Autonomous Learning Classifiers
Plamen Parvanov Angelov · 2012
One traditional approach to classifying data streams is the incremental classifier. In practice, nowadays, the classifiers need to cope with large quantities of data, often streaming with a fast rate. The challenges that classifiers face are related to the need to address nonlinearity and nonstationarity, large or even huge amounts of data, and real-time, recursive processing. The chapter discusses the autonomous self-learning classifier family AutoClassify. A classifier is a mapping from the feature space to the class label space. Learning AutoClassify0 is unsupervised and is based on focal points by clustering or partitioning into data clouds. Learning of AutoClassify1 is very similar to the autonomous learning system (ALS) of type B, which usually also consists of first order submodels. Controlled Vocabulary Terms learning systems; modelling; pattern classification; streaming media