检索结果-内蒙古大学图书馆

RAINBOW PO: A UNIFIED FRAMEWORK FOR COMBINING IMPROVEMENTS IN PREFERENCE optimization

学校读者我要写书评

暂无评论

arXiv 2024年

作者： Zhao, Hanyang Winata, Genta Indra Das, Anirban Zhang, Shi-Xiong Yao, David D. Tang, Wenpin Sahu, Sambit Columbia University United States Capital One United States

Recently, numerous preference optimization algorithms have been introduced as extensions to the Direct Preference optimization (DPO) family. While these methods have successfully aligned models with human preferences, there is a lack of understanding regarding the contributions of their additional components. Moreover, fair and consistent comparisons are scarce, making it difficult to discern which components genuinely enhance downstream performance. In this work, we propose RAINBOWPO, a unified framework that demystifies the effectiveness of existing DPO methods by categorizing their key components into seven broad directions. We integrate these components into a single cohesive objective, enhancing the performance of each individual element. Through extensive experiments, we demonstrate that RAINBOWPO outperforms existing DPO variants. Additionally, we provide insights to guide researchers in developing new DPO methods and assist practitioners in their implementations. © 2024, CC BY.

关键词： optimization algorithms

GENERALIZED ORDERED WEIGHTED AGGREGATION ROBUSTNESS TO SOLVE UNCERTAIN SINGLE OBJECTIVE optimization PROBLEMS

学校读者我要写书评

暂无评论

arXiv 2024年

作者： Kishor, Nand Ghosh, Debdas Zhao, Xiaopeng Varanasi221005 India School of Mathematical Sciences Tiangong University Tianjin300387 China

Robust optimization aims to find optimum points from the collection of points that are feasible for every possible scenario of a given uncertain set. An optimum solution to a robust optimization problem is commonly found by the min-max robust counterpart or by the best out of the worst-cases analysis. In this article, we introduce a new counterpart with the help of the generalized ordered weighted aggregation (GOWA) operator to solve uncertain single objective optimization problems. After introducing GOWA robustness, we analyze a few elementary properties of the GOWA robust objective function, like continuity, monotonicity, coerciveness, local Lipschitz property, and subdifferential regularity. An approach to computing the Clarke subdifferential of the GOWA robust objective function is also provided. We discuss the relationship between the concept of GOWA robustness with other existing robustness-flimsily, highly, min-max, light, and min-min robustness. We show that in a particular case, GOWA robustness reduces to the commonly used min-max robustness. The entire paper is supported by several geometrical and numerical illustrations. © 2024, CC BY.

关键词： optimization algorithms

Task Allocation Methods and optimization Techniques in Edge Computing: A Systematic Review of the Literature

学校读者我要写书评

暂无评论

FUTURE INTERNET 2023年第8期15卷 254-254页

作者： Patsias, Vasilios Amanatidis, Petros Karampatzakis, Dimitris Lagkas, Thomas Michalakopoulou, Kalliopi Nikitas, Alexandros Int Hellen Univ Dept Comp Sci Kavala 65404 Greece Univ Huddersfield Huddersfield Business Sch Dept Logist Mkt Hospitality & Analyt Huddersfield HD1 3DH England

Task allocation in edge computing refers to the process of distributing tasks among the various nodes in an edge computing network. The main challenges in task allocation include determining the optimal location for each task based on the requirements such as processing power, storage, and network bandwidth, and adapting to the dynamic nature of the network. Different approaches for task allocation include centralized, decentralized, hybrid, and machine learning algorithms. Each approach has its strengths and weaknesses and the choice of approach will depend on the specific requirements of the application. In more detail, the selection of the most optimal task allocation methods depends on the edge computing architecture and configuration type, like mobile edge computing (MEC), cloud-edge, fog computing, peer-to-peer edge computing, etc. Thus, task allocation in edge computing is a complex, diverse, and challenging problem that requires a balance of trade-offs between multiple conflicting objectives such as energy efficiency, data privacy, security, latency, and quality of service (QoS). Recently, an increased number of research studies have emerged regarding the performance evaluation and optimization of task allocation on edge devices. While several survey articles have described the current state-of-the-art task allocation methods, this work focuses on comparing and contrasting different task allocation methods, optimization algorithms, as well as the network types that are most frequently used in edge computing systems.

