Dynamic Classifier and Sensor Using Small Memory Buffers

A model presented in current paper designed for dynamic classifying of real time cases received in a stream of big sensing data. The model comprises multiple remote autonomous sensing systems; each generates a classification scheme comprising a plurality of parameters. The classification engine of each sensing system is based on small data buffers, which include a limited set of “representative” cases for each class (case-buffers). Upon receiving a new case, the sensing system determines whether it may be classified into an existing class or it should evoke a change in the classification scheme. Based on a threshold of segmentation error parameter, one or more case-buffers are dynamically regrouped into a new composition of buffers, according to a criterion of segmentation quality.