Fig 3 - uploaded by Sebastian Lang
Content may be subject to copyright.
Source publication
The article investigates the application of NeuroEvolution of Augmenting Topologies (NEAT) to generate and parameterize artificial neural networks (ANN) on determining allocation and sequencing decisions in a two-stage hybrid flow shop scheduling environment with family setup times. NEAT is a machine-learning and neural architecture search algorith...
Contexts in source publication
Context 1
... use the open-source library neat-python (McIntyre et al., 2019) to implement the different solution strategies. Fig. 3 shows NEAT as flow chart. In the following, we briefly explain the illustrated steps of the ...
Context 2
... planning on the second stage aims to find the best combined allocation and sequencing strategy, the purpose of the third-stage experiments is to maximize the solution quality of the best strategy. Our intention is to figure out which solution quality is possible if a high training time is negligible. For this purpose, we adapt the hyper- Fig. 13. Analysing the influence of NEAT's mutation features on the solution quality based on (a) the average and (b) the best fitness of the population as well as (c) the resulting total tardiness for all datasets. For the first and second experiment, we predefined an ANN structure (d). However, when allowing topological mutations, NEAT ...
Context 3
... to our design of experiments, we want to analyse first whether NEAT's mutation features for the parameters and topologies of ANNs have a positive impact on the solution quality. Fig. 13 shows for all three experiments in the first stage (a) the logarithmic average fitness of the population, (b) the logarithmic fitness of the best genome, and (c) the total tardiness for each dataset over 10 observations. Furthermore, Fig. 13 illustrates the topology of (d) the manually designed ANN and (e) the ANN created by NEAT. The ...
Context 4
... features for the parameters and topologies of ANNs have a positive impact on the solution quality. Fig. 13 shows for all three experiments in the first stage (a) the logarithmic average fitness of the population, (b) the logarithmic fitness of the best genome, and (c) the total tardiness for each dataset over 10 observations. Furthermore, Fig. 13 illustrates the topology of (d) the manually designed ANN and (e) the ANN created by NEAT. The results clearly show that the manually designed ANN with randomly initialized parameters provides a worse solution quality than both ANNs that NEAT evolved over several generations. Although the second experiment, in which only ANN parameters ...
Similar publications
In this paper, we present and discuss an innovative approach to solve Job Shop scheduling problems based on machine learning techniques. Traditionally, when choosing how to solve Job Shop scheduling problems, there are two main options: either use an efficient heuristic that provides a solution quickly, or use classic optimization approaches (e.g.,...
Citations
... Authors in [14] recommended the use of decision trees to select priority rules, associating performance indicators with specific rules. The application of Neuro Evolution of Augmenting Topologies (NEAT) [15] in complex scheduling problems has demonstrated its superiority over traditional methods, offering a new perspective for hybrid two-stage workshops. Authors in [16] considered the use of ML for estimating processing times, thus optimizing the scheduling of parallel machines. ...
Production planning in supply chain management faces considerable challenges due to the dynamics and unpredictability of the production environment. Decision support systems based on the evolution of artificial intelligence can provide innovative solutions. In this paper, an approach based on machine learning techniques to solve the problem of scheduling the production of N products on M non-identical parallel machines is proposed. Using regression and classification models, our approach aims to predict overall production costs and assign products to the right machines. Some experiments carried out on simulated data sets demonstrate the relevance of the proposed approach. In particular, the XGBoost model stands out for its superior performance compared with the other tested ML algorithms. The proposed approach makes a significant contribution to the optimization of production scheduling, offering significant potential for improvement in Supply Chain Management.
... Experiments proved that these agents, whether used individually or integrated, outperform existing scheduling strategies across a wide range of problem sizes. NEAT is an evolutionary algorithm that simultaneously optimizes the structure and weights of neural networks using genetic algorithms [8]. This research proposes an NEAT algorithm that integrates the features of the DFJSP and designs heuristic rules for job selection and machine allocation. ...
