Bias-variance tradeoff demo

a shiny app

Carto Wong
mathematician

What is bias-variance tradeoff?

In statistics and machine learning, the bias-variance tradeoff is a problem of trying to minimize two sources of error.

Bias refers to the distance between the mean prediction and the true value.
Variance refers to the variance of the prediction upon re-sampling of the training set.

trueValue <- 26
predictions <- c(24, 28, 17, 26, 23)
meanPredication <- mean(predictions)
(meanPredication - trueValue)^2  # bias squared

## [1] 5.76

mean((predictions - meanPredication)^2)  # variance

## [1] 13.84

\(k\)-nearest neighbors algorithm

The k-nearest neighbors (regression) algorithm is a non-parametric method which makes prediction using the average of the values of the \(k\)-nearest neighbors.

Shiny app demo of bias-variance tradeoff

This shiny app can be used to explore the bias-variance tradeoff.

Try it yourself

Fix \(n = 30\). Generate two density plots using the shiny app for \(k = 1\) and \(30\) to see what happens.