A dataset contains features with very different scales — age (0 to 100) and annual salary (0 to 200,000). A gradient descent-based algorithm is converging slowly. Which preprocessing step will most directly improve convergence?

Question

Accepted Answer

C. Feature scaling such as normalization or standardization. Feature scaling ensures all numeric features are on a similar scale, preventing features with large magnitudes from dominating gradient updates and significantly improving convergence speed for gradient descent-based algorithms.

A dataset contains features with very different scales — age (0 to 100) and annual salary (0 to 200,000). A gradient descent-based algorithm is converging slowly. Which preprocessing step will most directly improve convergence?

Related Questions

Discussion