Accession Number : AD1033819


Title :   Distance Metric Tracking


Descriptive Note : Technical Report


Corporate Author : MIT Lincoln Laboratory Lexington United States


Personal Author(s) : Kelley,Stephen J ; Greenewald,Kristjan ; Hero,Alfred III O


Full Text : https://apps.dtic.mil/dtic/tr/fulltext/u2/1033819.pdf


Report Date : 02 Mar 2016


Pagination or Media Count : 19


Abstract : Recent work in distance metric learning has produced numerous methods aimed at learning transformations of data that best align with provided sets of pairwise similarity and dissimilarity constraints. The learned transformations lead to improved retrieval, classification, and clustering algorithms due to the more accurate distance or similarity measures. Here, we introduce the problem of learning these transformations when the underlying constraint generation process is dynamic.These dynamics can be due to changes in either the ground-truth labels used to generate constraints or changes to the feature subspaces in which the class structure is apparent. We propose and evaluate an adaptive, online algorithm for learning and tracking metrics as they change over time. We demonstrate the proposed algorithm on both real and synthetic data sets and show significant performance improvements relative to previously proposed batch and online distance metric learning algorithms.


Descriptors :   machine learning , artificial intelligence , algorithms , information processing , signal processing , information systems , theorems , batch processing , clustering , data mining , game theory , surface truth


Subject Categories : Numerical Mathematics
      Cybernetics


Distribution Statement : APPROVED FOR PUBLIC RELEASE