... Dong et al. [33] explored the application of DRL and graph isomorphic networks in reducing the delay criteria for permutation flow shop scheduling issues. However, studies considering multiple dynamic events concurrently are scarce, with a notable example being Lang et al. [29] who developed the NEAT algorithm to address the two-stage HFSP incorporating product family set-up times. Consequently, this paper addresses the dynamic scheduling of hybrid flow shops, integrating flexible preventive maintenance for machines and multiple dynamic events such as dynamic job release and uncertain processing times. ...
... These functions assess the quality of the output produced by any artificial neural network solution. The NEAT algorithm enhances the weight parameters and the topology of neural networks, unlike traditional DRL algorithms such as DQN and AC [29,40]. NEAT utilises evolutionary algorithms to refine neural network structure and connection weights, addressing complex dynamic decision-making challenges. ...
... Offspring networks are produced via crossover and mutation, progressively evolving to yield an agent with the optimal scheduling policy. This methodology is illustrated in Figure 3. Based on the references [2,29], an interaction module between the agent and the hybrid flow shop system environment is crafted, with the corresponding pseudocode shown in Algorithm 1. The iteration process continues for several generations until convergence is achieved through evaluation, selection, crossover, and mutation. ...
A hybrid flow shop is pivotal in modern manufacturing systems, where various emergencies and disturbances occur within the smart manufacturing context. Efficiently solving the dynamic hybrid flow shop scheduling problem (HFSP), characterised by dynamic release times, uncertain job processing times, and flexible machine maintenance has become a significant research focus. A NeuroEvolution of Augmenting Topologies (NEAT) algorithm is proposed to minimise the maximum completion time. To improve the NEAT algorithm's efficiency and effectiveness, several features were integrated: a multi‐agent system with autonomous interaction and centralised training to develop the parallel machine scheduling policy, a maintenance‐related scheduling action for optimal maintenance decision learning, and a proactive scheduling action to avoid waiting for jobs at decision moments, thereby exploring a broader solution space. The performance of the trained NEAT model was experimentally compared with the Deep Q‐Network (DQN) and five classical priority dispatching rules (PDRs) across various problem scales. The results show that the NEAT algorithm achieves better solutions and responds more quickly to dynamic changes than DQN and PDRs. Furthermore, generalisation test results demonstrate NEAT's rapid problem‐solving ability on test instances different from the training set.
... Stochastic optimization methods, including Machine Learning, offer a way to overcome the challenge of exploring enormous state spaces by pre-selecting candidate solutions and finding acceptable-quality solutions in most cases. Due to the lack of labelled data (Schmidt and Stober, 2021), Deep Reinforcement Learning (Sutton and Barto, 2018) has emerged as a vital learning paradigm for scheduling problems (Gerpott et al., 2022;Kuhnle et al., 2021;Lang et al., 2021). However, these methods encounter challenges when the reward function is non-convex due to the use of gradient-based optimizations. ...
... To assess the practicality of our SA algorithm, we conducted an evaluation using the Tectron dataset (Gerpott et al., 2022;Lang et al., 2021), which consists of 170 orders from a real-world production facility. In order to provide a comprehensive comparison, we benchmarked our SA algorithm against several popular and state-of-the-art methods. ...
Inspired by the success of Simulated Annealing in physics, we transfer insights and adaptations to the scheduling domain, specifically addressing the one-stage job scheduling problem with an arbitrary number of parallel machines. In optimization, challenges arise from local optima, plateaus in the loss surface, and computationally complex Hamiltonian (cost) functions. To overcome these issues, we propose the integration of corrective actions, including symmetry breaking, restarts, and freezing out non-optimal fluctuations, into the Metropolis-Hastings algorithm. Additionally, we introduce a generalized Hamiltonian that efficiently fuses straightforward but widely applied processing-time cost functions. Our approach outperforms decision rules, meta-heuristics, and novel reinforcement learning algorithms. Notably, our method achieves these superior results in real-time, thanks to its computationally efficient evaluation of the Hamiltonian.
... Our proposed system aims to build a neuroEvolution-based classifier that leverages genetic programming to optimize both the network structure and weights. One of the most prominent neuroEvolution models is NeuroEvolution of Augmenting Topologies (NEAT) [7]. Our model's objective is to construct a high-accuracy classifier capable of detecting abnormalities in multivariate environments. ...
