Kiru Lab  /  Classical Machine Learning

Generalization and Its Failures

Overfitting, capacity, leakage — the honest evaluation of a model.

In this module you will

  • Explain overfitting as capacity exceeding evidence
  • Design a split that survives temporal and group structure
  • Recognize leakage before it flatters your results
Overfitting and Capacity
A model with enough capacity will memorize noise. Memorization looks like success on the data you trained on.
Concept · 30m
Splits, Leakage, and Honest Numbers
Most impressive results that fail in production were leaking. Learn the failure patterns.
Concept · 30m
← Classical Machine Learning