Skip to content

FRM Part I · FRM Exam Part I · Machine-Learning Methods

A deep neural network for credit scoring achieves very low training error but much higher validation error. Which single action is most directly aimed at reducing this problem?

Apply regularization such as dropout or weight penalties, together with early stopping on validation error. The gap between low training error and high validation error signals overfitting, which these methods reduce, whereas more capacity or longer training would make it worse.

  1. AAdd dropout or weight regularization and apply early stopping based on validation errorCorrect
  2. BIncrease the number of hidden layers and neurons to raise model capacity
  3. CTrain for more epochs until the training error reaches zero
  4. DRemove the validation set and use all data for training

Explanation

Low training error with high validation error indicates overfitting. Dropout, L1/L2 penalties and early stopping constrain effective complexity or halt training when validation error rises. More capacity or more epochs typically worsens overfitting, and removing the validation set removes the ability to detect it.

Did you get it right without looking?

One question tells you little. A timed set on Machine-Learning Methods shows your real accuracy, how long you take and where you lose marks.

Practise Machine-Learning MethodsFree diagnostic mock

More Machine-Learning Methods questions