Modified Bionic Strategy Of Bee Colony Algorithm

Bionic swarm intelligent optimization algorithm, a important branch of artificial intelligence, has become a research hotspot at present. Swarm intelligent optimization algorithm has some unique advantages, for example, its calculation process is relatively simple with flexible expandability, and data generation process doesn’t ask for high-performance CPU or large internal storage, besides it has a potential of parallel and distributed computing characteristics. Artificial bee colony algorithm is one of the newest probability search algorithm, carrying forward the theory of swarm intelligence optimization algorithm. Due to the less control parameters, easy application, simple calculation process and so on, bee colony algorithm has a wide application. However, there are still some limits, such as premature convergence, easy to fall into local optimum and low accuracy. Therefore, it is one of the key problems to choose appropriate stochastic process simulation and control the adjustment of the local search behavior.Blind source separation is one of the research focuses in signal processing field in recent years. When the source signal and the transmission channel are unknown, the source signals are recovered by separating the directly-observed mixed signals. Independent component analysis is a core algorithm with wide application, adopting the optimization algorithm based on gradient information. However, there are still some limits, such as slow convergence, easy to fall into local minima. Therefore, it is one of the key problems to choose appropriate optimization algorithm with simple structure, less parameters and flexible adjustment.According to above problems and research background, this thesis focuses on following aspects:1. At the beginning of the thesis, an overview is made on the blind source separation, bionic swarm intelligence optimization algorithm and bee colony algorithm, including system architecture, main algorithms and research progress as well as application both at home and abroad, etc.2. This thesis introduces the basic theory of blind source separation, including algorithm model, mathematical theory, mixed signal model, independent component analysis algorithm, signal preprocessing and function evaluation standard, etc. Detailed analysis is carried out on these theories.3. This thesis introduces bee colony algorithm in bionic swarm intelligence optimization algorithm, including the original biological model, basic principle and realization process. This thesis makes improvements to bionic strategy of bee colony algorithm, mainly including adopting the initial solution optimization strategy of reverse learning in the initialization phase, applying a improved search strategy with Levy flight characteristics in the phase of updating the population and local searching, and summarizing the process of improving algorithm.4. This thesis introduces the application of bee colony algorithm in blind source separation, improves bee colony algorithm based on biomimetic strategy and applies this algorithm as the calculation method for separation of matrix in blind source separation including its basic principle and theory, and summarizes the process of blind source separation after the application of modified artificial bee colony algorithm.5. This thesis introduces the contrast between the bee colony algorithm based on improved bionic strategy and the blind source separation method with application of bee colony algorithm. Experimental results show that modified bee colony algorithm based on reverse learning and Levy flight improves the convergence speed and ensures the accuracy of convergence. Besides when the modified bee colony algorithm is used as the optimization algorithm of blind source separation in the initial matrix, problems such as the contradiction of separation effect and convergence speed can be solved to a certain extent.6. At the end of this thesis, a summary is made on the improvements and the limits of our research work. The prospective research areas on bee colony algorithm are also proposed.