In gradient-based optimization, feature scaling is crucial because it:

The correct answer is B. Stabilizes gradient magnitude across dimensions. When features have very different scales (e.g., one feature ranges 0–1 and another 0–1,000,000), the loss surface becomes elongated and anisotropic. Gradient descent on such a surface oscillates along steep directions while barely moving along shallow ones, dramatically slowing c

NVIDIA Certified Associate (NCA) Core AI Concepts

Question

Options

ARemoves outliers
BStabilizes gradient magnitude across dimensions
CIncreases dataset size
DEliminates overfitting

How the community answered

(26 responses)

A
4% (1)
B
85% (22)
C
8% (2)
D
4% (1)

Explanation

When features have very different scales (e.g., one feature ranges 0–1 and another 0–1,000,000), the loss surface becomes elongated and anisotropic. Gradient descent on such a surface oscillates along steep directions while barely moving along shallow ones, dramatically slowing convergence. Scaling (normalization or standardization) makes the loss surface more spherical so gradients have comparable magnitudes across all dimensions and the optimizer converges faster and more reliably. Removing outliers (A) is data preprocessing unrelated to gradient behavior. Scaling never increases dataset size (C). It does not directly eliminate overfitting (D) - regularization and other techniques do that.

Topics

#Feature Scaling#Gradient Descent#Optimization#Machine Learning Fundamentals

Community Discussion

No community discussion yet for this question.

Full NCA-AIIO Practice