查看更多>>摘要：Evolutionary algorithms (EAs) show good performance in solving multi-objective optimization problems (MOPs). An EA needs to perform a substantial number of fitness evaluations. For the MOP with high complexity, the fit-ness evaluation functions are computationally expensive, making the evolutionary algorithms time-consuming. Surrogate-assisted evolutionary algorithms (SAEAs) that apply surrogate models instead of fitness exact evalua-tion functions have successfully reduced the computational complexity of fitness evaluations. However, because training a surrogate model requires a certain amount of calculation, a large amount of calculation is required by the SAEA to train multiple surrogate models. Furthermore, most existing surrogate models may not achieve desired evaluation accuracy when processing medium-dimensional and high-dimensional MOPs. This paper pro -poses a novel surrogate model. The surrogate model can be applied in multi-objective optimization evolutionary algorithm based on decomposition (MOEA/D), which is a classic decomposition-based multi-objective optimiza-tion algorithm. The surrogate model is designed based on the convolutional neural network structure, and it is called the multi-objective parallel fitness evaluation network (MPFEN). An MPFEN model contains multiple sub-networks which can be applied as the surrogate models. By training the MPFEN model, we can obtain all sur-rogate models required by a MOEA/D simultaneously without training each required surrogate model separately. Therefore, the amount of calculation of training surrogate models in a MOEA/D is reduced. The evaluation accu-racy of the MPFEN model is tested by experiments. The experimental results show that the evaluation accuracy of MPFEN model is higher than that of other classical surrogate models in most cases. By applying the MPFEN model, the solution quality of SAEA is also improved.
查看更多>>摘要：The solving of large-scale multi-objective optimization problem (LSMOP) has become a hot research topic in evolutionary computation. To better solve this problem, this paper proposes a self-organizing weighted optimization based framework, denoted S-WOF, for addressing LSMOPs. Compared to the original framework, there are two main improvements in our work. Firstly, S-WOF simplifies the evolutionary stage into one stage, in which the evaluating numbers of weighted based optimization and normal optimization approaches are adaptively adjusted based on the current evolutionary state. Specifically, regarding the evaluating number for weighted based optimization (i.e., t(1) ), it is larger when the population is in the exploitation state, which aims to accelerate the convergence speed, while t(1) is diminishing when the population is switching to the exploration state, in which more attentions are put on the diversity maintenance. On the other hand, regarding the evaluating number for original optimization (i.e., t(2) ), which shows an opposite trend to t(1) , it is small during the exploitation stage but gradually increases later. In this way, a dynamic trade-off between convergence and diversity is achieved in SWOF. Secondly, to further improve the search ability in the large-scale decision space, an efficient competitive swarm optimizer (CSO) is implemented in S-WOF, which shows efficiency for solving LSMOPs. Finally, the experimental results have validated the superiority of S-WOF over several state-of-the-art large-scale evolutionary algorithms.
查看更多>>摘要：Collective behaviours such as swarm formation of autonomous agents offer the advantages of efficient movement, redundancy, and potential for human guidance of a single swarm organism. However, tuning the behaviour of a group of agents so that they swarm, is difficult. Behaviour-bootstrapping algorithms permit agents to self-tune behaviour adapted for their physical form and associated movement constraints. This paper proposes a reinforce-ment learning framework to tune collective motion behaviours from random behaviours. The learning process is guided by a novel reward function capable of autonomously detecting generic collective motion behaviours from sensor data about the relative velocity and position of neighbouring agents. Our reward function is de-signed using a meta-learner trained on a human-labelled collective motion dataset. We demonstrate that our reinforcement learner can tune the behaviour of randomly moving groups so that structured collective motion emerges. We compare our framework to an existing developmental evolutionary framework for this purpose. Our results demonstrate that the proposed learning framework can generate behaviours with different collective motion characteristics more quickly than existing approaches. In addition, the trained reinforcement learner can tune the behaviour of robots with movement characteristics that it has not been trained on.
