Course Outline

• Optimal Bayesian classifiers: all probabilities are known.

• When probabilities are not known we must learn them from the data:

  • Parametric methods.
  • Non-parametric methods.

• Linear classifiers.

• Unsupervised learning.

• The relations between learning and compression:

  • Information theory and data compression.

• Patterns and randomness: Kolmogorov complexity.

• Sample complexity: how fast can we learn?

VC-theory.

If time allows: Bayesian networks, Support vector machines.

השקופית הקודמת

השקופית הבאה

חזור אל השקופית הראשונה

הצג גירסה גרפית