MLS-C01 Question 155
Select 3You are training a machine learning model on Amazon SageMaker to predict product demand for a retail store. The dataset contains millions of rows, and the training process is taking a significant amount of time. To optimize the training duration while ensuring the model still performs well, which of the following approaches would be most appropriate?
- A
Use the SageMaker built-in algorithm that supports distributed training and configure it to use multiple instances.
- B
Enable automatic model tuning (Hyperparameter Optimization) in SageMaker to find the best hyperparameters efficiently.
- C
Reduce the dataset size by removing random rows, as this will always improve training speed without affecting model performance.
- D
Use Amazon SageMaker Debugger to profile the training job and identify bottlenecks in resource utilization.
- E
Switch to a larger instance type for training to increase computational power.
Show answer and explanation
Correct answers: A, B, D
Explanation
To optimize training duration for a large dataset without compromising model performance, it is best to use distributed training with multiple instances, enable automatic hyperparameter optimization, and leverage SageMaker Debugger to identify and resolve training bottlenecks. These approaches address both computational scalability and performance optimization. Simply removing data or solely increasing instance size is not necessarily effective or recommended.
- A. Correct.
Using a built-in algorithm that supports distributed training and multiple instances allows for parallel processing, which can significantly reduce training time for large datasets.
- B. Correct.
Automatic model tuning (Hyperparameter Optimization) helps find the best hyperparameters efficiently, which can improve model performance without requiring exhaustive manual tuning.
- C. Incorrect.
Randomly removing rows from the dataset can lead to loss of valuable information, potentially degrading the model's performance. This is not a recommended approach.
- D. Correct.
Amazon SageMaker Debugger can help profile the training job, identify bottlenecks such as inefficient resource usage, and suggest optimizations to improve the training process.
- E. Incorrect.
Switching to a larger instance type may increase computational power, but it is not always the most efficient or cost-effective approach compared to using distributed training or profiling the job for bottlenecks.