Answer-first summary for fast verification
Answer: Use UNIT_LINEAR_SCALE for the embedding dimension, UNIT_LOG_SCALE for the learning rate, and a small number of parallel trials.
The correct answer is B. Here's why: For the embedding dimension, UNIT_LINEAR_SCALE is appropriate for integer hyperparameters with a continuous range like the embedding dimension. It linearly scales the search space from minValue to maxValue. For the learning rate, UNIT_LOG_SCALE is generally recommended for hyperparameters with values spanning multiple orders of magnitude, from 10e-05 to 10e-02. This ensures equal sampling probability across different log-scaled ranges. Since training time is not a concern, using a small number of parallel trials helps maintain the effectiveness of the Bayesian optimization algorithm, which can be compromised by too many parallel trials.
Author: LeetQuiz Editorial Team
Ultimate access to all questions.
You have recently created a proof-of-concept (POC) deep learning model for a machine learning project. The model architecture meets your expectations, but you need to fine-tune two hyperparameters for optimal performance: the embedding dimension for a categorical feature and the learning rate. To achieve this, you decide to use hyperparameter tuning on Vertex AI with the default Bayesian optimization algorithm. You configure the following hyperparameters: • Embedding dimension set as an INTEGER with a minValue of 16 and maxValue of 64. • Learning rate set as a DOUBLE with a minValue of 10e-05 and maxValue of 10e-02. Your primary goal is to maximize model accuracy, and you are not concerned about the time it takes to train the model. How should you set the hyperparameter scaling for each hyperparameter and the maxParallelTrials?
A
Use UNIT_LINEAR_SCALE for the embedding dimension, UNIT_LOG_SCALE for the learning rate, and a large number of parallel trials.
B
Use UNIT_LINEAR_SCALE for the embedding dimension, UNIT_LOG_SCALE for the learning rate, and a small number of parallel trials.
C
Use UNIT_LOG_SCALE for the embedding dimension, UNIT_LINEAR_SCALE for the learning rate, and a large number of parallel trials.
D
Use UNIT_LOG_SCALE for the embedding dimension, UNIT_LINEAR_SCALE for the learning rate, and a small number of parallel trials.
No comments yet.