SSLR basics

This package includes a collection of methods to create models for semi-supervised learning (e.g. fitting the model, making predictions, etc), with a fairly intuitive interface that is easy to use.

In Model list section you can see the list of different classification, regression and clustering models.

Motivation

Current packages to do semi-supervised learning do not use an intuitive interface. In this package, trying to use semi-supervised learning in an easy and intuitive way.

SSLR tries to solve this by providing an interface to use different models, mainly using the parsnip model interface to make the use of this package easier.

SSLR connects with parsnip to create different models without using too many arguments in the fit functions.

In addition, it uses other packages such as RSSL to use the same interface in an easy way.

For example, to use different ones like RSSL. It has a different interface. Thanks to SSLR you can use different options to use its fit functions.

Process

To fit the model (for example SelfTraining), you must:

Have a defined model using parsnip
Use your parameters or using by default
Call fit with formula, fit_xy with x and y, or fit_x_u with x and unlabeled data. See Model fitting section.

For example, with fit function:

rf <-  rand_forest(trees = 100, mode = "classification") %>%
  set_engine("randomForest")

m <- selfTraining(learner = rf) %>% fit(Wine ~ ., data = train)

Or with fit_xy function:

rf <-  rand_forest(trees = 100, mode = "classification") %>%
  set_engine("randomForest")

m <- selfTraining(learner = rf) %>% fit_xy(x = train[,-cls], y = train$Wine)

This uses the parsnip package that has an intuitive interface to create a Random Forest model and this can be used in the SSLR package in a simple way.