CPSC 352 -- Artificial Intelligence
Notes: Machine Learning
Introduction
Machine learning is an important topic in AI.
There are several different approaches. We begin
by discussing classical (search-based) approaches.
We also cover the neural-network approach and genetic
algorithms.
Partial Taxonomy of Learning Research
- Learning from instruction or analogy (task acquisition or revision)
- Learning from examples (supervised inductive learning)
- Learning from observation and discovery (unsupervised)
Some Knowledge Types: Representation Languages
- Concept Space: Version Space Search
- Decision trees (ID3 Algorithm)
- Production Rules (Soar)
- Parameter adjustment (connectionist learning)
- Genetic Algorithms
Some Types of Algorithms
- Inductive Learning -- ID3 Algorithm & Version Space Search
- Analytical Learning -- Soar (Explanation Based Learning)
- Connectionist Learning -- The Delta Rule
Historical Highlights
- Perceptrons (Rosenblatt, 1958)
- Genetic algorithms (Friedberg, 1958)
- Checkers (Samuels, 1959)
- Rule based systems (Feigenbaum, 1963)
- Knowledge Intensive Systems (mid 1970s)
Sub Topics