Reseach On The Interactive And Adaptive Evolutionary Algorithm

Optimization methods are playing an increasingly important role in many fields such as scientific research, social production, transportation and engineering technology at present. The typical modern optimization methods in solving optimization problems of display performance can solve arbitrary complexity problem, such as discontinuous, non-differentiable, non-convex, multimodal objective function, but they have some insurmountable difficulties in solving implicit performance optimization. Because of the capacity of solving implicit performance optimization problems, interactive evolutionary algorithm has gradually become the new hotspot of modern optimization method.As a modern optimization method of human-computer interaction, interactive evolutionary algorithm, having the unique properties of simulating the natural selection mechanism of "survival of the fittest" in the process of biological evolution and user participation, have been extensively and successfully applied in many fields of implicit performance optimization problems such as artistic designing, graph and image processing, industrial design, data mining, knowledge learning, musical composition etc. However, the practice shows that interactive evolutionary algorithm suffer from user fatigue problem in solving implicit performance optimization. Accelerating the algorithm convergence and integrating self-adaptive assessment mechanism in algorithm can directly reduce the assessment time of evolution individual that the user make, and then alleviate user fatigue. Based on the above purpose, this dissertation design and realize the interactive evolutionary algorithm (IGA) by learning from classical learning method in machine learning, including the naive Bayesian method (NB) and support vector machine (SVM), and then the adaptive interactive evolutionary algorithm is proposed for the design of commercial posters, and design and realize the interactive evolution platform. The main contents and innovations of this paper are summarized as follows:1. The application research of interactive evolutionary algorithms is explored, interactive evolutionary algorithm is applied to the design of commercial posters and the interactive evolutionary algorithm based on roulette wheel selection (RS-IGA) is proposed. Because the optimization scheme of commercial poster design cannot be clearly expressed by mathematical functions, typical modern optimization methods aren’t able to solve the problem,"human preferences" are needed to add into typical optimization methods to get the optimized design plan. Based on the above analysis, the commercial poster design problem is modeled and coded, and transform it into the implicit function optimization problem, which can be solved with the RS-IGA algorithm. Finally, the interactive evolutionary algorithm successfully solving this practical problem of commercial poster design is verified and obtains the ideal design scheme by analyzing the experimental results. 2. In order to solve the key problems of interactive evolutionary algorithm-the users’ fatigue problem, drawing lessons from naive Bayesian method in machine learning, the interactive evolutionary algorithm with adaptive mechanism is presented:Interactive evolutionary algorithm is based on Naive Bayesian method (NB-IGA). Firstly. The NB-IGA algorithm learns the experience of users, after learning some algebra, algorithm has the capability of evaluating new individual adaptive value. When the users begin to fatigue. NB-IGA displaces users to evaluate the evolution of the individual. In the Middle-Later Period of the evolution of algorithm, according to the requirement users can choose whether to go back to participate in algorithm, and then stop the algorithm running until it obtain the optimal solution. The naive Bayesian method directly reduces the number of individual that users evaluate, and can effectively relieve user fatigue.3. In order to solve the "user fatigue" problem of interactive evolutionary algorithm, the interactive evolutionary algorithm with adaptive mechanism is put forward by applying classification method of the support vector machine:Interactive evolutionary algorithm is based on support vector machine (SVM-IGA). In the initial stage of evolution, the user evaluates evolutionary individuals. When the user fatigues, evolutionary individual evaluated by user is used as the minimum sample set in SVM-IGA. After it is learnt, SVM-IGA displaces users to evaluate newly generated individual. In the Middle-Later Period of algorithm running, the users can participate in the algorithm at any time. Many experiments show that the SVM-IGA algorithm is an effectively self-adaptive assessment method to evaluate the individual adaptive value, which can directly reduce the time of individual that users evaluate, and can effectively ease user fatigue.4. Based on evolutionary algorithm design and analysis that integrates explicit and implicit performance optimization problem, an interactive evolutionary algorithm platform is designed and implemented. The platform, which has a friendly man-machine interface, provides several modern optimization methods, some representative Benchmark test functions and parameters of control window. The interactive platform not only can provide visual user interface for designing of interactive evolution algorithm of implicit performance optimization problem, but also can provide visual user interface and performance display regional for solving the explicit performance optimization problem, which accumulates many modern typical evolution algorithm, and guarantees the continuity of evolutionary computation research. The interactive platform provides a more universal support environment for comparative analysis, visual analysis, and the design of algorithm fusion etc.