had.co.nz/papers/density-estimation.pdf)** 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`.
## Citing mlr3proba
If you use mlr3proba, please cite our [Bioinformatics article](https://doi.org/10.1093/bioinformatics/btab039):
```{r echo = FALSE, comment = ""}
toBibtex(citation("mlr3proba"))
```