关键词： task offloading edge computing task allocation optimization algorithms

An accelerate Prediction Strategy for Dynamic Multi-Objective optimization

学校读者我要写书评

暂无评论

arXiv 2024年

作者： Lei, Ru Li, Lin Stolkin, Rustam Feng, Bin Key Laboratory of Information Fusion Technology of Ministry of Education Northwestern Polytechnical University Shaanxi Xi’an710072 China Extreme Robotics Laboratory University of Birmingham BirminghamB15 2TT United Kingdom

This paper addresses the challenge of dynamic multi-objective optimization problems (DMOPs) by introducing novel approaches for accelerating prediction strategies within the evolutionary algorithm framework. Since the objectives of DMOPs evolve over time, both the Pareto optimal set (PS) and the Pareto optimal front (PF) are dynamic. To effectively track the changes in the PS and PF in both decision and objective spaces, we propose an adaptive prediction strategy that incorporates second-order derivatives to predict and adjust the algorithms search behavior. This strategy enhances the algorithm’s ability to anticipate changes in the environment, allowing for more efficient population re-initialization. We evaluate the performance of the proposed method against four state-of-the-art algorithms using standard DMOPs benchmark problems. Experimental results demonstrate that the proposed approach significantly outperforms the other algorithms across most test problems. Copyright © 2024, The Authors. All rights reserved.

关键词： optimization algorithms

Enhancing Land Cover/Land Use (LCLU) classification through a comparative analysis of hyperparameters optimization approaches for deep neural network (DNN)

学校读者我要写书评

暂无评论

ECOLOGICAL INFORMATICS 2023年 78卷

作者： Azedou, Ali Amine, Aouatif Kisekka, Isaya Lahssini, Said Bouziani, Youness Moukrim, Said Ibn Tofail Univ Natl Sch Appl Sci BP 241 Kenitra Morocco Univ Calif Davis Land Air & Water Resources Dept Davis CA 95616 USA Univ Calif Davis Biol & Agr Engn Dept Davis CA 95616 USA Natl Sch Forestry Engn Sale 11000 Morocco Ibn Tofail Univ Fac Sci Kenitra Morocco Mohammed V Univ Rabat Fac Sci Res Ctr Plant & Microbial Biotechnol Biodivers & E Rabat Morocco

Sustainable natural resources management relies on effective and timely assessment of conservation and land management practices. Using satellite imagery for Earth observation has become essential for monitoring land cover/land use (LCLU) changes and identifying critical areas for conserving biodiversity. Remote Sensing (RS) datasets are often quite large and require tremendous computing power to process. The emergence of cloud based computing techniques presents a powerful avenue to overcome computing limitations by allowing machine-learning algorithms to process and analyze large RS datasets on the cloud. Our study aimed to classify LCLU for the Talassemtane National Park (TNP) using a Deep Neural Network (DNN) model incorporating five spectral indices to differentiate six land use classes using Sentinel-2 satellite imagery. optimization of the DNN model was conducted using a comparative analysis of three optimization algorithms: Random Search, Hyper band, and Bayesian optimization. Results indicated that the spectral indices improved classification between classes with similar reflectance. The Hyperband method had the best performance, improving the classification accuracy by 12.5% and achieving an overall accuracy of 94.5% with a kappa coefficient of 93.4%. The dropout regularization method prevented overfitting and mitigated over-activation of hidden nodes. Our initial results show that machine learning (ML) applications can be effective tools for improving natural resources management.

