MLA-C01 exam dumps

MLA-C01 practice question 97 of 458

AWS Certified Machine Learning Engineer - Associate. Associate level, Amazon Web Services. Free question with the correct answer and a full explanation.

MLA-C01 Question 97

Select 3

You are training a machine learning model to classify medical images into categories based on disease type. Upon analyzing your dataset, you discover a significant class imbalance, with certain disease categories having far fewer images than others. What are the most appropriate strategies to address this issue?

  1. A

    Generate synthetic images for the underrepresented classes using techniques like GANs (Generative Adversarial Networks).

  2. B

    Resample the dataset by oversampling the underrepresented classes or undersampling the overrepresented classes.

  3. C

    Use class weights during model training to penalize the model for misclassifying underrepresented classes.

  4. D

    Remove some images from the overrepresented classes to ensure an even class distribution.

  5. E

    Increase the number of epochs during training to allow the model to focus more on underrepresented classes.

Show answer and explanation

Correct answers: A, B, C

Explanation

Addressing class imbalance in datasets is crucial for training effective machine learning models. Techniques like synthetic data generation, resampling, and using class weights during training are well-established strategies to mitigate the impact of class imbalance. Removing data from overrepresented classes is generally avoided as it can lead to a loss of important information. Similarly, increasing the number of epochs does not specifically address the core issue of imbalance and can introduce overfitting.

  • A. Correct.

    Generating synthetic images using techniques like GANs can help augment the dataset for underrepresented classes, reducing class imbalance without losing valuable data.

  • B. Correct.

    Resampling, either by oversampling the underrepresented classes or undersampling the overrepresented classes, is a common strategy to address class imbalance in datasets.

  • C. Correct.

    Using class weights during training is an effective way to handle class imbalance, as it helps the model focus more on underrepresented classes by penalizing misclassifications.

  • D. Incorrect.

    While removing data from overrepresented classes can balance the dataset, it risks losing valuable information and reducing the overall dataset size, which is generally not recommended.

  • E. Incorrect.

    Increasing the number of epochs does not directly address class imbalance. It can lead to overfitting, especially if the dataset contains significant imbalance.

Timed practice exam

Take a MLA-C01 practice test under exam conditions

65 questions in 130 minutes, drawn from this bank, with a score report and a per-question review when you finish.

Start timed exam