A model generalizes poorly even with abundant training data. Which factor is MOST likely responsible?

The correct answer is B. Data distribution mismatch. When a model has abundant training data yet still generalizes poorly, the most common culprit is a distribution shift - the training data does not represent the deployment (test) distribution. This is known as covariate shift or domain shift. No amount of additional training…

NVIDIA Certified Associate (NCA) Core AI Concepts

Question

Options

AOver-regularization
BData distribution mismatch
CLow learning rate
DGradient clipping

How the community answered

(45 responses)

A
2% (1)
B
89% (40)
C
2% (1)
D
7% (3)

Explanation

When a model has abundant training data yet still generalizes poorly, the most common culprit is a distribution shift - the training data does not represent the deployment (test) distribution. This is known as covariate shift or domain shift. No amount of additional training data helps if that data is systematically different from what the model encounters in production. Option A (over-regularization) would cause underfitting, not a generalization gap specific to abundant data. Option C (low learning rate) slows convergence but does not inherently cause poor generalization. Option D (gradient clipping) is a training stability technique that does not explain poor generalization with large data.

Topics

#Model Generalization#Data Distribution Mismatch#Machine Learning Problems

Community Discussion

No community discussion yet for this question.

Full NCA-AIIO Practice