- Learning and Concepts
- Why Machine Learning
- Types of Learning
- Neural Networks
- Genetic Algorithm
- Fuzzy Learning
- Boltzman Machines
- Acquiring new knowledge
- Modifying existing knowledge
- Changes
- Generalization
- Improvement
Inductive Learning
Deductive Learning
Find patterns from data
Derive conclusion from facts
A=B,B=C -> A=C
Memorization
Learning by observing and analyzing human solutions to specific problems
After lerning, go directly from facts to solution
Prototypical EBL Architecture
- Needs many examples
- Broken Explanations?
- Inductive Learning
- Find Patterns
- Nature and Quality of Information
- High Level -> Abstract -> Broad Class of Problems
- Low Level -> Details -> Single Problem
- Noise Free
- Reliable
- Ordered
- Expressive
- Modifiable
- Extensibility
- Crossover
- Mutation
- Survival of the fittest
- Survival of the most diverse
- Heredity
- Variation
- Selection
A) generate initial population
B) evaluate fitness of all
C) select fitter population for crossover
D) cross-over
E) mutate
F) evaluate fitness of offspring
Loop while reaching desired fitness
Roulette Wheel
- Less efficient in terms of speed of convergence
- Tendency to get stuck in local maxima
Crisp
Fuzzy
Fuzzy Rule-Based System
Fuzzy Logic Operators
Fuzzy Logic Operators
De-Fuzzify Result
Research