Search or ask a question

What are the current advances in multi-objective optimization techniques for complex systems?

Robustness (evolution)

Particle swarm optimization

Evolutionary computation

Evolutionary algorithm

Best insight from top research papers

Multi-objective optimization techniques for complex systems have seen significant advances in recent years. These techniques have been applied to various fields such as bioinformatics, wireless networks, natural language processing, image processing, and astronomy. The use of multi-objective evolutionary algorithms (MOEAs) has proven to be effective in solving these optimization problems. MOEAs help users quickly obtain the desired data from complex network data, improving efficiency . Additionally, the integration of simulation-based optimization methods has shown promising results in optimizing complex urban mass rapid transit systems. The proposed two-phase algorithm combines the covariance matrix adaptation evolution strategy with a multi-directional local search, outperforming other comparative approaches . Furthermore, the use of surrogate models and the Dempster-Shafer theory has been successful in improving the computational efficiency of multi-objective network robustness optimization . These advancements in multi-objective optimization techniques have opened up new possibilities for solving real-world problems and have paved the way for future research in this field.

Answers from top 4 papers

PDF

Open Access

More filters

Papers (4)	Insight
Open access•Journal Article•DOI Multi-objective simulation optimization for complex urban mass rapid transit systems David Schmaranzer, Roland Braune, Karl F. Doerner - Show less +2 more 01 Oct 2021-Annals of Operations Research 11 Citations	The provided paper discusses a two-phase algorithm for multi-objective simulation-based headway optimization in complex urban mass rapid transit systems. It compares the proposed method with other multi-objective evolutionary algorithms and shows its superiority in terms of spread and accuracy of the resulting Pareto front approximation. However, the paper does not specifically mention the current advances in multi-objective optimization techniques for complex systems.
Open access•Journal Article•DOI Multi-Objective Optimization of the Robustness of Complex Networks Based on the Mixture of Weighted Surrogates Junfeng Nie, Zhuoran Yu, Junli Li - Show less +2 more 21 Apr 2023-Axioms	The provided paper discusses the use of surrogate models and the Dempster-Shafer theory for multi-objective optimization of network robustness. It does not specifically mention other current advances in multi-objective optimization techniques for complex systems.
Open access•Proceedings Article•DOI Dynamic multi‐objective optimisation of complex networks based on evolutionary computation Linfeng Huang 12 Oct 2022	The paper discusses the use of multi-objective evolutionary algorithms (MOEA) for optimizing complex networks, but does not specifically mention current advances in multi-objective optimization techniques for complex systems.
Journal Article•DOI Multi-objective optimization techniques: a survey of the state-of-the-art and applications Naveen Saini, Sriparna Saha - Show less +1 more 01 Sep 2021-European Physical Journal-special Topics 25 Citations	The paper provides a survey of recently developed multi-objective optimization algorithms, but does not specifically mention current advances in multi-objective optimization techniques for complex systems.

My columns

Related Questions

What is the latest research on integrating multi-objective optimization and deep learning?4 answersThe latest research on integrating multi-objective optimization and deep learning involves innovative approaches such as using variational auto-encoders (VAE) for optimizing different machine technologies simultaneously. Additionally, Deep Belief Networks (DBN) have been applied to optimize ship design in nonlinear waves, showcasing reductions in total resistance and transfer functions. Furthermore, a framework integrating multi-objective evolutionary algorithms with adaptive parameter control using Deep Reinforcement Learning (DRL) has been proposed, demonstrating improved solution quality and computation time for complex optimization problems like warehouse design. Moreover, the integration of IoT and Differential Evolution Algorithm (DEA) into deep learning models has been explored for multi-objective optimal scheduling of reservoirs, ensuring ecological flow and operational reliability.

What are the latest advances in quantum algorithm design and optimization techniques?4 answersQuantum computing is an emerging technology that utilizes quantum mechanical properties to perform compute-intensive operations faster than classical computers. Recent advances in quantum algorithm design and optimization techniques have shown promising results. One approach is the use of ZX calculus to optimize quantum arithmetic algorithms, reducing hardware resources and the number of qubits required. Another advancement is the development of a quantum algorithm for finding the minimum of a function using multistep quantum computation, which can efficiently solve optimization problems with continuous variables. Optimization tools have also been introduced for quantum information theory, including algorithms for quantum state estimation, certifying quantum non-locality, and solving the quantum marginal problem. The Quantum Approximate Optimization Algorithm has been proposed as a method for solving optimization problems on quantum processors, with optimal parameters reducing to one free variable. Additionally, the Quantum Multivariable Optimization Algorithm has been introduced for efficiently optimizing continuous multivariable functions, surpassing the limits of unstructured quantum search algorithms.

