This paper proposes a two-timescale compressed primal-dual (TiCoPD) algorithm for decentralized optimization with improved communication efficiency over prior works on primal-dual decentralized optimization. The algor...
详细信息
ISBN:
(数字)9798350368741
ISBN:
(纸本)9798350368758
This paper proposes a two-timescale compressed primal-dual (TiCoPD) algorithm for decentralized optimization with improved communication efficiency over prior works on primal-dual decentralized optimization. The algorithm is built upon the primal-dual optimization framework and utilizes a majorization-minimization procedure. The latter naturally suggests the agents to share a compressed difference term during the iteration. Furthermore, the TiCoPD algorithm incorporates a fast timescale mirror sequence for agent consensus on nonlinearly compressed terms, together with a slow timescale primal-dual recursion for optimizing the objective function. We show that the TiCoPD algorithm converges with a constant step size. It also finds an $\mathcal{O}(1/T)$ stationary solution after T iterations. Numerical experiments on decentralized training of a neural network validate the efficacy of TiCoPD algorithm.
In recent years, the increasing use of credit cards has led to a dramatic rise in fraudulent transactions. Identifying credit card fraud is crucial for preventing financial institutions and their customers from incurr...
详细信息
ISBN:
(数字)9798331508913
ISBN:
(纸本)9798331508920
In recent years, the increasing use of credit cards has led to a dramatic rise in fraudulent transactions. Identifying credit card fraud is crucial for preventing financial institutions and their customers from incurring substantial financial losses and for maintaining customer trust. Fraud detection systems play a key role in preventing unauthorized transactions. Auto-encoder networks, known for their ability to learn compact representations of data, can leverage this capability to identify anomalous transactions, which are typically indicative of fraud. However, their performance is often sensitive to hyperparameter choices, especially in imbalanced datasets. To address this, a genetic algorithm is utilized to dynamically optimize the hyperparameters of these networks, significantly improving fraud detection. This paper proposes using an optimized auto-encoder network, augmented with the genetic algorithm, to detect credit card fraud. On the Credit Card Fraud (CCF) dataset, the model achieved an F-score of 91.04 % and a Matthews correlation coefficient (MCC) of 0.8282, surpassing some baseline methods such as OCSVM, OCNN and COPOD. These results demonstrate the effectiveness of the genetic algorithm in improving the performance and robustness of fraud detection systems.
Aiming at the inefficiency of traditional protection systems in distribution network line faults, a distribution network line current automatic differential protection method based on DTW algorithm and phase character...
详细信息
ISBN:
(数字)9798331518806
ISBN:
(纸本)9798331518813
Aiming at the inefficiency of traditional protection systems in distribution network line faults, a distribution network line current automatic differential protection method based on DTW algorithm and phase characteristics is proposed. Firstly, a device framework integrating data acquisition, digital processing, and protection algorithms was designed to achieve precise processing and discrimination of fault currents and voltages. In terms of fault detection mechanism, differential protection criterion is introduced, which extracts current amplitude and phase characteristics, and uses line fault direction diagram to accurately locate the fault location and evaluate its severity. Innovatively applying the DTW algorithm allows for adaptive adjustment of the timeline length based on data features, thereby optimizing the similarity between data collection points and significantly improving the accuracy of fault feature recognition. By using the DTW distance as the objective function, this method can accurately calculate the position of short-circuit points, grounding resistance value, and fault phase current amplitude, effectively improving the reliability and efficiency of distribution network line protection. The simulation results have verified the improved protection effect of this method in terms of amplitude and phase difference, demonstrating the characteristics of rapid response and efficient action. Even in the case of frequent faults, its efficient performance can still be maintained.
We show by counterexample that the binary integer quadratic programming model proposed by Kaabi and Harrath [2019. "Scheduling on Uniform Parallel Machines with Periodic Unavailability Constraints."Internati...
详细信息
We show by counterexample that the binary integer quadratic programming model proposed by Kaabi and Harrath [2019. "Scheduling on Uniform Parallel Machines with Periodic Unavailability Constraints."International Journal of Production Research57: 216-227] is incorrect. We fix this model and propose two linear ones.
This paper presents a linear programming approach for tracking fundamental frequencies in acoustic signal that contains multiple speech sources and noise interference. A sparsity-based pitch estimation method is used ...
详细信息
This paper presents a linear programming approach for tracking fundamental frequencies in acoustic signal that contains multiple speech sources and noise interference. A sparsity-based pitch estimation method is used to obtain pitch candidates for each signal frame. With conventional methods like exhaustive searching, the computational complexity of multipitch tracking grows exponentially with the number of pitch tracks. We propose to use a linear programming relaxation approach to solve the multiple pitch track searching problem. This approach has low computational complexity while it was found to attain global optimal solution with high probability. Experimental results show that the proposed algorithm is more efficient and more accurate than the conventional tracking method, extended dynamic programming.
In this paper, we propose a fast algorithm for a multi-robot system which enables the robots to stably maneuver on domes. The multi-robot system includes robots, a leader robot and one or more follower robots, which a...
