An evolutionary hyper-heuristic to optimise deep belief networks for image reconstruction

Deep Belief Networks (DBN) have become a powerful tools to deal with a wide range of applications. On complex tasks like image reconstruction, DBN’s performance is highly sensitive to parameter settings. Manually trying out different parameters is tedious and time consuming however often required in practice as there are not many better options. This work proposes an evolutionary hyper-heuristic framework for automatic parameter optimisation of DBN. The hyper-heuristic framework introduced here is the first of its kind in this domain. It involves a high level strategy and a pool of evolutionary operators such as crossover and mutation to generates DBN parameter settings by perturbing or modifying the current setting of a DBN. Providing a large set of operators could be beneficial to form a more effective high level strategy, but in the same time would increase the search space hence make it more difficulty to form a good strategy. To address this issue, a non-parametric statistical test is introduced to identify a subset of effective operators for different phases of the hyper-heuristic search. Three well-known image reconstruction datasets were used to evaluate the performance of the proposed framework. The results reveal that the proposed hyper-heuristic framework is very competitive when compared to the state of art methods.     

Rough set and scatter search metaheuristic based feature selection for credit scoring

As the credit industry has been growing rapidly, credit scoring models have been widely used by the financial industry during this time to improve cash flow and credit collections. However, a large amount of redundant information and features are involved in the credit dataset, which leads to lower accuracy and higher complexity of the credit scoring model. So, effective feature selection methods are necessary for credit dataset with huge number of features. In this paper, a novel approach, called RSFS, to feature selection based on rough set and scatter search is proposed. In RSFS, conditional entropy is regarded as the heuristic to search the optimal solutions. Two credit datasets in UCI database are selected to demonstrate the competitive performance of RSFS consisted in three credit models including neural network model, J48 decision tree and Logistic regression. The experimental result shows that RSFS has a superior performance in saving the computational costs and improving classification accuracy compared with the base classification methods.     

A credibility-based fuzzy location model with Hurwicz criteria for the design of distribution systems in B2C e-commerce

Facility location problem is one of the most critical elements in the design of distribution systems, and numerous studies have focused on this issue. However, facility location theory and guidelines for B2C firms are sparse. In this paper, with regard to the customer characteristics peculiar to B2C e-commerce and the turbulence of the competitive market, a new fuzzy location model is proposed to optimize the distribution system design in B2C e-commerce. The model adopts a hierarchical agglomerative clustering method to classify customers and estimate the fuzzy delivery cost. At the same time, due to the turbulence of competitive market, both market supply and customer demand are treated as fuzzy variables in the model. Afterward, the credibility measure and Hurwicz criterion are introduced to convert the model into a crisp one which has NP-hard complexity. In order to solve the crisp model, an improved genetic algorithm with particle swarm optimization is developed. Finally, the computational results of some numerical examples are used to illustrate the application and performance of the proposed model and algorithm.     

A variable neighborhood search based approach for uncapacitated multilevel lot-sizing problems

In this paper, an effective approach based on the variable neighborhood search (VNS) algorithm is presented to solve the uncapacitated multilevel lot-sizing (MLLS) problems with component commonality and multiple end-items. A neighborhood structure for the MLLS problem is defined, and two kinds of solution move policies, i.e., move at first improvement (MAFI) and move at best improvement (MABI), are used in the algorithm. A new rule called Setup shifting is developed to conduct a more efficient neighborhood search for the MLLS problems. Computational studies are carried out on two sets of benchmark problems. The experimental results show that the VNS algorithm is capable of solving MLLS problems with good optimality and high computational efficiency as well, outperforming most of the existing algorithms in comparison.     

A new stochastic approach for a reliable p-hub covering location problem

A hub location problem (HLP) is a fertile research field, in the aspect of interdisciplinary studies, such as transportation, operation research, network design, telecommunication and economics. The location of hub facilities and allocation of non-hub nodes to hubs configure the backbone of HLPs. This study presents a new mathematical model for a reliable HLP by a new stochastic approach to minimize the total transportation cost and obtain maximum flows that the network can carry, when its link capacities are subject to stochastic degradations, as in a form of daily traffic, earthquake, flood, etc. We consider the road capacity reliability as a probability that ensures the maximum network capacity is greater than or equal to the total incoming flow to the network by considering the road capacity as random variable. As a result, this paper assumes that link capacities satisfy in a Truncated Erlang $(\mathit{TErl})$ distribution function. Due to complexity of the HLP, a meta-heuristic algorithm, namely differential evolution (DE) algorithm, is applied on the problem in order to achieve near-optimal solutions. Furthermore, the performance of the proposed algorithm (i.e., DE) is evaluated by the performance of the genetic algorithm (GA) applied on the given problem. Some computational experiments are presented to illustrate the effectiveness of the presented model and proposed algorithm. Finally the conclusion is provided.     

