Batch Normalization
55.8% acceptance
MediumNeural Networks
Commonly asked at:
GoogleWhy this matters:
Tests understanding of training stability techniques – essential knowledge for deep learning roles.
Batch normalization normalizes the inputs of each layer to have zero mean and unit variance, then applies a learnable scale and shift.
Examples
Constraints
- Normalize over batch dimension
- Include learnable gamma and beta parameters
- Handle training vs inference mode (use running averages during inference)
- Use epsilon = 1e-5 for numerical stability
Tags
distributed-trainingmlopsmodel-serving