Mini-Batch Gradient Descent
Domain: AI Fundamentals
🎯 Motivation
Update weights using small batches of data - a balance between speed and stability.
📋 Concrete Example
Batch size 32: each update uses 32 examples. Faster than full-batch, smoother than one-at-a-time.
🔗 Analogy
Like a teacher grading 32 papers at a time to estimate the class average, instead of all at once or one by one.