Design of unsupervised classifier

Reports a feedback unsupervised classifier formulated by differential equations with no external control and few tuning parameters. This classifier is called the Lyapunov associative memory, in order to emphasize the importance of the Lyapunov (energy) function in the design of the associative memory. A vigilance parameter is built in to the dynamics of the classifier. Its architecture consists of two modules: the learning and recall modules. The learning module shapes the recall module, energy function with the arrival of new input information. The classifier was tested with an analog input pattern used by the ART-2 (adaptive resonance theory) model.< >