Continuous Optimization Algorithm Based On The Ib Theory

IB method, as a data analysis method based on information theory, effectively solves the trade-off between the accuracy and the efficiency existed in complex problems by balancing information compression and information preservation, further produces several IB algorithms, of which the sequential optimization algorithm(sIB) is a better one, but this algorithm has some problems in efficiency, accuracy and so forth.Aimed to the problems show above, this thesis proposes isIB algorithm by introducing mutation method. Based on the initial solution space formed by many times of Depth-First Search(DFS) from random sIB, this algorithm enlarges effectual search space by local Breadth-First Search(BFS), finally attains better solutions. Compared with sIB algorithm on nine document datasets from 20NG corpus, this thesis verifies the validity of isIB algorithm: special experiments first attain an effective mutation rate, and then illustrate that isIB algorithm outperforms the sIB algorithm in both the accuracy and the efficiency. At the same time, we also get the best repeating times for attaining better solutions; Along with the increase of repeating times, the addition of algorithm's accuracy gradually minishes, this indicates that isIB algorithm is convergent. The algorithm also shows Monte Carlo method can play an important role for solving complex problems.The high-effective algorithms for solving complex optimization problems can improve efficiency for dealing with problems. Ths isIB algorithm, proposed in this thesis, can be used for information retrieval and searching useful information by key words, etc. Additionally, IB method successfully applied in many fields, the novel isIB algorithm can also be used for image processing and other fields. The applications imply the extensive realistic meaning of algorithm research.