Which technique increases the size and diversity of the training data through small automatic transformations?

Which technique increases the size and diversity of the training data through small automatic transformations?

• A. Data augmentation
• B. Data sampling
• C. Data reduction
• D. Data cleaning

Answer: (A)

Data augmentation is a technique that expands the training set by applying small, automated transformations to existing data. In practice, especially with images, this means flipping, rotating, cropping, or adjusting color and brightness to create new labeled examples from the originals. These modest changes increase both the size and the variety of the data the model sees, helping it learn to recognize patterns under different appearances and conditions, which improves generalization and reduces overfitting. Other approaches don’t add new examples in the same way. Data sampling selects or reweights a subset of the data, which doesn’t increase overall size or diversity. Data reduction aims to shrink the dataset. Data cleaning focuses on improving data quality by removing or correcting errors, not on creating new training instances.

Data augmentation is a technique that expands the training set by applying small, automated transformations to existing data. In practice, especially with images, this means flipping, rotating, cropping, or adjusting color and brightness to create new labeled examples from the originals. These modest changes increase both the size and the variety of the data the model sees, helping it learn to recognize patterns under different appearances and conditions, which improves generalization and reduces overfitting.

Other approaches don’t add new examples in the same way. Data sampling selects or reweights a subset of the data, which doesn’t increase overall size or diversity. Data reduction aims to shrink the dataset. Data cleaning focuses on improving data quality by removing or correcting errors, not on creating new training instances.