Abnormal behaviors degrade overall system efficiency and may lead to system suspension. Extensive academic research has focused on anomaly detection across various sectors, including industrial, information security, and artificial intelligence. In this study, we propose a NeuroEvolution AutoEncoder (NEAE) for classifying and predicting abnormalities in internet-based applications. The NEAE is genetically programmed to process dataset features that encompass abnormal behaviors in internet activities. Our approach targets multivariate anomalies using parallel dimension processing, aiming for synchronized intelligence. Genetic programming is central to our technique, enhancing optimization. We evaluate our proposed technique using a substantial internet-based dataset, specifically network traffic. Simulation results demonstrate the effectiveness of the NEAE in detecting abnormalities based on performance metrics. Given its relevance and integration with diverse industries such as the Internet of Things, wireless communication, network traffic, and the web, we choose an Internet-based application dataset to validate the NEAE’s performance.
... There have been very few publications of NEAT techniques being used for scheduling tasks. According to [23], there are only three publications to date excluding the aforementioned work on using NEAT for scheduling tasks. This work presents a perspective not yet realized for the use of NEAT as a scheduling learning model. ...
... The approach is based on the property of HFS to be presented as either an allocation or a sequencing problem. Their work concludes that NEAT can compete with state-of-the-art metaheuristic techniques in terms of quality of the solution presented and once trained they can determine the schedule in real time with the time complexity growing with production volume and complexity [23]. ...
This work presents a novel approach by utilizing Heterogeneous Activation Neural Networks (HA-NNs) to evolve the weights of Artificial Neural Networks (ANNs) for reinforcement learning in console and arcade computer games like Atari's Breakout and Sonic the Hedgehog. It is the first study to explore the potential of HA-NNs as potent ANNs in solving gaming-related reinforcement learning problems. Additionally, the proposed solution optimizes data transmission over networks for edge devices, marking a novel application of HA-NNs. The study achieved outstanding results, outperforming recent works in benchmark environments like CartPole-v1, Lunar Lander Continuous, and MountainCar-Continuous, with HA-NNs and ANNs evolved using the Neuroevolution of Augmenting Topologies (NEAT) algorithm. Notably, the key advancements include exceptional scores of 500 in CartPole-v1 and 98.2 in Mountain Car Continuous, demonstrating the efficacy of HA-NNs in reinforcement learning tasks. Beyond gaming, the research addresses the challenge of efficient data communication between edge devices, which has the potential to enhance performance in smart cities while reducing the load on edge devices and supporting seamless entertainment experiences with minimal commuting. This work pioneers the application of HA-NNs in reinforcement learning for computer games and introduces a novel approach for optimizing edge device communication, promising significant advancements in the fields of AI, neural networks, and smart city technologies.
... Wang, J.J. and Wang, L. studied an energy-aware distributed hybrid flow shop scheduling method based on reinforcement learning [4]. Lang et al. presented a dynamic scheduling method based on the NEAT algorithm for a two-stage hybrid flow shop scheduling problem with family setup times [5]. ...
... These tools include Reinforcement Learning (RL), Bayesian Optimization, Particle Swarm Optimization (PSO), and NeuroEvolution of Augmenting Topologies (NEAT), which are highly dependent upon advanced computational capabilities. [249,250] During the lifecycle analysis stage, advanced computer technology (e.g., ANSYS Granta Edupack) can aid in conducting lifecycle assessments to evaluate the environmental impact of battery designs and manufacturing processes. [251,252] It enables the identification of opportunities for optimization, resource efficiency, and reduced waste generation. ...
