Exam DP-100 topic 3 question 163 discussion

NO, YES, NO, NO

NO, YES, NO, NO

A is false. Grid sampling (the only way to exhaustively search all combinations of parameters) only allows for Choice(...). Reference: https://learn.microsoft.com/en-us/azure/machine-learning/how-to-tune-hyperparameters?view=azureml-api-2

the keep_probability parameter value will always be either 0.05 ot 0.1: NO. The keep_probability parameter value will not always be either 0.05 or 0.1 based on the code. The keep_probability is defined with a Uniform distribution ranging from 0.05 to 0.1. This means that during hyperparameter tuning, the value of keep_probability can take on any value between 0.05 and 0.1, and it will not necessarily be limited to just these two specific values. The Uniform distribution will sample values uniformly within the specified range, so you may get values like 0.06, 0.07, 0.08, and so on during the hyperparameter tuning experiment. It will not be constrained to only 0.05 or 0.1.

A. NO -> As Uniform option is used in "keep_probability" param, Random or Bayesian should be used, and NOT all the combinations will be tested. The rest are OK: B.YES, C.NO, D.NO

according to chat gtp: a. No. By defining the sampling in this manner, not every possible combination of the parameters will be tested. Instead, the hyperparameter tuning experiment will use different sampling techniques to explore the hyperparameter space efficiently. b. Yes. Random values of the learning_rate parameter will be selected from a normal distribution with a mean of 10 and a standard deviation of 3. c. Yes. The keep_probability parameter value will always be either 0.05 or 0.1 because it is defined as a uniform distribution between these two values. d. No. The number_of_hidden_layers parameter is defined as a choice between the values of 3 and 4, not as a normal distribution.