Theory of Intelligence
MACHINE LEARNING•Foundations•The Machine Learning Landscape•1.3 Batch vs Online Learning

Introduction

Learning from all the data at once versus learning continuously from a stream, the learning rate, and why models go stale (data drift).

Introduction

Think of two ways to stay up to date with the news:

  • Once a year, you sit down and read every newspaper from the past year in one go. In between, your knowledge doesn't change.
  • Every morning, you read today's paper and update what you know, a little at a time.

Machine learning systems face the same choice. They can learn from all the data at once (batch learning) or continuously, a bit at a time as new data arrive (online learning). The choice affects cost, speed, and how quickly the system notices that the world has changed.

Discuss theoretical concepts, mathematical formulas, and architectural foundations.

Supports inline code with `backticks`
No comments yet. Be the first to start the discussion!