Plotting a learning curve from image recognition data.

learning-curve.png

learning-curve.R

library(caret)
library(reshape2)

dataTrain <- data.frame(x = trainData$x[1:30000,], y = trainData$y[1:30000])
dataTest <- data.frame(x = testData$x, y = testData$y)

# Train.
trainctrl <- trainControl(verboseIter = TRUE, number=5, repeats=1, method='repeatedcv')

results <- data.frame()
for (i in 1:30) {
  partialSet <- dataTrain[1:(1000 * i),]
  
  fit <- train(y ~ ., data=partialSet, method = 'LogitBoost', trControl = trainctrl)

  correct1 <- length(which(predict(fit, partialSet) == partialSet$y)) / nrow(partialSet)
  correct2 <- length(which(predict(fit, dataTest) == dataTest$y)) / nrow(dataTest)
  
  # Record accuracy history.
  results <- rbind(results, c(correct1, correct2))
  
  # Plot learning curve.
  names(results) <- c('Train', 'CV')
  r <- melt(results)
  r <- cbind(r, seq(from = 1000, to = nrow(results) * 1000, by = 1000))
  names(r) <- c('Set', 'Accuracy', 'Count')
  print(ggplot(data = r, aes(x = Count, y = Accuracy, colour = Set)) + geom_line() + geom_smooth(method = 'lm', se=F))
}