Evaluating the fitness of individuals in the initial population of an evolutionary algorithm (EA) is usually straightforward and poses few theoretical problems. In asynchronous steady-state EAs (ASSEAs), however, the ...
详细信息
ISBN:
(纸本)9798400701207
Evaluating the fitness of individuals in the initial population of an evolutionary algorithm (EA) is usually straightforward and poses few theoretical problems. In asynchronous steady-state EAs (ASSEAs), however, the choice of initialization strategy can significantly alter the algorithm's long-term behavior. ASSEAs have long been recognized as an important alternative to parallel and distributed evolution, because unlike the more commonly used generational model, they avoid leaving processing resources idle while waiting for a generation boundary. In our work on ASSEAs, we have observed that colleagues' intuitions tend to differ about what a basic asynchronous initialization strategy should look like-but the implications of this choice have not been studied before. This paper analyzes the results of three competing initialization strategies that have appeared in prior literature-the immediate, until-finished, and extra strategies-and concludes that the immediate strategy in particular, which relies on maintaining a queue of jobs waiting to be evaluated, incurs a slow evolutionary feedback loop and should be used with caution. Queuing is a necessary part of scaling ASSEAs up to be resilient to node failures in large HPC environments, however-so this work suggests that further research is needed to understand how asynchronous initialization can be used most effectively for expensive optimization.
Human machine interfaces (HMIs) are an essential part of our daily lifes. Generally, a shift from mechanical HMIs such as knobs to more electronics-based solutions can be observed. A well-known example are touch inter...
详细信息
Human machine interfaces (HMIs) are an essential part of our daily lifes. Generally, a shift from mechanical HMIs such as knobs to more electronics-based solutions can be observed. A well-known example are touch interfaces which may be found in a variety of products ranging from phones over cars to home appliances. Touch interfaces offer more usability but pose increased technical challenges which must be faced by the development team in order to meet customer demands. Additionally, legal obligations regarding their robustness must be fullfilled (even under noise) in order to allow the novel products to enter the market. This is often enabled by employing a dedicated signal filter chain which is to be set accordingly. The corresponding calibration process is often time intensive and error-prone if performed manually. This is for example the case for our industrial partner which is a major producer in the consumer electronics market. This work analyses if the pain of a manual calibration can be resolved. This task consists of probing new calibration(s) and measuring their performance as well as choosing a calibration. The former is automated by developing a corresponding testbed which can simulate various noise classes and can perform touch interactions on a real user interface using a six-axis robot. The novel calibrations are determined by a genetic algorithm (GA) which is a special optimizer out of the family of evolutionary algorithms. The optimization is the main focus of this study. Quantitatively the GA was analysed by benchmarking it successfully against several other optimizers as well as against a human expert. Furthermore, the search time of the GA could be reduced by improving the stopping criterion. The analysis is rounded up by a qualitative analysis which showed that the GA’s calibrations can withstand noise to a certain degree whilst still identifying touch events if and only if they occurred. Over all the study outlined that an automation of the manu
Efficient energy management is critical to building inclusive, safe, resilient, and sustainable cities and human settlements. Optimizing the operation and planning of smart grids is crucial in this regard and remains ...
详细信息
ISBN:
(纸本)9798400701207
Efficient energy management is critical to building inclusive, safe, resilient, and sustainable cities and human settlements. Optimizing the operation and planning of smart grids is crucial in this regard and remains an active research area. The "Competition on evolutionary computation in the Energy Domain" has been held annually since 2017. Its 2023 edition focuses on two problems: Risk-based optimization of energy resource management considering the uncertainty of high penetration of distributed energy resources, and Long-term transmission network expansion planning. In this paper, we apply the RCED-UMDA algorithm to solve these problems, and our experimental results demonstrate its superiority over the top three algorithms of the 2022 and 2021 competition editions.
With countries' work in reducing emissions and customers' preference to environmental protection, electric vehicles increase rapidly following by the urgent need for charging stations. The vehicle energy stati...
详细信息
ISBN:
(纸本)9798350321456
With countries' work in reducing emissions and customers' preference to environmental protection, electric vehicles increase rapidly following by the urgent need for charging stations. The vehicle energy station distribution problem tries to guarantee the convenience of both gas vehicle and electric vehicle customers by reasonably allocating the gas stations and charging stations. In practice, the vehicle energy station investors are driven by individual profits, the government needs to guide the investors by using incentive. In order to solve this problem, we design a social-benefit based incentive distribution method with the following three features. Firstly, we design a social-benefit indicator to evaluate the vehicle energy station distribution program. Secondly, we apply the network-based evolutionary game as simulator to model investors' decision-making under incentive. Thirdly, we propose an evolutionary framework to determine the incentive distribution program according to the evolutionary game simulation result. Experiments are performed on the real station map and the results show our method can appropriately determine the incentive distribution program and increase the social benefit of the vehicle energy station distribution program.
Neural architecture search (NAS) is becoming increasingly popular for its ability to automatically search for an appropriate network architecture, avoiding laborious manual designing processes, and potentially introdu...
详细信息
ISBN:
(纸本)9798400701191
Neural architecture search (NAS) is becoming increasingly popular for its ability to automatically search for an appropriate network architecture, avoiding laborious manual designing processes, and potentially introducing novel structures. However, many NAS methods suffer from heavy computational consumption. One-shot NAS alleviates this issue by training a big supernet and allowing all the candidates to inherit weights from the supernet, avoiding training from scratch. However, the performance evaluations in one-shot methods might not always be reliable due to the weight co-adaption issue inside the supernet. This paper proposes a supernet fine-tuning strategy to allow the supernet's weights to adapt to the new focused search region along with the evolutionary process. Furthermore, a new genetic algorithm-based search method is designed to offer an effective path-sampling strategy in the search region and provide a new population generation method to preclude unfair fitness comparisons between different populations. The experimental results demonstrate the proposed method achieves promising results compared with 32 peer competitors in terms of the algorithm's computational cost and the searched architecture's performance. Specifically, the proposed method achieves classification error rates of 2.50% and 17.07% within only 0.50 and 0.92 GPU-days on CIFAR10 and CIFAR100, respectively.