Optimizing 3D Irregular Object Packing from 3D Scans Using Metaheuristics

For efficient construction-assemblies transportation, volume constrained 3D printing, dry stacking, and facility waste management, a common problem must be solved. It is the practical problem of packing irregular 3D rigid objects into a container with fixed dimensions so that the volume of the final packed objects is minimized. To solve this problem, a methodology is presented that begins with capturing the initial as-is 3D shape data for each object, followed by a metaheuristic-based packing optimization algorithm. This methodology is demonstrated to be effective in two situations with known optimum solutions and in a third situation involving packing of real-life as-is objects. A high-level selection algorithm that is designed to guide the search of possible object subsets, when not all objects can fit into a single predefined container, is discussed as well. Performance is examined for variations, and a preliminary sensitivity analysis is performed. The methodology and its key algorithms are demonstrated to produce effective packing solutions in a mostly automatic manner. Object packing for this class of applications in civil engineering can thus be potentially improved in terms of outcome efficiency and level of planning effort required.     

Lot sizing and furnace scheduling in small foundries

A lot sizing and scheduling problem prevalent in small market-driven foundries is studied. There are two related decision levels: (1) the furnace scheduling of metal alloy production, and (2) moulding machine planning which specifies the type and size of production lots. A mixed integer programming (MIP) formulation of the problem is proposed, but is impractical to solve in reasonable computing time for non-small instances. As a result, a faster relax-and-fix (RF) approach is developed that can also be used on a rolling horizon basis where only immediate-term schedules are implemented. As well as a MIP method to solve the basic RF approach, three variants of a local search method are also developed and tested using instances based on the literature. Finally, foundry-based tests with a real-order book resulted in a very substantial reduction of delivery delays and finished inventory, better use of capacity, and much faster schedule definition compared to the foundry's own practice.     

Meta-heuristics for stable scheduling on a single machine

This paper presents a model for single-machine scheduling with stability objective and a common deadline. Job durations are uncertain, and our goal is to ensure that there is little deviation between planned and actual job starting times. We propose two meta-heuristics for solving an approximate formulation of the model that assumes that exactly one job is disrupted during schedule execution, and we also present a meta-heuristic for the global problem with independent job durations.     

随机蛙跳算法的研究进展

随机蛙跳算法(Shuffled Frog Leaping Lgorithm,SFLA)是进化计算领域中一种新兴、有效的亚启发式群体计算技术,近几年来逐渐受到学术界和工程优化领域的关注.SFLA结合了具有较强局部搜索(Local Search,LS)能力的元算法(Memetic Algorithm,MA)和具有良好全局搜索(Global Search,GS)性能的粒子群算法(Particle Swarm Optimization,PSO)的特点,因此其寻优能力强,易于编程实现.详细阐述了SFLA的基本原理和流程,总结了SFLA目前在优化和工程技术等领域中的研究,展望了SFLA的发展前景.     

Simulated annealing and ant colony optimization algorithms for the dynamic throughput maximization problem

In many industries, inspection data is determined to merely serve for verification and validation purposes. It is rarely used to directly enhance the product quality because of the lack of approaches and difficulties of doing so. Given that a batch of subassembly items have been inspected, it is sometimes more profitable to exploit the data of the measured features of the subassemblies in order to further reduce the variation in the final assemblies so the rolled yield throughput is maximized. This can be achieved by selectively and dynamically assembling the subassemblies so we can maximize the throughput of the final assemblies. In this paper, we introduce and solve the dynamic throughput maximization (DTM) problem. The problem is found to have grown substantially by increasing the size of the assembly (number of subassembly groups and number of items in each group). Therefore, we resort to five algorithms: simple greedy sorting algorithm, two simulated annealing (SA) algorithms and two ant colony optimization (ACO) algorithms. Numerical examples have been solved to compare the performances of the proposed algorithms. We found that our ACO algorithms generally outperform the other algorithms.