What are some of the latest advances in AI?4 answersSignificant growth in the development and deployment of artificial intelligence (AI) is being witnessed. Various AI sensors provide cost-effective approaches for monitoring applications in smart homes and personal healthcare. These sensors include complementary metal oxide semiconductor image sensors for computer vision, microelectromechanical systems, microphone sensors for voice recognition, and wearable sensors for gesture recognition. Explainable Artificial Intelligence (E-AI) is a method that enables the understanding and improvement of outcomes generated by AI algorithms. E-AI has the potential to accelerate drug discovery and reduce costs. Intelligent systems and AI techniques have advanced significantly, leading to more efficient and accurate systems in computer vision, natural language processing, and autonomous systems. Future advancements include deep learning, explainable AI, transfer learning, and human-AI collaboration. AI tools, such as machine learning algorithms, have been applied to enhance wireless connectivity in cyber-physical systems (CPS) and Internet of Things (IoT). ML paradigms like transfer learning, distributed learning, and federated learning have been used to leverage large data for learning and adaptation in wireless CPS and IoT systems.

How do different multi-objective optimization algorithms compare?5 answersDifferent multi-objective optimization algorithms have been compared in various studies. One study compared heuristic algorithms and the sequential quadratic programming (SQP) method for obtaining Pareto-optimal solutions. The particle swarm optimization algorithm was the only heuristic algorithm that consistently obtained optimal solutions, while the SQP method was superior in terms of speed and accuracy for certain functions. Another study evaluated the speed of evolution towards the edge of chaos using NSGA-II and SPEA2 algorithms, with NSGA-II showing better convergence speed. A novel ranking scheme was proposed to compare meta-heuristic multi-objective optimization algorithms, reducing potential information loss when statistical significance is not observed in high-dimensional data. A study proposed and evaluated different algorithmic implementations of multi-objective optimization, robustness consideration, and multi-criterion decision-making together, discussing the advantages of different integration approaches. Finally, a study proposed a graphical interface to compare the performance of NSGA-II and MOPSO algorithms, allowing non-expert users to analyze and determine the best-suited algorithm for their needs.

What are the latest advances in machine learning?5 answersMachine learning has seen several recent advances. Optimal Transport has emerged as a probabilistic framework in machine learning, offering new solutions for generative modeling and transfer learning. Computational Optimal Transport has also developed, impacting machine learning practice. Another area of progress is the usage of machine learning for electrochemical sensors, particularly in the analysis of data generated by sensing and biosensing methods. Additionally, machine learning models have been widely used for predicting the toxicity of small molecules, aiding in drug discovery by filtering out molecules with a high probability of failing in clinical trials. These models have been applied to various toxic endpoints, such as acute oral toxicity, hepatotoxicity, and mutagenicity. Overall, these advancements highlight the growing importance and diverse applications of machine learning in different domains.

How can multi-objective optimization be used to solve complex problems?5 answersMulti-objective optimization can be used to solve complex problems by simultaneously optimizing multiple conflicting objectives. This approach involves finding a set of compromise solutions, known as non-dominated or Pareto-optimal solutions, that provide the best trade-offs between the objectives. Various algorithms and techniques have been proposed to tackle multi-objective optimization problems. For example, one approach integrates multi-objective evolutionary algorithms with adaptive parameter control using deep reinforcement learning. Another approach combines neural networks with multi-objective genetic algorithms to handle strict constraints effectively. Additionally, a guided multi-objective equilibrium optimizer algorithm has been developed, which utilizes an external archive and Pareto dominance to achieve better exploration and diversity during the optimization process. Furthermore, the Corona algorithm, inspired by the epidemic behavior of the Corona virus, has been expanded to solve multi-objective programming problems. Finally, a multi-strategy Remora Optimization Algorithm has been proposed, incorporating random restart, information entropy evaluation, and visual perception strategies to improve optimization capabilities.

See what other people are reading

What is formula of smote pso?

