Skip to content

FRM Part I · FRM Exam Part I · Machine Learning and Prediction

During training of a deep neural network for credit scoring, the training loss keeps decreasing while the validation loss begins rising after epoch 40. Which action is most appropriate?

Apply early stopping or regularization such as dropout or weight decay. Falling training loss with rising validation loss indicates overfitting, so capacity should be restrained rather than increased. More layers or epochs worsen it, and discarding the validation set removes the ability to monitor generalization.

  1. AApply early stopping or regularization such as dropout or weight decayCorrect
  2. BIncrease the number of hidden layers to raise model capacity
  3. CTrain for many more epochs until training loss reaches zero
  4. DAdd the validation data to the training set and stop monitoring it

Explanation

Diverging training and validation losses signal overfitting. Early stopping, dropout, or weight decay limit the model's ability to fit noise. Adding layers or training longer increases overfitting, and merging validation data removes the means of detecting it.

Did you get it right without looking?

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

Practise Machine Learning and PredictionFree diagnostic mock

More Machine Learning and Prediction questions