Flute: fuzzy learning in unfamiliar teacher environments

Two conceptually elegant ideas of learning with an unfamiliar teacher and fuzzy models are synergistically combined to derive a new fuzzy pattern recognition methodology for operating in imperfectly supervised environments. Under this new methodology, which is applicable to, multiple pattern class problems also, the fuzziness introduced into the recognition problem by the imperfectness of the supervision in the environment is modeled with fuzzy membership functions. These functions are learnt during the training phase by employing the unfamiliar teacher concepts, and deployed during the ensuing classification phase to take into account the imperfectness of the learning environment.<>