检索结果-内蒙古大学图书馆

Proceedings of the 38th International Conference on Neural Information Processing systems

作者： Xuanjia Zhao Jian Guan Congyi Fan Dongli Xu Youtian Lin Haiwei Pan Pengming Feng College of Computer Science and Technology Harbin Engineering University Independent Researcher School of Intelligence Science and Technology Nanjing University State Key Laboratory of Space-Ground Integrated Information Technology

ISBN: (纸本)9798331314385

Drag-based image editing using generative models provides precise control over image contents, enabling users to manipulate anything in an image with a few clicks. However, prevailing methods typically adopt n-step iterations for latent semantic optimization to achieve drag-based image editing, which is time-consuming and limits practical applications. In this paper, we introduce a novel one-step drag-based image editing method, i.e., FastDrag, to accelerate the editing process. Central to our approach is a latent warpage function (LWF), which simulates the behavior of a stretched material to adjust the location of individual pixels within the latent space. This innovation achieves one-step latent semantic optimization and hence significantly promotes editing speeds. Meanwhile, null regions emerging after applying LWF are addressed by our proposed bilateral nearest neighbor interpolation (BNNI) strategy. This strategy interpolates these regions using similar features from neighboring areas, thus enhancing semantic integrity. Additionally, a consistency-preserving strategy is introduced to maintain the consistency between the edited and original images by adopting semantic information from the original image, saved as key and value pairs in self-attention module during diffusion inversion, to guide the diffusion sampling. Our FastDrag is validated on the DragBench dataset, demonstrating substantial improvements in processing time over existing methods, while achieving enhanced editing performance. Project page: https://***/.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Evaluating Performance Portability of SYCL and Kokkos: A Case Study on LBM Simulations

Evaluating Performance Portability of SYCL and Kokkos: A Cas...

引用

IEEE International Conference on Big Data and Cloud Computing (BdCloud)

作者： Yue Ding Chuanfu Xu Haozhong Qiu Qingsong Wang Weixi Dai Yongzhen Lin Yonggang Che Institute for Quantum Information & State Key Laboratory of High Performance Computing National University of Defense Technology Changsha China

Since modern high performance computing systems are evolving towards diverse and heterogeneous architectures, the emergence of high-level portable programming models leads to a particular focus on performance portability. In this paper, we evaluate the performance portability and explore performance optimization methods for two portable programming models SYCL and Kokkos. We take an open-source multi-phase Lattice Boltzmann Method (LBM) flow simulation code as a case study and implement portable versions with different optimizations. Then we compare our portable implementations with engineer-tuned OpenMP and CUDA versions on Intel CPUs and NVIDIA GPUs. Experimental results show that both SYCL and Kokkos can deliver superior performance than traditional programming models, but the best performance of the portable versions depends heavily on platform-specific optimizations. There is no single implementation that can achieve the best performance on both CPUs and GPUs. Consequently, we conclude that the performance portability still needs to be further improved for both SYCL and Kokkos. In addition, we present a comparative analysis of different optimization methods that qualify the performance enhancement when using SYCL and Kokkos on CPUs and GPUs. Our work offers valuable references for the development of both portable programming models and applications.

关键词：

来源：评论

学校读者我要写书评

暂无评论

An adaptive cubic regularisation algorithm based on interior point methods for solving nonlinear inequality constrained optimization

arXiv

引用

arXiv 2024年

作者： Pei, Yonggang Guo, Jingyi Zhu, Detong Engineering Laboratory for Big Data Statistical Analysis and Optimal Control College of Mathematics and Information Science Henan Normal University China Mathematics and Science College Shanghai Normal University China

Nonlinear constrained optimization has a wide range of practical applications. In this paper, we consider nonlinear optimization with inequality constraints. The interior point method is considered to be one of the most powerful algorithms for solving large-scale nonlinear inequality constrained optimization. We propose an adaptive regularisation algorithm using cubics based on interior point methods (ARCBIP) for solving nonlinear inequality constrained optimization. For solving the barrier problem, we construct ARC subproblem with linearized constraints and the well-known fraction to the boundary rule that prevents slack variables from approaching their lower bounds prematurely. The ARC subproblem in ARCBIP can avoid incompatibility of the intersection of linearized constraints with trust-region bounds in trust-region methods. We employ composite-step approach and reduced Hessian methods to deal with linearized constraints, where the trial step is decomposed into a normal step and a tangential step. They are obtained by solving two standard ARC subproblems approximately with the fraction to the boundary rule. Requirements on normal steps and tangential steps are given to ensure global convergence. To determine whether the trial step is accepted, we use exact penalty function as the merit function in ARC framework. The updating of the barrier parameter is implemented by adaptive strategies. Global convergence is analyzed under mild assumptions. Preliminary numerical experiments and some comparison results are reported. Copyright © 2024, The Authors. All rights reserved.

关键词： Constrained optimization

来源：评论

学校读者我要写书评

暂无评论

SoK: A Review of Differentially Private Linear Models For High-Dimensional Data

SoK: A Review of Differentially Private Linear Models For Hi...