关键词： Deep learning optimization algorithms Image processing Remote sensing Land-use and land-cover classification Google Earth Engine

Symmetric, optimization-based, cross-element compatible nodal distributions for high-order finite elements

学校读者我要写书评

暂无评论

arXiv 2024年

作者： Kaufmann, Julian M. Zahr, Matthew J. Department of Aerospace and Mechanical Engineering University of Notre Dame Notre DameIN46556 United States

We present a general framework to construct symmetric, well-conditioned, cross-element compatible nodal distributions that can be used for high-order and high-dimensional finite elements. Starting from the inherent symmetries of an element geometry, we construct node groups in a systematic and efficient manner utilizing the natural coordinates of each element, while ensuring nodes stay within the elements. Proper constraints on the symmetry group lead to nodal distributions that ensure cross-element compatibility (i.e., nodes of adjacent elements are co-located) on both homogeneous and mixed meshes. The final nodal distribution is defined as a minimizer of an optimization problem over symmetry group parameters with linear constraints that ensure nodes remain with an element and enforce other properties (e.g., cross-element compatibility). We demonstrate the merit of this framework by comparing the proposed optimization-based nodal distributions with other popular distributions available in the literature, and its robustness by generating optimized nodal distributions for otherwise difficult elements (such as simplex and pyramid elements). Copyright © 2024, The Authors. All rights reserved.

关键词： optimization algorithms

Theoretically Grounded Pruning of Large Ground Sets for Constrained, Discrete optimization

学校读者我要写书评

暂无评论

arXiv 2024年

作者： Nath, Ankur Kuhnle, Alan Department of Computer Science and Engineering Texas A&M University United States

Modern instances of combinatorial optimization problems often exhibit billion-scale ground sets, which have many uninformative or redundant elements. In this work, we develop light-weight pruning algorithms to quickly discard elements that are unlikely to be part of an optimal solution. Under mild assumptions on the instance, we prove theoretical guarantees on the fraction of the optimal value retained and the size of the resulting pruned ground set. Through extensive experiments on real-world datasets for various applications, we demonstrate that our algorithm, QuickPrune, efficiently prunes over 90% of the ground set and outperforms state-of-the-art classical and machine learning heuristics for pruning. © 2024, CC BY.

关键词： optimization algorithms

A new Simheuristics procedure for stochastic combinatorial optimization

学校读者我要写书评

暂无评论

arXiv 2024年

作者： Berkhout, Joost Department of Mathematics Vrije Universiteit Amsterdam De Boelelaan 1105 Amsterdam1081 HV Netherlands

Ignoring uncertainty in combinatorial optimization leads to suboptimal decisions in practice. Nevertheless, the focus is often on deterministic combinatorial optimization problems, mainly because they are already challenging enough without stochasticity. To make it easier to address stochasticity in combinatorial optimization, Simheuristics have been developed that allow solving stochastic combinatorial optimization problems. We propose a new Simheuristic procedure that dynamically changes the optimization focus between a deterministic and stochastic perspective based upon a statistical model. By doing so, an adequate trade-off is made between exploration and exploitation of the solution space during the optimization. We numerically show that the new Simheuristic procedure solves real-life stochastic scheduling problems more efficiently than standard Simheuristics strategies. © 2024, CC BY.

关键词： optimization algorithms

Evolutionary Pre-Prompt optimization for Mathematical Reasoning

学校读者我要写书评

暂无评论

arXiv 2024年

作者： Videau, Mathurin Leite, Alessandro Schoenauer, Marc Teytaud, Olivier Meta AI France TAU INRIA LISN CNRS Univ. Paris-Saclay France

Recent advancements have highlighted that large language models (LLMs), when given a small set of task-specific examples, demonstrate remarkable proficiency, a capability that extends to complex reasoning tasks. In particular, the combination of few-shot learning with the chain-of-thought (CoT) approach has been pivotal in steering models towards more logically consistent conclusions [Wei et al. 2022b]. This paper explores the optimization of example selection for designing effective CoT pre-prompts and shows that the choice of the optimization algorithm, typically in favor of comparison-based methods such as evolutionary computation, significantly enhances efficacy and feasibility. Specifically, thanks to a limited exploitative and overfitted optimization, Evolutionary Pre-Prompt optimization (EPPO) brings an improvement over the naive few-shot approach exceeding 10 absolute points in exact match scores on benchmark datasets such as GSM8k and MathQA. These gains are consistent across various contexts and are further amplified when integrated with self-consistency (SC). © 2024, CC BY.

关键词： optimization algorithms