Multimedia data stream information mining algorithm based on jointed neural network and soft clustering

As one non-surveillance study method, soft clustering is well applied in the data mining, the imagery processing, the pattern recognition, the spatial remote sensing technology and the characteristic extraction and so on state-of-the-art applications in many domains all have the widespread application. Inspired by the combination of neural network and soft computing model, in this paper, we propose the novel multimedia data stream information mining model based on jointed neural network and soft clustering. In the self-training process, we train a learner on a tagged sample set and then use the learner to mark an unlabeled sample that it considers to be highly reliable and add the newly labeled sample to the original training set, then use this new training set to re-train the learner and repeat the above process until the iteration condition is terminated. To better play to the performance of the main processor and then save address space resources with the wishbone protocol implementation the separation of the high-low speed hierarchical interconnection structure GPU is applied. Experimental result proves the effectiveness of the proposed model.