Outline
-
A general model of learning agent
-
The learning element
-
The performance element
-
The learning paradigms
-
The subsymbolic paradigm
-
The symbolic paradigm
-
The learning strategies
-
Inductive learning
-
Deductive learning
-
Abductive learning
-
Multistrategy learning
-
The representation problem
-
Feature vectors and attribute-based representations
-
Structural representations
-
Machine learning issues
-
Inductive learning for attribute-based representations
-
Statistical methods
-
Parametric methods
-
Non-parametric methods
-
Relevance for document analysis and recognition
-
Assessing the performance of an inductive learning algorithm
-
Decision trees
-
Main problems in decision tree induction
-
The decision tree learning algorithm
-
Handling continuous attributes
-
Measures for attribute selection
-
Pruning to avoid overfitting the data
-
From decision trees to rules
-
Case study: C4.5
-
Further issues in decision tree learning
-
Incremental learning of decision trees
-
Case study: ITI
-
Relevance for document analysis and recognition
-
Inductive learning for structural representations
-
Concept learning as search
-
The instance space and the hypotheses space
-
General-to-specific ordering of hypotheses
-
Generalization
-
Specialization
-
Sequential covering algorithms
-
Learning first-order rules
-
First-order Horn clauses
-
Mapping clauses to graphs and viceversa
-
Generalization
-
Specialization
-
Top-down and bottom-up learning algorithms
-
Case study: FOIL
-
Using prior knowledge in learning
-
Case study: FOCL
-
Relevance for document analysis and recognition
-
Learning in document analysis and recognition: open
problems and research directions.