Dynamic Optimization Problems (DOPs) are characterized by changes in the fitness landscape that can occur at any time and are common in real world applications. The main issues to be considered include detecting the c...
详细信息
ISBN:
(纸本)9798400701207
Dynamic Optimization Problems (DOPs) are characterized by changes in the fitness landscape that can occur at any time and are common in real world applications. The main issues to be considered include detecting the change in the fitness landscape and reacting in accord. Over the years, several evolutionary algorithms have been proposed to take into account this characteristic during the optimization process. However, the number of available tools or open source codebases for these approaches is limited, making reproducibility and extensive experimentation difficult. To solve this, we developed a component-oriented framework for DOPs called Adjustable Components for Dynamic Problems (AbCD), inspired by similar works in the Multiobjective static domain. Using this framework, we investigate components that were proposed in several popular DOP algorithms. Our experiments show that the performance of these components depends on the problem and the selected components used in a configuration, which differs from the results reported in the literature. Using irace, we demonstrate how this framework can automatically generate DOP algorithm configurations that take into account the characteristics of the problem to be solved. Our results highlight existing problems in the DOP field that need to be addressed in the future development of algorithms and components.
The depiction of a city's facade can have various purposes, from purely decorative use to documentation for future restoration. This representation is often a manual and time-consuming process. This paper describe...
详细信息
ISBN:
(纸本)9783031490101;9783031490118
The depiction of a city's facade can have various purposes, from purely decorative use to documentation for future restoration. This representation is often a manual and time-consuming process. This paper describes the co-creative system Evolving Urban Landscapes, which uses evolutionary computation to produce images that represent the landscape of an input city. In order to evaluate the creativity of the system, we conducted a study with 23 users. The results show that the system we created can be considered creative and, above all, that it generates diverse results, allowing the users to evolve landscapes according to their tastes.
Flapping-Wing Micro Air Vehicles (FW-MAVs) may lose the ability to reliably follow waypoint trajectories because ongoing accumulation of wing damage can render invalid the system models implicit in the vehicle's f...
详细信息
ISBN:
(纸本)9798350395747
Flapping-Wing Micro Air Vehicles (FW-MAVs) may lose the ability to reliably follow waypoint trajectories because ongoing accumulation of wing damage can render invalid the system models implicit in the vehicle's flight controller. Previously we reported using an evolutionary Algorithm (EA) to continuously adapt hardware-coded core wing motions to force the vehicle to remain compliant with the needs of the controller, rather than perhaps more apparent technique of adapting the controller models to the hardware. In those works, we provided strong evidence that the proposed technique could restore proper waypoint tracking behavior. In additional work, we also demonstrated for a simplified version of the problem that we could use the same machine learning to actually determine the nature of the wing damage and update the models internal to the main flight controller. We also argued that this indirect means of recovering models was likely more applicable for situations in which online, ongoing, learning was required and in which more traditional system identification was difficult to apply. In this work, we will provide evidence, for the first time, that this model extraction capability is feasible for the full, unconstrained, 3D flight. In addition, as a consequence of this additional analysis, we will further improve the yield of successful learning trials.
Recently, Boundary Control Methods (BCMs) have become increasingly relevant in the field of metaheuristic algorithms. In this study, we investigate the relationship between the activation frequency of different BCMs a...
详细信息
ISBN:
(纸本)9798400701207
Recently, Boundary Control Methods (BCMs) have become increasingly relevant in the field of metaheuristic algorithms. In this study, we investigate the relationship between the activation frequency of different BCMs and the problem's dimensionality. Additionally, we analyze each problem dimension independently. Our research primarily concentrates on the top three algorithms from the IEEE CEC 2020 competition: AGSK, IMODE, and j2020, utilizing the competition benchmark set to conduct experiments. Our findings provide valuable insights into the metaheuristic domain, underlining the significance of comprehending BCM activation patterns to improve algorithm design and benchmarking practices.
We propose a two-stage surrogate-assisted evolutionary approach to address the computational issues arising from using Genetic Algorithm (GA) for feature selection in a wrapper setting for large datasets. The proposed...
详细信息
ISBN:
(纸本)9798400701207
We propose a two-stage surrogate-assisted evolutionary approach to address the computational issues arising from using Genetic Algorithm (GA) for feature selection in a wrapper setting for large datasets. The proposed approach involves constructing a lightweight qualitative meta-model by sub-sampling data instances and then using this meta-model to carry out the feature selection task. We define "Approximation Usefulness" to capture the necessary conditions that allow the meta-model to lead the evolutionary computations to the correct maximum of the fitness function. Based on our procedure we create CHCQX a Qualitative approXimations variant of the GA-based algorithm CHC (Cross generational elitist selection, Heterogeneous recombination and Cataclysmic mutation). We show that CHCQX converges faster to feature subset solutions of significantly higher accuracy, particularly for large datasets with over 100K instances. We also demonstrate the applicability of our approach to Swarm Intelligence (SI), with results of PSOQX, a qualitative approximation adaptation of the Particle Swarm Optimization (PSO) method. A GitHub repository with the complete implementation is available(2). This paper for the Hot-off-the-Press track at GECCO 2023 summarizes the original work published at [3].
暂无评论