This paper presents a new class of local similarity metrics, called AASM, that are not symmetric and that can be adopted as basic retrieval method in a CBR system. We also introduce an anytime learning procedure that starting from an initial set of stored cases improves the retrieval accuracy by modifying the local definition of the metric. The learning procedure is an implementation of the reinforcement learning paradigm and can be run as a black box as no particular setting is required. We show with the aid of classical test sets tat AASM cab improve in many cases the accuracy of both nearest Neighbour methods and Salzberg’s NGE. Moreover AASM cab achieve significant data compression (10%) still maintaining the same accuracy of NN
Learning an Asymmetric and Anisotropic Similarity Metric for Case-Based Reasoning
Ricci, Francesco;Avesani, Paolo
1995-01-01
Abstract
This paper presents a new class of local similarity metrics, called AASM, that are not symmetric and that can be adopted as basic retrieval method in a CBR system. We also introduce an anytime learning procedure that starting from an initial set of stored cases improves the retrieval accuracy by modifying the local definition of the metric. The learning procedure is an implementation of the reinforcement learning paradigm and can be run as a black box as no particular setting is required. We show with the aid of classical test sets tat AASM cab improve in many cases the accuracy of both nearest Neighbour methods and Salzberg’s NGE. Moreover AASM cab achieve significant data compression (10%) still maintaining the same accuracy of NNI documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
