Ill-posed problems are widely existed in signal processing. In this paper, we review popular regularization models such as truncated singular value decomposition regularization, iterative regularization, variational regularization. Meanwhile, we also retrospect popular optimization approaches and regularization parameter choice methods. In fact, the regularization problem is inherently a multiobjective problem. The traditional methods usually combine the fidelity term and the regularization term into a singleobjective with regularization parameters, which are difficult to tune. Therefore, we propose a multi-objective framework for ill-posed problems, which can handle complex features of problem such as non-convexity, discontinuity. In this framework, the fidelity term and regularization term are optimized simultaneously to gain more insights into the ill-posed problems. A case study on signal recovery shows the effectiveness of the multi-objective framework for ill-posed problems.
In this paper, we propose a new hybrid method called SQPBSA which combines backtracking search optimization algorithm (BSA) and sequential quadratic programming (SQP). BSA, as an exploration search engine, gives a good direction to the global optimal region, while SQP is used as a local search technique to exploit the optimal solution. The experiments are carried on two suits of 28 functions proposed in the CEC-2013 competitions to verify the performance of SQPBSA. The results indicate the proposed method is effective and competitive.
Particle swarm optimization (PSO) is one of the most popular population-based stochastic algorithms for solving complex optimization problems. While PSO is simple and effective, it is originally defined in continuous space. In order to take advantage of PSO to solve combinatorial optimization problems in discrete space, the set-based PSO (SPSO) framework extends PSO for discrete optimization by redefining the operations in PSO utilizing the set operations. Since its proposal, S-PSO has attracted increasing research attention and has become a promising approach for discrete optimization problems. In this paper, we intend to provide a comprehensive survey on the concepts, development and applications of S-PSO. First, the classification of discrete PSO algorithms is presented. Then the S-PSO framework is given. In particular, we will give an insight into the solution construction strategies, constraint handling strategies, and alternative reinforcement strategies in S-PSO together with its different variants. Furthermore, the extensions and applications of S-PSO are also discussed systemically. Some potential directions for the research of S-PSO are also discussed in this paper.
Particle swarm optimizer (PSO) is an effective tool for solving many optimization problems. However, it may easily get trapped into local optimumwhen solving complex multimodal nonseparable problems. This paper presents a novel algorithm called distributed learning particle swarm optimizer (DLPSO) to solve multimodal nonseparable problems. The strategy for DLPSO is to extract good vector information from local vectors which are distributed around the search space and then to form a new vector which can jump out of local optima and will be optimized further. Experimental studies on a set of test functions show that DLPSO exhibits better performance in solving optimization problems with few interactions between variables than several other peer algorithms.
The multiple knapsack problem (MKP) forms a base for resolving many real-life problems. This has also been considered with multiple objectives in genetic algorithms (GAs) for proving its efficiency. GAs use selfadaptability to effectively solve complex problems with constraints, but in certain cases, self-adaptability fails by converging toward an infeasible region. This pitfall can be resolved by using different existing repairing techniques; however, this cannot assure convergence toward attaining the optimal solution. To overcome this issue, gene position-based suppression (GPS) has been modeled and embedded as a new phase in a classical GA. This phase works on the genes of a newly generated individual after the recombination phase to retain the solution vector within its feasible region and to improve the solution vector to attain the optimal solution. Genes holding the highest expressibility are reserved into a subset, as the best genes identified from the current individuals by replacing the weaker genes from the subset. This subset is used by the next generated individual to improve the solution vector and to retain the best genes of the individuals. Each gene’s positional point and its genotype exposure for each region in an environment are used to fit the best unique genes. Further, suppression of expression in conflicting gene’s relies on the requirement toward the level of exposure in the environment or in eliminating the duplicate genes from the environment. The MKP benchmark instances from the OR-library are taken for the experiment to test the new model. The outcome portrays that GPS in a classical GA is superior in most of the cases compared to the other existing repairing techniques.