Case-Based Reasoning CBR Cycle CBR Problem Issues Knowledge Representation Case Retrival Case Adaptation Strategies Learning Mechanism
Knowledge Representation The Dynamic Memory Model Memory Organization Packets (MOP) Norms (common features) Cases Indexes (to discriminate features…) The Category & Exemplar Model Categories Examplars Features Network
Knowledge Representation The Category & Exemplar Model Categories Exemplars Features Network
Case Retrieval Memory (Case-Base) Organization Search Strategies Match Index Schemes Match Weight Matching Algorithm (Nearest Neighbors)
1. Memory Organization and Retrieval Algorithm Flat Memory, Serial Search Shallow indexing Partitioning of the case library Parallel retrieval Shared Feature Networks(Based on MOP), tree/graph Provide a means of clustering cases so that cases that share many features are clustered together. (Clustering-Discrimination)
1. Memory Organization and Retrieval Algorithm Discrimination Networks Discrimination first, Clustering second. Index Scheme Does a system need Index Scheme ? How to organize index (Scheme)? What kinds of features can be used as indexes (Discrimination)?
Nearest Neighbor Matching (Weight Matching) Algorithm NumericEvaluation: nk=1 Wk * Sim( fIk, fRk ) nk=1 Wk Wk : Weight of feature k Sim : similarity function of the primitives fIk, fRk : the values for feature fk in the input and retrieved cases.
Adaptation Methods and Strategies Simple Substitution Parameter Adjustment Global Constraint Satisfactions
Learning in CBR--Learning from Problem Solving What can be learned ? New experience (New case) Strategies: Storing cases in the case base