Skip to content

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

A bank compares a neural network with a regularized linear model for predicting corporate defaults using a modest dataset of 800 firms and 15 predictors. Which statement best describes a risk of using a deep neural network here?

A deep neural network is prone to overfitting with a small dataset. Its many parameters let it fit noise in 800 observations, so out-of-sample performance may be poor. Networks do capture nonlinearity and use gradient-based training, and they do not require a normally distributed target.

  1. AIt is prone to overfitting given limited data relative to its many parametersCorrect
  2. BIt cannot model any nonlinear relationships between predictors
  3. CIt requires the target variable to be normally distributed
  4. DIt cannot be trained using gradient-based methods

Explanation

Deep networks have many parameters, so with a small dataset they can memorize noise and overfit. They do model nonlinearity well, impose no normality requirement on the target, and are trained by gradient-based backpropagation.

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