The formula for SMOTE-PSO (Synthetic Minority Oversampling Technique - Particle Swarm Optimization) involves optimizing the selection of key parameters in SMOTE using metaheuristic optimization algorithms like PSO. SMOTE is a technique used to address imbalanced datasets by generating synthetic samples for the minority class. In the context of optimizing SMOTE parameters, metaheuristic algorithms like PSO and BAT (Bat-inspired algorithm) are employed to find the best parameter values for enhancing classifier performance in data mining tasks. Additionally, in the context of breast cancer patient classification, a hybrid algorithm combining SMOTE, PSO, and classifiers like C5 decision tree model is proposed to improve the effectiveness of classification for imbalanced datasets. This integration of SMOTE with PSO showcases enhanced performance in dealing with imbalanced data for classification tasks.Can reinforcement learning be used to improve decision-making processes in complex systems such as autonomous vehicles or financial markets?

Reinforcement learning (RL) is a valuable tool for enhancing decision-making in intricate systems like autonomous vehicles and financial markets. RL methods, which do not heavily rely on model assumptions, have been successfully applied in finance for tasks such as portfolio optimization, optimal execution, and market making. However, challenges like designing reward functions in noisy environments can be mitigated by imitation learning (IL) algorithms, which learn from expert demonstrations to improve decision-making processes. Moreover, innovative frameworks like Cognitive Escape Reinforcement Learning (CERL) have been introduced to address issues like local optima trapping and weak generalization, showcasing superior performance in decision-making tasks such as visual navigation. Additionally, in safety-critical domains like autonomous driving, constrained adversarial reinforcement learning approaches have been proposed to ensure robust decision-making against environmental uncertainties, demonstrating effectiveness in highway on-ramp merging scenarios.What is value of pre existing knowledge?

Pre-existing knowledge holds significant value in various domains. In the realm of machine learning, leveraging prior knowledge through informed pre-training on condensed knowledge prototypes enhances learning speed, generalization capabilities, and model robustness. In the educational context, understanding correlations between students' past performance can shed light on their ability to apply acquired knowledge in future studies. Moreover, in the field of medical education, pre-existing knowledge of CPR guidelines among 6th-year medical students positively influences their training outcomes, with those possessing prior knowledge showing higher levels of improvement post-training. Therefore, pre-existing knowledge acts as a foundation that can significantly impact learning outcomes and skill development across various disciplines.What is reliability in electrical engineering?

Reliability in electrical engineering refers to the ability of electrical components and systems to function without failure. In the power industry, reliability engineering has become crucial for decision-making, emphasizing the importance of reliable electrical power systems. Power electronics play a significant role in energy systems, impacting security, safety, efficiency, and cost, making the reliability of electronic components and converters vital for overall system reliability. Predicting reliability indices involves analyzing temperature fields of power semiconductor devices, where factors like ambient temperature and dimensional orientation affect failure rates. To enhance power system security and reliability, tools like graph theory-based methodologies are developed for classifying bus reliability based on plant engagement, aiding in mitigating blackout risks. Reliability testing is essential to investigate failure mechanisms and ensure robustness in components and converters.How software grasshopper is more faster than others using grasshopper?

The software Grasshopper is faster than others due to various enhancements and modifications made to the Grasshopper Optimization Algorithm (GOA). These improvements include dynamic dual elite learning and sinusoidal mutation in the Enhanced Grasshopper Optimization Algorithm (EMGOA), the integration of a crazy factor to induce sudden direction changes for enhanced diversity in the Crazy-GOA, and the combination of Gaussian and chaos theory to improve search ability in the Gaussian-Chaos Grasshopper Optimization Algorithm (GC-GOA). These modifications lead to faster convergence speeds, avoidance of local optima, and better search performance compared to other swarm intelligence algorithms like ALO and PSO. Additionally, the utilization of Grasshopper in parametric design platforms contributes to rapid modeling of spatial structures, improving efficiency in design realization.What is multimodal learning?

Multimodal learning involves learning across various input modalities like video, audio, and text. It explores how models adapt when the modalities differ between training and deployment, a common scenario in applications involving hardware platforms. This field addresses challenges in utilizing diverse data sources for intelligent agent design, encompassing linguistic, acoustic, visual, tactile, and physiological information. Techniques like multimodal robustness frameworks and deep hierarchical multimodal metric learning have been developed to enhance learning outcomes by leveraging multiple modalities effectively. Despite facing obstacles like noise and lack of information, multimodal learning models have shown promise in improving learning outcomes and teaching effectiveness, especially in open learning environments.What is mismatched disturbance?

