AWS Certified AI PractitionerFoundation ModelsEasy
A team of researchers is training a new, very large foundation model with billions of parameters on a massive, diverse dataset. During the training process, they observe that the model's performance on the training data is exceptionally high, achieving near-perfect scores. However, when evaluated on a separate, unseen validation dataset, the model's performance is significantly worse. This discrepancy indicates which common issue in machine learning?
- AUnderfitting.
- BHaving a high bias.
- CConverging too slowly.
- DOverfitting.
Show answer & explanationAnswer & explanation
Correct answer: D. Overfitting.
The scenario describes a classic case of overfitting: the model performs extremely well on the data it was trained on ('exceptionally high' on training data) but fails to generalize to new, unseen data ('significantly worse' on validation data). This means the model has learned the training data too specifically, including noise, rather than the underlying patterns.
Why the other options are wrong
- A. Underfitting occurs when a model is too simple to capture the underlying patterns, leading to poor performance on both training and validation data.
- B. High bias is often associated with underfitting, where the model makes strong assumptions about the data, leading to errors on both training and test sets.
- C. Converging too slowly relates to training efficiency, not the generalization gap described.
Overfitting
A phenomenon in machine learning where a model learns the training data too precisely, including noise and specific examples, resulting in poor generalization to new, unseen data.
- High performance on training data, low performance on validation/test data.
- Often occurs in complex models with many parameters trained on limited data.
- Can be mitigated by regularization, more data, or simpler models.
Memory trick: Overfitting: Too much detail, loses the big picture.