massOfai

Batch Size

AI/ML Fundamentals

Number of samples processed before updating weights

What is Batch Size?

Larger batches: more stable gradients but more memory. Smaller: faster updates but noisier.

Real-World Examples

  • •Batch size 32 for most neural networks
  • •Batch size 1 = stochastic gradient descent

When to Use This

Important hyperparameter affecting training speed and convergence

Related Terms

Learn more about concepts related to Batch Size

Training

Process of teaching an ML model using data

Gradient Descent

Optimization algorithm minimizing loss

Epoch

One complete pass through training data

Hyperparameter

Configuration setting for learning algorithm