Mismatched disturbance refers to disturbances in a system that are not accounted for or directly measured in the control design, leading to potential performance degradation or instability. In the context of vehicle platoons and power converters, mismatched disturbances such as velocity and acceleration disturbances in platoons, uncertainties in DC-DC boost converters, and load mutations in solar electric vehiclesare considered. Strategies like finite-time disturbance observers, robust H∞ control with mismatched disturbance compensation, and repetitive control with mismatched disturbance observers are proposed to estimate and counteract these disturbances, enhancing system robustness and performance. These approaches aim to mitigate the impact of mismatched disturbances on system stability and control effectiveness.How has neural network been applied in the epidemiology compartment model of COVID-19 ?

Neural networks have been innovatively applied to enhance the epidemiological compartment models of COVID-19, addressing some of the traditional models' limitations and improving prediction accuracy and robustness. Specifically, a deep learning approach utilizing a multi-variate Long Short-term Memory (LSTM) recurrent neural network has been proposed to predict future COVID-19 infection cases and deaths with improved predictive performance by incorporating mobility data and training on multiple time series samples simultaneously. This approach offers fine geographic granularity in predictions and reduces data requirements and computational intensity by substituting a single model for multiple models in an ensemble model, demonstrating similar accuracy and spatial patterns with standard ensemble models used as benchmarks. Furthermore, neural networks have been applied through a deep recurrent reinforced learning-based model to evaluate and predict the spatiotemporal effects of various control measures on COVID-19 cases and mortality. This model uses a combination of control measures, official statistics, and government data to closely match actual values of cases and mortality, highlighting the importance of adjunctive measures such as international travel and quarantine restrictions. Additionally, machine learning techniques, including neural networks, have been employed to project COVID-19 infections and deaths, comparing traditional time series statistical models with modern machine learning approaches. This research identifies the logistic growth model as the best fit for the pandemic's behavior in Mexico, emphasizing the correlation of climate and mobility variables with disease numbers and proposing a model to predict daily cases and fatalities using time series data, mobility, and weather variables. These applications of neural networks in epidemiological compartment models of COVID-19 represent a significant advancement in the field, offering more accurate, robust, and geographically detailed predictions that can better inform public health decisions and interventions.How to improve performances in Graph Convolutional Network?

To enhance performance in Graph Convolutional Networks (GCNs), various strategies have been proposed. One approach involves utilizing t-hopGCN to describe t-hop neighbors by the shortest path between nodes, enhancing node classification on graphs without node features. Another method restricts polynomial coefficients in spectral filters to prevent overfitting, significantly improving GCN performance in large receptive fields and across datasets. Additionally, incorporating wavelet denoising filters into GCNs unifies different architectures under a common framework, offering a trade-off between local and global information for improved performance on diverse datasets. Furthermore, the deep graph convolutional generative adversarial network (DGCGAN) combines GCN and DCGAN to address non-robustness and low accuracy issues, outperforming baselines in semi-supervised node classification and image classification tasks.Does placement of IMU impact accuracy when measuring movemente?

The placement of Inertial Measurement Units (IMUs) significantly impacts the accuracy when measuring movements. Research highlights that variations in sensor attachment locations can lead to differences in the accuracy of measuring joint angles and gait parameters. Studies emphasize the necessity to reevaluate sensor systems and algorithms when the sensor position changes to maintain reliability in monitoring systems. Different sensor positions on the body can result in varying bias and root mean square error values, affecting the precision of IMU-based kinematic measurements for different functional movements. Therefore, optimizing IMU placement is crucial for accurate joint angle estimation and gait analysis, especially in clinical settings where IMUs are used for diagnosing and rehabilitating gait pathologies.How does optical camera communication differ from other forms of communication in IoV systems?

Optical camera communication (OCC) in the Internet of Vehicles (IoV) systems stands out due to its utilization of image sensors for data reception, offering spatial separation properties and robustness to ambient noise, making it ideal for outdoor applications like intelligent transportation systems (ITSs). OCC, a form of visible light communication, leverages commercial off-the-shelf devices, providing a wide range of services in various areas such as intelligent transportation, indoor positioning, and the Internet of Things (IoT). By integrating Color Shift Keying modulation with the Rolling Shutter mechanism, OCC can achieve higher data rates compared to other modulation schemes, making it a promising technology for wireless communication. Additionally, OCC systems have been experimentally demonstrated to achieve high data rates, such as 150 kbit/s, using innovative transmitter designs.