In this letter we revisit the famous heavy ball method and study its global convergence for a class of non-convex problems with sector-bounded gradient. We characterize the parameters that render the method globally c...
详细信息
In this letter we revisit the famous heavy ball method and study its global convergence for a class of non-convex problems with sector-bounded gradient. We characterize the parameters that render the method globally convergent and yield the best R-convergence factor. We show that for this family of functions, this convergence factor is superior to the factor obtained from the triple momentum method.
In modern urban planning, agricultural production, and environmental monitoring, classification tasks necessitate sophisticated approaches due to the inherent complexity of hyperspectral images (HSIs), characterized b...
详细信息
In modern urban planning, agricultural production, and environmental monitoring, classification tasks necessitate sophisticated approaches due to the inherent complexity of hyperspectral images (HSIs), characterized by their abundant spectral bands and consequent high dimensionality. Such profusion poses significant challenges for effective data processing, analysis, and classification. Addressing these challenges, the application of deep learning, particularly Convolutional Neural Networks (CNNs), has emerged as a pivotal advancement. By exploiting the inherent spectral correlation and spatial context, these networks are adept at extracting pertinent features from the high-dimensional data of HSIs, thereby significantly enhancing classification performance. This study introduces four novel deep learning models optimized with the Adam algorithm: a 3-Dimensional Convolutional Neural Network (3D-CNN), a 2-Dimensional Convolutional Neural Network (2D-CNN), a recurrent 3D-CNN (R-3D CNN), and a recurrent 2D-CNN (R-2D-CNN). The Adam optimizer, known for its adaptive nature and the utilization of moving averages of gradients and their squares, is demonstrated to efficiently handle sparse gradients, thus providing stability during the optimization process. Comprehensive analyses were conducted on two publicly available databases-Indian Pine and Pavia University-yielding notable results. The employment of the Adam optimizer facilitated the attainment of exceptional performance metrics, evidenced by a kappa coefficient of 99.6%, a processing time of 322.18 seconds, an overall accuracy of 99.97%, and an average accuracy of 97.6% for the Indian Pine dataset;similarly, for the Pavia University dataset, results showcased a kappa coefficient of 99.1%, a processing time of 293.74 seconds, an overall accuracy of 99.98%, and an average accuracy of 99.99%. These findings underscore the superiority of the proposed deep learning models, particularly the R-3D-CNN and R-2D-CNN, over tradi
Iterative learning control (ILC) is a control strategy for repetitive tasks wherein information from previous runs is leveraged to improve future performance. optimization-based ILC (OB-ILC) is a powerful design frame...
详细信息
Iterative learning control (ILC) is a control strategy for repetitive tasks wherein information from previous runs is leveraged to improve future performance. optimization-based ILC (OB-ILC) is a powerful design framework for constrained ILC where measurements from the process are integrated into an optimization algorithm to provide robustness against noise and modelling error. This paper proposes a robust ILC controller for constrained linear processes based on the forward-backward splitting algorithm. It demonstrates how structured uncertainty information can be leveraged to ensure constraint satisfaction and provides a rigorous stability analysis in the iteration domain by combining concepts from monotone operator theory and robust control. Numerical simulations of a precision motion stage support the theoretical results.
High-performance concrete (HPC) outperforms regular concrete due to incorporating additional components that go beyond the typical ingredients used in standard concrete. Various artificial analytical methods were empl...
详细信息
High-performance concrete (HPC) outperforms regular concrete due to incorporating additional components that go beyond the typical ingredients used in standard concrete. Various artificial analytical methods were employed to assess the compressive strength (CS) of high-performance concrete containing fly ash (FA) and blast furnace slag (BFS). The primary objective of this study was to present a practical approach for a comprehensive evaluation of machine learning algorithms in predicting the CS of HPC. The study focuses on utilizing the adaptive neuro-fuzzy inference system (ANFIS) to develop models for predicting HPC characteristics. To enhance the performance of ANFIS methods, the study incorporates the arithmetic optimization algorithm (AOA) and equilibrium optimizer (EO) (abbreviated as ANAO and ANEO, respectively). Notably, this research introduces novelty through the application of the AOA and EO, the evaluation of HPC with additional components, the comparison with prior literature, and the utilization of a large dataset with multiple input variables. The results indicate that the combined ANAO and ANEO systems demonstrated strong estimation capabilities, with R-2 values of 0.9941 and 0.9975 for the training and testing components of ANAO, and 0.9878 and 0.9929 for ANEO, respectively. The results comparison of this study presented the comprehensiveness and reliability of the created ANFIS model optimized with AOA for predicting the HPC's CS improved with FA and BFS, which could be applicable for practical usages.
In model predictive control fast and reliable quadratic programming solvers are of fundamental importance. The inherent structure of the subsequent optimal control problems can lead to substantial performance improvem...
详细信息
In model predictive control fast and reliable quadratic programming solvers are of fundamental importance. The inherent structure of the subsequent optimal control problems can lead to substantial performance improvements if exploited. Therefore, we present a structure-exploiting solver based on proximal augmented Lagrangian, extending the general-purpose quadratic programming solver QPALM. Our solver relies on semismooth Newton iterations to solve the inner sub-problem while directly accounting for the optimal control problem structure via efficient and sparse matrix factorizations. The matrices to be factorized depend on the active-set and therefore low-rank factorization updates can be employed like in active-set methods resulting in cheap iterates. We compare our solver with QPALM and other well-known solvers and show its benefits in a numerical example.
Regression techniques were developed to determine the concrete initial (Gf) and total (GF) fracture energy based on prior data using mechanical features and mixed design elements. There were 264 samples retrieved from...
