| Title: | Lightweight interface to sklearn, nnetsauce and unifiedbooster with conformal prediction |
|---|---|
| Description: | Lightweight interface to Python packages sklearn, nnetsauce and unifiedbooster with conformal prediction. |
| Authors: | T. Moudiki [aut, cre] |
| Maintainer: | T. Moudiki <[email protected]> |
| License: | BSD_3_clause Clear + file LICENSE |
| Version: | 0.10.0 |
| Built: | 2026-05-23 09:15:31 UTC |
| Source: | https://github.com/Techtonique/tisthemachinelearner_r |
An R6 Class that provides an interface to gradient boosting with neural network feature transformation.
estimatorsList of fitted base models
learning_rateLearning rate for boosting
lossesVector of training losses
n_estimatorsNumber of estimators used
new()
Create a new Booster object
Booster$new( model_name = "ExtraTreeRegressor", n_estimators = 100L, learning_rate = 0.1, tolerance = 1e-04, calibration = FALSE, seed = 123L, show_progress = TRUE, verbose = FALSE, venv_path = "./venv" )
model_nameName of the base model
n_estimatorsNumber of boosting iterations
learning_rateLearning rate for boosting
toleranceConvergence tolerance
calibrationWhether to calibrate the model
seedRandom seed
show_progressWhether to show progress bar
verboseWhether to print detailed output
venv_pathPath to the virtual environment
fit()
Fit the boosting model to training data
Booster$fit(x, y)
xFeature matrix
yTarget vector
The fitted object (invisible)
predict()
Make predictions on new data
Booster$predict(newdata)
newdataNew data to predict on
Vector of predictions
clone()
The objects of this class are cloneable with this method.
Booster$clone(deep = FALSE)
deepWhether to make a deep clone.
Fit a boosting model with neural network feature transformation
boosterCpp( x, y, model_name, n_estimators = 100L, learning_rate = 0.1, tolerance = 1e-04, calibration = FALSE, seed = 123L, show_progress = TRUE, verbose = FALSE, venv_path = "./venv" )boosterCpp( x, y, model_name, n_estimators = 100L, learning_rate = 0.1, tolerance = 1e-04, calibration = FALSE, seed = 123L, show_progress = TRUE, verbose = FALSE, venv_path = "./venv" )
x |
Input matrix |
y |
Target vector |
model_name |
Name of the base model |
n_estimators |
Number of boosting iterations |
learning_rate |
Learning rate for boosting |
tolerance |
Convergence tolerance |
calibration |
Whether to calibrate the model |
seed |
Random seed |
show_progress |
Whether to show progress bar |
verbose |
Whether to print detailed output |
venv_path |
Path to the virtual environment |
This function retrieves a list of all models available in scikit-learn. It imports the necessary Python modules and retrieves all estimators, filtering them into classifiers and regressors.
get_model_list(venv_path = "./venv")get_model_list(venv_path = "./venv")
A list with two elements: - 'classifiers': A character vector of all classifier models - 'regressors': A character vector of all regressor models
# model_list <- get_model_list() # print(model_list$classifiers)# model_list <- get_model_list() # print(model_list$classifiers)
Predict using a boosted model
## S3 method for class 'booster' predict(object, newdata, ...)## S3 method for class 'booster' predict(object, newdata, ...)
object |
A boosted model object |
newdata |
New data to predict on |
... |
Additional arguments |
Predict method for regressor objects
## S3 method for class 'regressor' predict( object, newdata, nsim = 250L, level = 95, method = c("none", "splitconformal", "surrogate", "bootstrap", "tsbootstrap", "bayesian"), seed = 123, ... )## S3 method for class 'regressor' predict( object, newdata, nsim = 250L, level = 95, method = c("none", "splitconformal", "surrogate", "bootstrap", "tsbootstrap", "bayesian"), seed = 123, ... )
object |
A regressor object |
newdata |
New data to predict on |
nsim |
Number of simulations for bootstrap/tsbootstrap |
level |
Confidence level for prediction intervals |
method |
Method for computing prediction intervals |
seed |
Seed for the random number generator |
... |
Additional arguments |
Predict using a boosted model
predictBoosterCpp(booster, x)predictBoosterCpp(booster, x)
booster |
A boosted model object |
x |
New data to predict on |
An R6 Class that provides an interface to scikit-learn regression models.
modelThe underlying sklearn model
residualsModel residuals
df.residualDegrees of freedom of residuals
new()
Create a new Regressor object
Regressor$new(model_name, venv_path = "./venv", ...)
model_nameName of the sklearn model to use
venv_pathPath to the virtual environment
...Additional parameters passed to the sklearn model
fit()
Fit the model to training data
Regressor$fit(x, y, calibration = FALSE, seed = 42L, ...)
xFeature matrix
yTarget vector
calibrationLogical flag to indicate if calibration of residuals should be used
seedSeed for random number generator
predict()
Make predictions on new data
Regressor$predict(
newdata,
method = c("none", "splitconformal", "surrogate", "bootstrap", "tsbootstrap",
"bayesian"),
nsim = 250L,
level = 95,
seed = 123
)newdataNew data to predict on
methodMethod for computing prediction intervals
nsimNumber of simulations for bootstrap/tsbootstrap
levelConfidence level for prediction intervals
seedRandom seed
clone()
The objects of this class are cloneable with this method.
Regressor$clone(deep = FALSE)
deepWhether to make a deep clone.
Setup Python environment using uv
setup_sklearn(venv_path = "venv")setup_sklearn(venv_path = "venv")
venv_path |
Path to virtual environment (default: "./venv") |
## Not run: # After creating venv with: uv venv venv setup_sklearn() ## End(Not run)## Not run: # After creating venv with: uv venv venv setup_sklearn() ## End(Not run)
Simulate method for regressor objects
## S3 method for class 'regressor' simulate( object, newdata, nsim = 250L, level = 95, method = c("surrogate", "bootstrap", "tsbootstrap", "bayesian"), seed = 123, venv_path = "./venv", ... )## S3 method for class 'regressor' simulate( object, newdata, nsim = 250L, level = 95, method = c("surrogate", "bootstrap", "tsbootstrap", "bayesian"), seed = 123, venv_path = "./venv", ... )
object |
A regressor object |
newdata |
New data to predict on |
nsim |
Number of simulations for bootstrap/tsbootstrap |
level |
Confidence level for prediction intervals |
method |
Method for computing prediction intervals |
seed |
Seed for the random number generator |
... |
Additional arguments |