ssc (version 1.0)

democratic: Train the Democratic model

Description

Builds and trains a model to predict the label of instances, according to Democratic algorithm.

Usage

democratic(x, y, bclassifs, dist = "matrix", dist.use = NULL)

Arguments

x

A object that can be coerced as matrix. This object have various interpretations depending on the value set in dist argument. See dist argument.

y

A vector with the labels of training instances. In this vector the unlabeled instances are specified with the value NA.

bclassifs

A list with the classifiers specification. For defining new base classifiers see bClassif.

dist

Distance information. Valid options are:

• "matrix": this string indicates that x is a distance matrix or a list of distance matrices.
• string: the name of a distance method available in proxy package. In this case x is interpreted as a matrix of instances.
• function: a function defined by the user that computes the distance between two vectors. This function is called passing the vectors in the firsts two arguments. If the function have others arguments, those arguments must be have default values. In this case x is interpreted as a matrix of instances.

dist.use

Boolean matrix that indicates which base classifier uses each distance in case that more than one distance is supplied. Each row represents a classifier and each column a distance.

Value

The trained model stored in a list with the following named values:

References

Yan Zhou and Sally Goldman. Democratic co-learning. In IEEE 16th International Conference on Tools with Artificial Intelligence (ICTAI), pages 594–602. IEEE, Nov 2004. doi: 10.1109/ICTAI.2004.48.

Examples

# Note: democratic assumes that the classifiers provided are from different
# learning paradigms. In this examples we simulated this requirement by using two
# 1-NN classifiers with different distance measures instead of two different
# learning paradigms.

library(ssc)

## Load Wine data set
data(wine)

x <- wine[, -14] # instances without classes
y <- wine[, 14] # the classes
x <- scale(x) # scale the attributes

## Prepare data
set.seed(20)
# Use 50% of instances for training
tra.idx <- sample(x = length(y), size = ceiling(length(y) * 0.5))
xtrain <- x[tra.idx,] # training instances
ytrain <- y[tra.idx]  # classes of training instances
# Use 70% of train instances as unlabeled set
tra.na.idx <- sample(x = length(tra.idx), size = ceiling(length(tra.idx) * 0.7))
ytrain[tra.na.idx] <- NA # remove class information of unlabeled instances

# Use the other 50% of instances for inductive testing
tst.idx <- setdiff(1:length(y), tra.idx)
xitest <- x[tst.idx,] # testing instances
yitest <- y[tst.idx] # classes of testing instances

## Example
bclassifs <- list(bClassifOneNN(), bClassifOneNN())
dist <- list("Euclidean", "Manhattan")
dist.use <- matrix(
  data = c(
#   Euclidean Manhattan
    TRUE,     FALSE,    # First  1-NN
    FALSE,    TRUE      # Second 1-NN
  ),
  nrow = 2, byrow = TRUE
)
m <- democratic(x = xtrain, y = ytrain,
                bclassifs, dist, dist.use)
pred <- predict(m, xitest)
caret::confusionMatrix(table(pred, yitest))

## Example: Using distance matrices instead of the instances
dtEuclidean <- proxy::dist(x = xtrain, method = "Euclidean", by_rows = TRUE)
dtManhattan <- proxy::dist(x = xtrain, method = "Manhattan", by_rows = TRUE)
m2 <- democratic(x = list(dtEuclidean, dtManhattan), y = ytrain,
                bclassifs, dist = "matrix", dist.use = dist.use)

ditEuclidean <- proxy::dist(x = xitest, y = xtrain[m2$included.insts,],
                            method = "Euclidean", by_rows = TRUE)
ditManhattan <- proxy::dist(x = xitest, y = xtrain[m2$included.insts,],
                            method = "Manhattan", by_rows = TRUE)
pred <- predict(m2, ditEuclidean, ditManhattan)
caret::confusionMatrix(table(pred, yitest))