详细信息
Regression techniques were developed to determine the concrete initial (Gf) and total (GF) fracture energy based on prior data using mechanical features and mixed design elements. There were 264 samples retrieved from prior investigations in all. Research contributes to the field by improving the accuracy of predicting concrete fracture energy, offering a methodology for optimizing hyperparameters, and providing a model comparison that demonstrates the practical value of the new approach. These findings can benefit the construction and engineering industries by enhancing the accuracy of material property predictions and improving the quality and safety of constructed structures. This study merged support vector regression (SVR) assessment with arithmetic optimization algorithm (AOA) and whale optimization algorithm (WOA) to predict the Gf and GFF of concrete. The aim of combining the optimization algorithms with SVR analysis was to determine the optimal values of hyperparameters that play pivotal role in developed models' accuracy. The computation and analysis for Gf and GF using five criteria shows that optimized SVR-AOA and SVR-WOA analyses can do admirably well throughout the forecasting model. When the outperforming SVR analysis was compared to the library, it was discovered that the newly constructed SVR-AOA also present a small raise in accuracy, with modification in all metrics. In conclusion, while the SVR-WOA demonstrates its effectiveness in the forecasting outline, the SVR-AOA analysis appears to be a reliable approach for determining accurate Gf values (R2train = 0.921, and R2test = 0.9853) and GF values (R2train = 0.9281, and R2test = 0.9236, as supported by the arguments and feasibility of the models.
Distributed generators (DGs), which can be traditional fossil fuel generators or renewable energy sources (RES), must be appropriately planned in order to reduce a power network's overall generating cost. Renewabl...
详细信息
Distributed generators (DGs), which can be traditional fossil fuel generators or renewable energy sources (RES), must be appropriately planned in order to reduce a power network's overall generating cost. Renewable energy sources (RES) should be prioritized because they provide a clean and sustainable energy supply and are abundant in nature. Demand side management (DSM) optimizes the scheduling of flexible loads to reduce peak demand and improve the load factor, while keeping daily demand unchanged. The test system in this research employs a dependable and effective hybrid optimization tool to plan the DGs of a dynamic system in a way that matches low active power production costs with low pollutant emissions. The fitness functions used in the test system were non-linear due to the presence of the valve point effect (VPE). The costs and emissions were evaluated for various fitness functions which included involvement of wind, DSM, and different types of combined economic emission dispatch (CEED) methods. The test system's peak demand was cut by 12% and the load factor was raised from 0.7528 to 0.85 when DSM technique was used. The generation cost has been reduced from $1,014,996 to $1,012,182 using CSAJAYA algorithm which was further reduced to $1,007,441 after incorporating DSM. Likewise, the CEEDppf was also observed to be reduced to $1,231,435 and $1,216,885 with and without DSM compared to $1,232,001 from reported literature. Numerical results show that both the cost and emission were reduced significantly using the proposed CSAJAYA compared to a long-sighted list of algorithms published in literature.
The notion of fitness landscape (FL) has shown promise in terms of optimization. In this paper we propose a machine learning (ML) prediction approach to quantify FL ruggedness by computing the entropy. The approach ai...
详细信息
The notion of fitness landscape (FL) has shown promise in terms of optimization. In this paper we propose a machine learning (ML) prediction approach to quantify FL ruggedness by computing the entropy. The approach aims to build a model that could reveal information about the ruggedness of unseen instances. Its contribution is attractive in many cases like black-box optimization and in case we can rely on the information of small instances to discover the features of larger and timeconsuming ones. The experiment consists in evaluating multiple ML models for the prediction of the ruggedness of the traveling salesman problem (TSP). The results show that ML can provide, for instances of a similar problem, acceptable predictions and that it can help to estimate ruggedness of large instances in that case. However, the inclusion of several features is necessary to have a more predictable landscape, especially when dealing with different TSP instances.
This letter presents an enhanced Trust Region Method (TRM) for Sequential Linear Programming (SLP) designed to improve the initial feasible solution to a constrained nonlinear programming problem while maintaining the...
详细信息
This letter presents an enhanced Trust Region Method (TRM) for Sequential Linear Programming (SLP) designed to improve the initial feasible solution to a constrained nonlinear programming problem while maintaining the interim solutions feasibility throughout the SLP iterations. The method employs a polytopic sub-approximation of the feasible region, defined around the interim solution as a level set based on variable limits for the linearization error. This polytopic feasible region is established by using a trust region that ensures that maximum limits of the linearization errors are respected. The method adaptively adjusts the size of the feasible region during iterations to achieve convergence to a local optimum by employing variable linearization error limits. Local convergence is attained by reducing the size of the trust radius. A case study illustrates the effectiveness of the proposed method, which is compared to the benchmark TRM that uses heuristic limits on the permissible changes in manipulated variables.
One of the most challenging aspects of nonsmooth analysis is to overcome non-differentiability. A possible approach is to use the generalized notions of the classical gradient and directional derivatives. In this lett...
详细信息
One of the most challenging aspects of nonsmooth analysis is to overcome non-differentiability. A possible approach is to use the generalized notions of the classical gradient and directional derivatives. In this letter we define a generalized directional derivative, the Mandalay derivative, based on set-valued Lie derivatives. For this operator, we derive the analogues to the classical chain rule, superposition rule (for linear combinations of functions), product rule, and quotient rule in the form of inequalities, which facilitate the computation of the Mandalay derivative in the context of nonsmooth system analysis and design. Moreover, we demonstrate the application of the Mandalay derivative for both first and high-order nonsmooth Control Barrier Functions in multiple examples.
暂无评论