Lithium‐ion batteries (LIBs) have significantly impacted the daily lives, finding broad applications in various industries such as consumer electronics, electric vehicles, medical devices, aerospace, and power tools. However, they still face issues (i.e., safety due to dendrite propagation, manufacturing cost, random porosities, and basic & planar geometries) that hinder their widespread applications as the demand for LIBs rapidly increases in all sectors due to their high energy and power density values compared to other batteries. Additive manufacturing (AM) is a promising technique for creating precise and programmable structures in energy storage devices. This review first summarizes light, filament, powder, and jetting‐based 3D printing methods with the status on current trends and limitations for each AM technology. The paper also delves into 3D printing‐enabled electrodes (both anodes and cathodes) and solid‐state electrolytes for LIBs, emphasizing the current state‐of‐the‐art materials, manufacturing methods, and properties/performance. Additionally, the current challenges in the AM for electrochemical energy storage (EES) applications, including limited materials, low processing precision, codesign/comanufacturing concepts for complete battery printing, machine learning (ML)/artificial intelligence (AI) for processing optimization and data analysis, environmental risks, and the potential of 4D printing in advanced battery applications, are also presented.
... However, to check that our results are not particular to this choice, the Supplemental Material reports also results obtained by varying the number of allowed time durations N τ . To learn optimal policies solving the target-search problem, we exploit the evolutionary algorithm NeuroEvolution of Augmenting Topologies (NEAT) [49,50]. Therefor we equip each agent with an ANN taking as input the current state (BP or ABP) and outputs an action chosen among the set of actions described above, see Figure (1)c. ...
... To investigate how an evolutionary pressure allows the adaptive particles to develop successful target-search strategies, we resort to the genetic algorithm NeuroEvolution of Augmenting Topologies (NEAT) [49,50,61]. To do so, a simple Artificial Neural Network (ANN) is associated with our adaptive particle. ...
... During mutations, hidden nodes are generated with a probability p add and deleted with probability p del . Following standard practice [50,61], we set p add = p del = 0.05. ...
Developing behavioral policies designed to efficiently solve target-search problems is a crucial issue both in nature and in the nanotechnology of the 21st century. Here, we characterize the target-search strategies of simple microswimmers in a homogeneous environment containing sparse targets of unknown positions. The microswimmers are capable of controlling their dynamics by switching between Brownian motion and an active Brownian particle and by selecting the time duration of each of the two phases. The specific conduct of a single microswimmer depends on an internal decision-making process determined by a simple neural network associated with the agent itself. Starting from a population of individuals with random behavior, we exploit the genetic algorithm NeuroEvolution of Augmenting Topologies to show how an evolutionary pressure based on the target-search performances of single individuals helps to find the optimal duration of the two different phases. Our findings reveal that the optimal policy strongly depends on the magnitude of the particle's self-propulsion during the active phase and that a broad spectrum of network topology solutions exists, differing in the number of connections and hidden nodes.
... To learn optimal policies solving the target-search problem, we exploit the evolutionary algorithm NeuroEvolution of augmenting topologies (NEAT) [49,50]. Therefor we equip each agent with an ANN taking as input the current state (BP or ABP) and outputs an action chosen among the set of actions described above, see figure 1(c). ...
... To investigate how an evolutionary pressure allows the adaptive particles to develop successful target-search strategies, we resort to the genetic algorithm NEAT [49,50,61]. To do so, a simple ANN is associated with our adaptive particle. ...
... During mutations, hidden nodes are generated with a probability p add and deleted with probability p del . Following standard practice [50,61], we set p add = p del = 0.05. ...
Developing behavioral policies designed to efficiently solve target-search problems is a crucial issue both in nature and in the nanotechnology of the 21st century. Here, we characterize the target-search strategies of simple microswimmers in a homogeneous environment containing sparse targets of unknown positions. The microswimmers are capable of controlling their dynamics by switching between Brownian motion and an active Brownian particle and by selecting the time duration of each of the two phases. The specific conduct of a single microswimmer depends on an internal decision-making process determined by a simple neural network associated with the agent itself. Starting from a population of individuals with random behavior, we exploit the genetic algorithm NeuroEvolution of Augmenting Topologies to show how an evolutionary pressure based on the target-search performances of single individuals helps to find the optimal duration of the two different phases. Our findings reveal that the optimal policy strongly depends on the magnitude of the particle’s self-propulsion during the active phase and that a broad spectrum of network topology solutions exists, differing in the number of connections and hidden nodes.