nerdexam
Microsoft

DP-100 · Question #521

Drag and Drop Question You are managing an Azure Machine Learning workspace. You must tune a hyperparameter for a neural network model. The learning rate must be a continuous hyperparameter between 0.

The correct answer is Uniform; Choice. The question requires matching hyperparameter types- continuous range versus discrete list- to their appropriate search space distributions in Azure Machine Learning.

Explore data and run experiments

Question

Drag and Drop Question You are managing an Azure Machine Learning workspace. You must tune a hyperparameter for a neural network model. The learning rate must be a continuous hyperparameter between 0.001 and 0.1. The batch size can be 32, 64, or 128. You need to select the appropriate search space for each parameter. Which search space should you use? To answer, move the appropriate search spaces to the correct hyperparameters. You may use each search space option once, more than once, or not at all. You may need to move the split bar between panes or scroll to view content. NOTE: Each correct selection is worth one point. Answer:

Exhibit

DP-100 question #521 exhibit

Answer Area

Drag items

ChoiceNormalQNormalUniform

Correct arrangement

  • Uniform
  • Choice

Explanation

The question requires matching hyperparameter types- continuous range versus discrete list- to their appropriate search space distributions in Azure Machine Learning.

Approach. To correctly answer, 'Uniform' must be dragged to 'Learning rate', and 'Choice' must be dragged to 'Batch size'.

  • The problem states that the 'Learning rate' must be a 'continuous hyperparameter between 0.001 and 0.1'. The 'Uniform' search space is used when a hyperparameter can take any value within a specified continuous range, with each value having an equal probability of being selected. This perfectly matches the definition of the learning rate requirement.

  • The problem states that the 'Batch size' 'can be 32, 64, or 128'. This is a discrete list of specific, predefined values. The 'Choice' search space is designed for selecting a value from a finite, explicit list of options. Therefore, 'Choice' is the correct search space for the batch size.

Common mistakes.

  • common_mistake. Using 'Normal' or 'QNormal' would be incorrect. 'Normal' (or 'QNormal' for quantized normal) search spaces are used for continuous hyperparameters that are expected to follow a normal (Gaussian) distribution, where values closer to the mean are more likely. The 'Learning rate' is specified as a simple continuous range with no mention of a normal distribution. Using 'Normal' or 'QNormal' for 'Batch size' would also be wrong because batch size is defined as a fixed set of discrete values, not a continuous range that might follow a distribution.

Concept tested. Hyperparameter search spaces and distributions for automated machine learning (AutoML) or hyperparameter tuning in Azure Machine Learning. Specifically, understanding when to use discrete (Choice) versus continuous (Uniform, Normal) search spaces.

Topics

#Hyperparameter Tuning#Azure Machine Learning#Search Space Definition#Neural Networks

Community Discussion

No community discussion yet for this question.

Sign up to join the discussion
Full DP-100 Practice