You have an Azure Machine Learning workspace. You plan to tune a model hyperparameter when you train the model. You need to define a search space that returns a normally distributed value. Which…

The correct answer is D. LogNormal. In Azure ML hyperparameter tuning, the search space sampling functions map to statistical distributions. 'LogNormal(mu, sigma)' returns a value whose logarithm follows a normal distribution - making it the distribution in the list rooted in the normal (Gaussian) curve…

Train and deploy models

Question

You have an Azure Machine Learning workspace. You plan to tune a model hyperparameter when you train the model. You need to define a search space that returns a normally distributed value. Which parameter should you use?

Options

AQUniform
BLogUniform
CUniform
DLogNormal

How the community answered

(29 responses)

A
3% (1)
B
3% (1)
D
93% (27)

Explanation

In Azure ML hyperparameter tuning, the search space sampling functions map to statistical distributions. 'LogNormal(mu, sigma)' returns a value whose logarithm follows a normal distribution - making it the distribution in the list rooted in the normal (Gaussian) curve. 'Uniform' samples evenly between two bounds. 'LogUniform' samples uniformly in log-space (the exponent of a uniform draw), producing no normal-distribution properties. 'QUniform' is a quantized (rounded) variant of Uniform. Because LogNormal is the only choice whose underlying math involves a normal distribution, it is the correct answer.

Topics

#Hyperparameter Tuning#Azure Machine Learning#Search Space#Statistical Distributions

Community Discussion

No community discussion yet for this question.

Full DP-100 Practice