详细信息
In this paper, we propose a fast algorithm for a multi-robot system which enables the robots to stably maneuver on domes. The multi-robot system includes robots, a leader robot and one or more follower robots, which are connected to each other by strings and make a ring around the dome. The proposed algorithm employs multi-rapidly random trees (multi-RRTs) and linear programming (LP) to plan a stable path for the whole system which enables the leader robot to cover the whole dome. The main contribution of the paper is in reducing the high dimensionality of the multi-robot configuration space using linear programming. This would allow the multi-RRT grow faster to cover the whole configuration space. To show the effectiveness of the approach we simulated the behavior of a 4-robot system on a spherical dome which denotes the efficiency of our algorithm.
Detailed placement is a crucial stage in VLSI design that starts from the global placement result to determine the final legal locations of each cell through fine-grained optimization. Traditional detailed placement m...
详细信息
ISBN:
(数字)9783982674100
ISBN:
(纸本)9798331534646
Detailed placement is a crucial stage in VLSI design that starts from the global placement result to determine the final legal locations of each cell through fine-grained optimization. Traditional detailed placement methods focus on minimizing the half-perimeter wire length (HPWL) as in global placement. However, incorporating timing-driven placement becomes essential with the increasing complexity of VLSI designs and tighter performance constraints. In this paper, we propose a timing-driven detailed placement framework that leverages unsupervised graph learning techniques. Specifically, we integrate timing-related metrics into the objective function for detailed placement and formulate it into the loss function of a graph neural network (GNN) model. The loss function includes overlap, legality, and timing-related arc lengths, with appropriate weights using Bayesian optimization. Experimental results show that our framework achieves comparable or improved HPWL while significantly reducing total negative slack (TNS) by 5.5%, compared to existing methods.
Optimization is a computational process to find the best solution by minimizing or maximizing an objective function. This research compares the performance of three optimization algorithms Genetic Algorithm (GA), Part...
详细信息
ISBN:
(数字)9798331508616
ISBN:
(纸本)9798331508623
Optimization is a computational process to find the best solution by minimizing or maximizing an objective function. This research compares the performance of three optimization algorithms Genetic Algorithm (GA), Particle Swarm Optimization (PSO), and Simulated Annealing (SA) in determining the shortest path optimization between coordinate points extracted from detected facial edge images. Using the Canny edge detection method, 10 single facial images with a resolution of 120 x 120 pixels were processed to produce (X, Y) coordinates as input for the optimization process. The results show that GA consistently outperforms the other methods, achieving the shortest total distance in 9 out of 10 datasets. The results of the first image data comparison, the GA method produced the shortest total distance with a value of 18,199.59 pixels, while (SA) 21,072.28 pixels had a performance 15.78% worse than GA, and (PSO) 28,381.58 (pixels) showed a performance 55.94% worse than GA. SA showed competitive and optimal results in one dataset, while PSO consistently provided the longest distance, indicating its lower efficiency in this research.
This paper presents an optimization method to the demand side Energy Management System (EMS) of a given consumer (*** industrial compound or university campus) with respect to hourly electricity prices. This paper con...
详细信息
This paper presents an optimization method to the demand side Energy Management System (EMS) of a given consumer (*** industrial compound or university campus) with respect to hourly electricity prices. This paper considers a cluster of interconnected price responsive demands in an Academic Campus. The demands can be supplied through the main grid and stochastic Distributed Energy Resources (DERs), such as wind and solar power sources. In addition, the cluster of demands owns an energy storage facility. The proposed EMS has ability that each consumer can employ their own strategy to regulate the present load and prices in the power distribution system. To solve this EMS problem and optimization algorithm based on linear programming (LP) approach has been implemented. In addition to LP algorithm an Artificial Neural Network was applied to predict the future power consumption of the cluster of price responsive demands. The objective of the proposed method is to maximize the utilization of the cluster of demands when it is subjected to a set of constrains. This LP algorithm allows the cluster of demand to buy, store and sell energy at suitable times to adjust the hourly load level. To evaluate the performance of the proposed algorithm an IEEE 14 bus system was considered. The results shows that the cluster of demands of energy management system using the proposed approach increasing the efficiency and minimizing the losses than the existing methods.
Misclassification minimization is an important and interesting topic in classification problem. Obviously, exploring the solution for this topic will benefit to many real life problems, such as credit card clients cla...
详细信息
Misclassification minimization is an important and interesting topic in classification problem. Obviously, exploring the solution for this topic will benefit to many real life problems, such as credit card clients classification. This paper focuses on misclassification minimization based on multiple criteria linear programming (MCLP), proposing two different schemes to minimize the number of misclassified points in original MCLP. Especially, the complementarity is used to construct the first scheme and linear approximation technique is applied to solve it. Furthermore, successive linearization algorithm (SLA) is employed to achieve minimization the second scheme. Finally, numerical experiment tests the effect of this idea.
暂无评论