By controlling a bifurcation parameter from the chaotic dynamics regime to the fixed-point regime gradually the system may eventually reach the global optimum state or its good approximation with very high probability.
Through compare and analysis, we get a satisfied improved hybrid Genetic Algorithm with (μ + λ) select, edge probability recombine and 2-opt local search. It can quickly converge to the global optimum solution.
Genetic Algorithm(for short, GA or GAs), which has a simple, all-purpose, sure character, has made great achievements in the solution to difficult, complex problems, and can convergence to the global minimum. However, random search techniques handicap convergence steep of Genetic Algorithm which likely to seek out a local optimum solution but a global sub-optimization.
In Part three, based on a kind of new simulated annealing, it is proved that, in theory, such algorithms asymptotically converge to the global minimum point with probability one under suitable conditions.
In this paper, a new hybrid algorithm which combines the chaos optimization method and the conjugate gradient approach having an effective convergence property, is proposed. The hybrid algorithm can help the conjugate gradient approach to skip the local minimum. At the end, it can find the global minimum.
Experimental results have shown that a global optimal solution can be quickly obtained using the proposed method and the precision requirement for target location is satisfied.
In addition, considering the non-convex and non-concave nature of the sub-problem of combinational optimization, the branch-and-bound technique was adopted to obtain or approximate a global optimal solution.
To speed up the search process and guarantee a global optimal result, the extended compact genetic algorithm (ECGA) is used to carry out the search process.
These schemes are based on a unified theoretical base-sufficient conditions for the global optimal known in optimal control theory.
2D and 3D ASMs are combined to obtain a "global optimal" segmentation of the 3D object embedded in the data set, rather than the "locally optimal" segmentation on separate slices.