The Application Research On Network Boosting Learning

Network Boosting Algorithm combined the classifiers network and Boosting learning to construct a classifiers network system with stronger learning ability and generalization ability. On this basis, The Network Boosting algorithm is applied to SAR target recognition and intrusion detection. According to the existing problem, we have researched them and proposed the solvent. The main contributions can be summarized as follows:(1) The self-tuning Network Boosting Learning based on ICSA (ICSNB) is proposed. A weight-tuning method (WNB) of Network Boosting Algorithm is proposed through introducing the weight factors, which are used to modify the decision function so that the final hypothesis is the weighted mean of the learned hypotheses. WNB algorithm has the better stability and the stronger anti-noise ability. In order to avoid luxury searching artificially a combination of multi-parameter in Network Boosting algorithm, which needs expensive computing cost, we make use of the global optimal searching ability of Immune Clonal Selection Algorithm (ICSA) to optimize multi-parameter to coordinate with each other. ICSNB can automatically adjust parameters which affect the structure and performance of classifiers network while can speed up the selection of multi-parameter. The experimental result show that ICSNB can save much time while maintaining higher classification accuracy compared with the method of parameters selected manually. And the present algorithm can attain the weightiness order of parameters.(2) The target recognition accuracy of SAR images is not satisfied. The labels of images acquisition and recollecting are difficult and expensive. In order to solve the problem, we construct the transfer model of the base classifier and introduce transfer learning into Network Boosting algorithm (NB) to propose Transfer Network Learning algorithm (TNL), in which other out-date data are reused to instruct the SAR target recognition. TNL is suitable to improve the performance of SAR target recognition, in which instances transfer learning is adopted for domain adaptation. The experimental results show that the proposed algorithm has better performance and achieves different domains learning.(3) In order to solve the problem that there exists unbalanced detection performance on different types of attacks in current large-scale network intrusion detection algorithms, Distributed Transfer Network Learning algorithm (DTNL) is proposed. The algorithm introduces transfer learning into Distributed Network Boosting algorithm for instructing the attacks learning with poor performance, in which the instances transfer learning is adopted for domain adaptation. The experimental results on the KDD CUP'99 Data Set show that the proposed algorithm has higher efficiency and better performance. Further, the detection accuracy of R2L attacks has been improved greatly while maintaining higher detection accuracy of other attack types.This paper was supported by the Eleventh Five-Year Pre-research Project (No. 51307040103), the National High Technology Research and Development Program (973 Program) of China (No.2006CB705707),, the Cultivation Fund of the Key Scientific and Technical Innovation Project, Ministry of Education of China (No.706053), the National Natural Science Foundation of China (Nos.60672126, 60703109,60803098), the Provincial Natural Science Foundation of Shaanxi of China (No.2009JQ8016).