查看更多>>摘要：A critical problem in the Orthogonal Frequency Division Multiplexing (OFDM) framework for 5G communication is the high Peak to Average Power Ratio (PAPR) of the transmitted signal. The Tone Reservation (TR) scheme is employed to minimize the OFDM signal's PAPR in time domain. It is comprised of some reserved subcarriers to generate a Peak Reduction Tone (PRT) set. The selection of these PRT set and the optimal target clipping level are the depending parameters of the PAPR reduction performance. A novel Differential Evolution based Tone Reservation (DE-TR) scheme is proposed to find an optimal PRT set and then the PAPR is reduced by employing an Optimized Iterative Clipping and Filtering (OICF) technique. Compared to other related techniques like Clipping Control Tone Reservation (CC-TR), Curve Fitting based Tone Reservation (CF-TR) and Parallel Tabu Search Tone Reservation (PTS-TR), the proposed technique DE-TR(Pro.) contributes to a substantial PAPR reduction output with lower computational complexity and maintaining an adequate bit error rate (BER) and spectral spreading efficiency. In addition, the theoretical relationship between the Error Vector Magnitude (EVM) and PAPR as well as CCDF performance with EVM are analyzed.
查看更多>>摘要：There is evidence that many real-world applications can be characterized as sparse multiobjective problems (SMOPs), where most variables of their Pareto optimal solutions are zero. Existing multiobjective evolution-ary algorithms (MOEAs) have shown their competitiveness on conventional SMOPs. However, they may en-counter difficulties when tackling large-scale SMOPs (LSMOPs). This paper thereby proposes a two-stage MOEA tailored to LSMOPs, named TS-SparseEA. TS-SparseEA integrates the prior information into the evolution and enables the population to spread over the Pareto front by two stages. In the first stage, TS-SparseEA adopts a new binary weight optimization framework, transforming the original large-scale optimization problem into a low-dimensional one via a set of low-dimensional binary weights. In the second stage, TS-SparseEA employs an improved evolutionary algorithm, including a hybrid encoding and a specialized matching strategy, where each solution is reproduced by a conditional combination between two types of variables. To summarize, the proposed binary weight optimization can better address large-scale sparse variables by generating a high-quality initial population, whereas the new hybrid encoding can facilitate the offspring evolution. Extensive experiments have verified the effectiveness of TS-SparseEA on LSMOPs, by comparing it with several state-of-the-art MOEAs on both benchmark problems and real-world applications.
查看更多>>摘要：Fog computing is a new computational paradigm that emerged from the need to reduce network usage and la-tency in the Internet of Things (IoT). Fog can be considered as a continuum between the cloud layer and IoT users that allows the execution of applications or storage/processing of data in network infrastructure devices. The het-erogeneity and wider distribution of fog devices are the key differences between cloud and fog infrastructure. Genetic-based optimization is commonly used in distributed systems; however, the differentiating features of fog computing require new designs, studies, and experimentation. The growing research in the field of genetic-based fog resource optimization and the lack of previous analysis in this field have encouraged us to present a compre-hensive, exhaustive, and systematic review of the most recent research works. Resource optimization techniques in fog were examined and analyzed, with special emphasis on genetic-based solutions and their characteristics and design alternatives. We defined a taxonomy of the optimization scope in fog infrastructures and used this optimization taxonomy to classify the 70 papers in this survey. Subsequently, the papers were assessed in terms of genetic optimization design. Finally, the benefits and limitations of each surveyed work are outlined in this paper. Based on these previous analyses of the relevant literature, future research directions were identified. We concluded that more research efforts are needed to address the current challenges in data management, work-flow scheduling, and service placement. Additionally, there is still room for improved designs and deployments of parallel and hybrid genetic algorithms that leverage, and adapt to, the heterogeneity and distributed features of fog domains.
