Nearest Neighbor is a well-known algorithm extensively studied by the Pattern Recognition and Machine Learning communities and widely exploited in Case Based Reasoning applications. The notion of metric is central to Nearest Neighbor's working and different feature weighting metrics have been proposed in order to increase its performance.  Probability Based Metrics are metrics for classification tasks that are based on estimates of the posterior probabilities. Some of these metrics derive from optimality criteria, like the Short & Fukunaga metric. In this talk I will describe an original Probability Based Metrics called Minimum Risk Metric. Experimental comparisons with Short & Fukunaga, Value Difference, and Euclidean metrics on benchmark datasets are presented. Finally, a possible use of the metric in Case Based Reasoning applications is proposed.

Return to the seminar schedule