Semi-supervised learning method system for data classification according to discriminating parameters

摘要

The method involves constructing a local module storing data represented by a prototype, and classifying the prototype with a label. A generative graph is defined, and weight is allocated to arcs of the graph. Labels are spread in the graph until a convergence state, where propagation of the label of the prototype towards another prototype modifies probabilities of the label of latter prototype based on the weight and probabilities of the label of the former prototype. The data is classified based on the label of prototypes representing the data after the propagation of the label. Independent claims are also included for the following: (1) a computer program comprising instructions for executing a semi-supervised learning method for data classification (2) a recording medium comprising a computer program with instructions for executing a semi-supervised learning method for data classification.