PAC — Probably Approximately Correct Learning

Fact

A framework for saying when a model learns well enough, with probability.

In Plain Words

PAC learning is like a frozen pizza brand. Not perfect, but it should not ruin Friday night often.

It is a basic ruler for learning theory. It checks if you have enough examples. It asks if the model will work on new cases.

Related Concepts

• Statistical Learning Theory
• Supervised Learning
• Empirical Risk Minimization
• Bias-Variance Tradeoff

SLT
PAC is one classic way SLT defines learning.

Supervised Learning
PAC often studies when Supervised Learning can learn from examples.

ERM
PAC explains why low training error may not mean good new-case performance.

Bias-Variance Tradeoff
Both ask how model complexity affects new-case performance.