Skip to content

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

During training of a deep neural network for credit scoring, the training error keeps falling while the validation error begins to rise after epoch 40. Which action is most directly aimed at addressing this problem?

Apply early stopping. Falling training error with rising validation error indicates overfitting, and stopping training near the point where validation error is lowest limits it. More layers would worsen overfitting, and discarding the validation set would remove the ability to detect it.

  1. AApply early stopping at around epoch 40Correct
  2. BIncrease the number of hidden layers and nodes
  3. CRaise the learning rate so training converges faster
  4. DRemove the validation set and train on all the data

Explanation

Diverging training and validation errors signal overfitting. Early stopping halts training when validation error stops improving, limiting overfitting. Adding capacity worsens overfitting, and removing the validation set removes the means to detect it. A higher learning rate does not address overfitting.

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