IES 511 Machine Learning Dr. Türker İnce (Lecture notes by Prof. T. M IES 511 Machine Learning Dr. Türker İnce (Lecture notes by Prof. T. M. Mitchell, Machine Learning course at CMU) Concept Learning General-to-Specific Ordering of Hypothesis Find-S and Candidate Elimination Algorithms Inductive Bias
Learning System Design Example - Play Checkers
Concept Learning Example
Hypothesis representation
Concept Learning Task
Fundamental assumption of inductive learning
General-to-specific ordering of hypothesis
The Find-S Algorithm
Hypothesis space search by Find-S
Limitations of Find-S Can’t tell whether it has learned concept Can’t tell when training data inconsistent Picks a maximally specific h (why?) Depending on H, there might be several!
Consistent hypothesis and Version Space
The List-Then-Eliminate Algorithm
Version Space of EnjoySport Concept Learning
Version Space of EnjoySport Concept Learning
Candidate Elimination Algorithm
Candidate Elimination Algorithm
EnjoySport Example
EnjoySport Example
EnjoySport Example
Limitations of Candidate Elimination Training data contains errors Target concept is not in H, it can not be described in current hypothesis representation
Partially learned concept
Inductive Bias
Inductive learners modeled by equivalent deductive systems