Anyuan Deng

Novel Binary Differential Evolution Algorithm for Discrete Optimization

New Binary Differential Evolution algorithm was proposed for the combinatorial optimization problem. With the same framework of the original Differential Evolution algorithm, three new operators were used to expand the continuous field of the original Differential Evolution to the discrete field. Firstly, a new operator, mapping operator, in the new algorithm was used to ensure the original mutation operator still effective. Then a new S operator, with sigmoid function, was used to keep the result of the mutation operator falls in the interval. Before the crossover operator, an inverse mapping operator transformed the continuous numbers to discrete. Two initial simulation results show it is effective and useful.

Novel binary Differential Evolution without scale factor F

Differential Evolution is a competitive optimization technique over continuous space. The operation in the original Differential Evolution is simple, however, the mechanism, in the Differential Evolution makes it practically impossible to effectively use the original Differential Evolution to the binary space. A novel binary mutation operation was defined to enable the Differential Evolution to operate within the binary space. The new binary mutation works well in the binary space without the original scale factor F. Initial experimental results of three different sizes of knapsack problems and the One-Max problem indicate the effectiveness and validity of the binary Differential Evolution operating in binary space.

Hybrid-Coding Binary Differential Evolution Algorithm with Application to 0-1 Knapsack Problems

Binary Differential Evolution algorithm was proposed for discrete optimization problem. Firstly, a new operator, boundary-handling operator, was added to the original Differential Evolution to ensure each population generated by the mutation and crossover operator comply with the boundary constraints. Then a new hybrid coding Differential Evolution algorithm with mapping method was put forward to deal with the discrete optimization problem. And lastly, a new selection operator was employed to deal with constraints directly. Two initial simulation results of knapsack problem with different variables show it is effective and useful. Hybrid coding Differential Evolution algorithm is a new effective way for solving the discrete optimization problem.

New penalty function with differential evolution for constrained optimization

The penalty function is one of the most commonly used approaches for constrained optimization problems. However, it often leads to additional parameters and the parameters are not easy for the users to select. A new way without additional parameters to deal the constrained optimizations was proposed. Firstly, a new penalty function was defined using the constrained functions without additional parameters. Secondly, combining the penalty function and the original objective function, a new objective function without any constrained conditions was got. Then differential evolution algorithm was used to solve the non-constrained optimization problem. The numerical experiments show its advantage over the other existing method.

Hybrid differential evolution for knapsack problem

A hybrid Differential Evolution algorithm with double population was proposed for 0-1 knapsack problem The two populations play different roles during the process of evolution with the floating-point population as an engine while the binary population guiding the search direction Each gene of every chromosome in the floating-point encoding population is restricted to the range [-1, 1], while each gene of every chromosome in the binary encoding population is zero or one A new mapping operation based on sign function was proposed to generate the binary population from the floating-point population And a local refining operation called discarding operation was employed in the new algorithm to fix up the solutions which are infeasible Three benchmarks of 0-1 knapsack problem with different sizes were used to verify the new algorithm and the performance of the new algorithm was also compared with that of other evolutionary algorithms The simulation results show it is an effective and efficient way for the 0-1 Knapsack problem.

Integer Encoding Differential Evolution Algorithm for Integer Programming

A novel integer encoding Differential Evolution (IEDE) algorithm was proposed for integer optimization problems in this paper. Based on the standard framework of the traditional DE, the population was encoding with integer. The IEDE inherited the crossover operator and selection operator from the traditional DE directly. And a new integer mutation operator was defined to deal with the integer encoding individual. Several initial simulation results show it is effective and efficient in solving the integer optimization problems. The IEDE is a new effective way for the integer optimization problems.

Modified Differential Evolution for Hard Constrained Optimization

Hard constrained optimization problems in science and engineering are common computationally very expensive. This leads to serious impediment to the successful application of evolutionary optimization techniques. A modified Differential Evolution with hybrid mutation and new selection rules was proposed to solve the hard constrained optimization problem. The hybrid mutation is the linear combination of two different traditional mutations schemes using simulated annealing way. The new selection rule can handle the constraints directly without choosing the penalty factors difficultly. The simulation results on the Bump function show it is effective and robust way for solving the hard constrained optimization.

Modified Dynamic Differential Evolution for 0-1 Knapsack Problems

A modified dynamic differential evolution was proposed for discrete optimization. Based on the new framework of dynamic differential evolution, two additional operators were used to extend the dynamic differential evolution to the field of discrete optimization. The first operator was the mapping operator, which could map the continuous value into zero or one. The other new operator was the boundary constraints handling operator, which ensured the results gotten by the mutation operation fall in some range. The results of the simulation on four different sizes of the knapsack problems show it is efficient and effective way for solving 0-1 knapsack problems.

Notice of Retraction Mixed-coding Differential Evolution Algorithm for Integer Programming

Mixed-coding population Differential Evolution algorithm was proposed for integer programming. Firstly, each individual in the initial population with two different fields, one is real and the other is integer, is generated. Then a new operator, boundary-handling operator, was added to the original Differential Evolution to ensure each population generated by the mutation operator and crossover operator satisfying the boundary constraints. Lastly a new mapping operator was put forward to map one real number into an integer number, which is stored in the integer field. Two initial simulation results of two integer programming with the size of 30 variables show it is effective and efficient. Mixed-coding Differential Evolution algorithm is a new effective way for solving the integer programming problems.

Differential Evolution with dual population for static Weapon-Target assignment problem

A new Differential Evolution algorithm with dual population was proposed for the static Weapon-Target problem. There are two different types of population in the new algorithm with one is floating-point coded and the other is sequence number coded. During the evolution a new mapping operation was proposed to generate the corresponding sequence population from the floating-point population. And the sequence population was used to guide the direction of the new algorithm by deciding which one between the floating-point individual and its rival will exist in the next generation. Initial simulation results the new algorithm is effective and efficient in solving the static Weapon-Target problem which is NP-complete.