MLS-C01 Question 213
Single answerYou are training a linear regression model using the Amazon SageMaker built-in Linear Learner algorithm. During training, you notice that the loss function is decreasing very slowly. Which of the following actions is most likely to resolve the issue?
- A
Increase the learning rate parameter.
- B
Decrease the learning rate parameter.
- C
Increase the number of epochs.
- D
Switch to a non-linear model.
Show answer and explanation
Correct answer: A
Explanation
The learning rate is a critical hyperparameter that controls the size of the steps the model takes during the optimization process. If the learning rate is set too low, the model will make very small updates to weights, causing the loss function to decrease slowly. By increasing the learning rate, the model can converge faster to the optimal solution. However, it's essential to monitor training closely, as setting the learning rate too high could cause the model to overshoot the optimum or diverge.
- A. Correct.
Increasing the learning rate parameter allows the model to take larger steps in the direction of minimizing the loss, which can speed up convergence. However, care must be taken to avoid setting it too high, as this can lead to instability.
- B. Incorrect.
Decreasing the learning rate makes the model take smaller steps, which can slow down the training process further. This option would not resolve the issue of slow convergence.
- C. Incorrect.
Increasing the number of epochs might improve the training over time but does not address the root cause of the slow progress in each iteration.
- D. Incorrect.
Switching to a non-linear model is unrelated to the issue described, as the problem lies in the training dynamics of the current linear model, not the type of model being used.