Research On Quantum-behaved Particle Swarm Algorithm And Data Classification

The mutual penetration, cross and promotion with each other between information science and life science is a notable feature in the development of the modern science and technology. Intelligent computation is a new interdisciplinary subject formed by the combination of information science and life science. As a representative algorithm of swarm intelligence, particle swarm algorithm (PSO) is simple in concept, convenient to implement, and have the characteristics of fast convergence speed, less parameter Settings etc., which make the algorithm efficient and get widespread attention in academic circles. In recent years, with the development of society, the actual problems becoming more and more complex, the precocious faults of particle swarm algorithm constantly expose. As one of the most thrilling found of physics in the20th century, quantum compute, produced by the fusion of quantum mechanics and informatics, has the properties of superposition of the quantum state, entanglement and interference, which is different from traditional calculation in nature. In this paper, against the frame of particle swarm algorithm and the quantum theory, a novel optimization method named Multiple Collapse Orthogonal Crossover Quantum Particle Swarm algorithm is put forward. To solve classification problems in optimization way, quantum-behaved particle swarm optimization (QPSO) algorithm and a kind of nearest prototype method are tried to combine together, and the novel QPSO is also brought in to solve data classification. The main contributions can be listed as follows:1) A new quantum particle swarm algorithm on the basis of the cross many collapse is put forward. First, making full of the uncertainty of the quantum mechanism, the method collapse in many times from the quantum state to classic state, which enhances the diversity of population. Then we to get the individuals orthogonal cross communicate with each other, which make full use of the effective information carried by every individual and eventually search out the optimal solution. In order to prove the performance of the algorithm, some typical benchmark test function optimization are tested and experimental results show that this method can accelerate the convergence speed and converge more easily to the global extreme value point, moreover, when applied in difficult CEC05composite function, it also can search to the optimal solutions rapidly.2) Based on the compute complexity feature of the nearest neighborhood classification, cooperating with the uncertainty characteristic of quantum system, we combine QPSO and nearest neighborhood prototype to solve data classification problem. The code of a particle in this method possessing multiple prototypes and each prototype corresponds to a class mark. After choose effective prototype with QPSO, test data just need to calculate the distance to the prototypes, the number of which is much smaller compared with a lot of training data, hence reducing the computational complexity in great degree. Comparative experiments show that this algorithm is of much improvement in speed and classification results.3) An improved algorithm is proposed. Based on the successful experience of data classification using quantum-inspired nearest prototype algorithm, we bring in the improved quantum-behaved particle swarm optimization algorithm further in combination with nearest neighborhood prototype to solve data classification problem. The experiments showed that the improved one can get better result than the original one and other traditional classification approaches.This research is supported by the National High Technology Research and Development Program (863Program) of China (Grant No.2009AA12Z210), the Key Scientific and Technological Innovation Special Projects of Shaanxi "13115",(No.2008ZDKG-37), the National Natural Science Foundation of China(Grant No.60703107and60703108),the Natural Science Basic Research Plan in Shaanxi Province of China(Grant No.2007F32), the China Postdoctoral Science Foundation Special funded project (No.200801426), the China Postdoctoral Science Foundation funded project (No.20080431228) and the Fundamental Research Funds for the Central Universities (No.JY10000902040).