查看更多>>摘要：Research into the applicability of ant-based optimisation techniques for hyper-heuristics is largely limited. This paper expands upon the existing body of research by presenting a novel ant-based generation constructive hyper heuristic and then investigates how different pheromone maps affect its performance. Previous work has focused on applying ant-based optimisation techniques that work in the solution space directly to the heuristic space and we hypothesise that this may be problematic for the hyper-heuristic's efficacy. The focus of this analysis is primarily on how the pheromone map, 2D and 3D, of ant-based methods, can be used for this hyper-heuristic task. 2D pheromone maps are the predominant pheromone map type used by ant-based algorithms. Thus the comparison here is between the existing 2D pheromone map and the newly introduced 3D pheromone map. The analysis consists of multiple experiments with algorithms in the TSP and 1DBPP domain which are assessed in terms of optimality and generality. The results of the experiment demonstrate key differences in performance between the two different pheromone spaces. The 3D pheromone map showed better generality and optimality in the 1DBPP domain whereas the 2D pheromone map showed better generality and only marginally better optimality for the TSP domain. The analysis indicated that the different pheromone maps work most optimally for different types of optimisation problems. The hybrid method showed some improvements in generality but showed little improvements in optimality overall.
查看更多>>摘要：In this paper, a surrogate-assisted hybrid swarm optimization algorithm is proposed to solve high-dimensional computationally expensive problems. Two swarms are, respectively, used in different optimization states. The first swarm uses the teaching-learning-based optimization in the early stage to enhance the exploration. The second swarm uses the particle swarm optimization in the later stage to accelerate convergence. Two different pre-screening criteria based on the corresponding evolutionary rules are proposed to select promising individuals for exact function evaluations. Several commonly used benchmark functions with their dimensions varying from 30 to 200 and an engineering optimization problem are used to validate the efficiency of the proposed algorithm. In addition, a comprehensive analysis is conducted to demonstrate the effectiveness of each main component of the proposed algorithm. The experimental results demonstrate the superiority of the proposed algorithm over other compared algorithms.
查看更多>>摘要：Multi-objective Optimization Evolutionary Algorithms (MOEAs) face numerous challenges when they are used to solve Many-objective Optimization Problems (MaOPs). Decomposition-based strategies, such as MOEA/D, divide an MaOP into multiple single-optimization sub-problems, achieving better diversity and a better approximation of the Pareto front, and dealing with some of the challenges of MaOPs. However, these approaches still require one to solve a multi-criteria selection problem that will allow a Decision-Maker (DM) to choose the final solution. Incorporating preferences may provide results that are closer to the region of interest of a DM. Most of the proposals to integrate preferences in decomposition-based MOEAs prefer progressive articulation over the "a priori " incorporation of preferences. Progressive articulation methods can hardly work without comparable and transitive preferences, and they can significantly increase the cognitive effort required of a DM. On the other hand, the "a priori " strategies do not demand transitive judgements from the DM but require a direct parameter elicitation that usually is subject to imprecision. Outranking approaches have properties that allow them to suitably handle non-transitive preferences, veto conditions, and incomparability, which are typical characteristics of many real DMs. This paper explores how to incorporate DM preferences into MOEA/D using the "a priori " incorporation of preferences, based on interval outranking relations, to handle imprecision when preference parameters are elicited. Several experiments make it possible to analyze the proposal's performance on benchmark problems and to compare the results with the classic MOEA/D without preference incorporation and with a recent, state-of-the-art preference-based decomposition algorithm. In many instances, our results are closer to the Region of Interest, particularly when the number of objectives increases.
Ji, JunzhongWeng, YannanYang, CuicuiWu, Tongxuan...
14页
查看更多>>摘要：In recent years, bacterial foraging optimization (BFO) has been used to solve multiobjective optimization problems (MOPs). However, BFO has not fully developed its potentials on MOPs for the reason of lacking of in-depth research on the optimization mechanisms and the diversity maintenance strategies. To solve it, this paper develops a multi-resolution grid-based BFO algorithm (called as MRBFO). MRBFO redesigns four tailored optimization mechanisms for MOPs including chemotaxis, conjugation, reproduction, and elimination and dispersal to search optimal nondominated solutions. Moreover, MRBFO defines a multi-resolution grid strategy to produce well-distributed diverse nondominated solutions. The performance of MRBFO is comprehensively evaluated by comparing it with several state-of-the-art algorithms on many benchmark test problems. The empirical results have sufficiently verified the advantages of MRBFO.
NSTL
Elsevier