引用

Secure and Trustworthy Machine Learning (SaTML), IEEE Conference on

作者： Amol Khanna Edward Raff Nathan Inkawhich Booz Allen Hamilton Annapolis Junction MD USA Booz Allen Hamilton University of Maryland Baltimore County Annapolis Junction MD USA Air Force Research Laboratory Rome NY USA

Linear models are ubiquitous in data science, but are particularly prone to overfitting and data memorization in high dimensions. To guarantee the privacy of training data, differential privacy can be used. Many papers have proposed optimization techniques for high-dimensional differentially private linear models, but a systematic comparison between these methods does not exist. We close this gap by providing a comprehensive review of optimization methods for private high-dimensional linear models. Empirical tests on all methods demonstrate robust and coordinate-optimized algorithms perform best, which can inform future research. Code for implementing all methods is released online.

关键词：

来源：评论

学校读者我要写书评

暂无评论

optimization of Heavy Haul Train Speed Profiles Under Discrete Notches and Nonlinear Braking Forces

Optimization of Heavy Haul Train Speed Profiles Under Discre...

引用

International Conference on Intelligent Transportation

作者： Zhiyuan Zhang Hongwei Wang Xi Wang Jun Jin Zhipeng Ying Pengfei Ning School of Automation and Intelligence Beijing Jiaotong University Beijing China China State Railway Group Co. Ltd. Beijing China Signals and Communication Research Institute China Academy of Railway Sciences Beijing China State Key Laboratory of Advanced Rail Autonomous Operation Beijing Jiaotong University Beijing P. R. China Collaborative Innovation Center of Railway Traffic Safety Beijing Jiaotong University Beijing China

ISBN: (数字)9798331505929

ISBN: (纸本)9798331505936

Due to the large mass of heavy haul trains, operating on downhill sections poses significant challenges on train control. Drivers must perform air braking and releasing at the appropriate times to prevent the movement of heavy hual train from decoupling and derailment. To promote the smooth operation of the heavy haul train, this research investigates the optimization problem of heavy haul train speed profiles under discrete notches and nolinear braking forces. An optimization model is first established considering constraints such as speed and air-refilling time to maximize the running distance and minimize the air braking time of a train on downhill sections. This model takes into account discrete notches of electric braking force, each with a nonlinear braking profile. To linearize the model, piecewise approximation (PWA) function and auxiliary variables are introduced, transforming the original problem into a mixed integer linear programming (MILP) problem. The proposed MILP problem is then solved using a solver. Simulations using actual data from the Shuohuang Railway illustrate the effectiveness of the proposed methods.

关键词： Atmospheric modeling Railway accidents optimization models Force Linear programming Rail transportation Data models Mixed integer linear programming optimization Intelligent transportation systems

来源：评论

学校读者我要写书评

暂无评论

Reinforcement Learning-Based Adaptive Search Strategy Combination Algorithm 22

Reinforcement Learning-Based Adaptive Search Strategy Combin...

引用

22nd IEEE International Symposium on Parallel and Distributed Processing with Applications, ISPA 2024

作者： Liu, Xiaotong Zeng, Zhiqiang Wan, Sicheng Tian, Jiajie Wang, Tianlei Hong, Zhiyong Wuyi University School of Mechanical and Automation Engineering Jiangmen China Jiangmen Key Laboratory of Kejie Semiconductor Bonding Technology and Control System Jiangmen China Wuyi University School of Electronic and Information Engineering Jiangmen China Guangdong Kejie Technology Co. Ltd. Jiangmen China

ISBN: (纸本)9798331509712

Integrating multiple search operators to leverage the characteristics of different search operators is one of the common methods to enhance the performance of evolutionary algorithms. Most of these combination algorithms adopt a fixed framework, which makes it difficult to adaptively select the appropriate search operator during the algorithm iteration process, resulting in poor performance on certain problems. To overcome the aforementioned shortcomings of combination algorithms, an adaptive search strategy combination algorithm based on reinforcement learning and neighborhood search (RLASCA) is proposed. During the algorithm iteration process, a reinforcement learning-based adaptive search operator selection method (RLAS) is designed, enabling the algorithm to adaptively select the appropriate search operator based on the state of the individual. Furthermore, to address the issue of premature convergence, a neighborhood search strategy based on differential evolution (NSDE) is designed to increase the diversity of the population. To verify the effectiveness of the proposed algorithm, a comprehensive testing was conducted using the CEC2017 test suite and two constrained engineering design problems. The algorithm was compared with four basic algorithms that compose RLASCA and four classical combination algorithms. The experimental results indicate that RLASCA outperforms the other eight algorithms in terms of convergence speed and accuracy. © 2024 IEEE.

关键词： optimization algorithms

来源：评论

学校读者我要写书评

暂无评论

TEMPORAL ALIGNMENT PREDICTION FOR SUPERVISED REPRESENTATION LEARNING AND FEW-SHOT SEQUENCE CLASSIFICATION 10

TEMPORAL ALIGNMENT PREDICTION FOR SUPERVISED REPRESENTATION ...

引用

10th International Conference on Learning Representations, ICLR 2022

作者： Su, Bing Wen, Ji-Rong Beijing Key Laboratory of Big Data Management and Analysis Methods Gaoling School of Artificial Intelligence Renmin University of China Beijing100872 China

Explainable distances for sequence data depend on temporal alignment to tackle sequences with different lengths and local variances. Most sequence alignment methods infer the optimal alignment by solving an optimization problem under pre-defined feasible alignment constraints, which not only is time-consuming, but also makes end-to-end sequence learning intractable. In this paper, we propose a learnable sequence distance called Temporal Alignment Prediction (TAP). TAP employs a lightweight convolutional neural network to directly predict the optimal alignment between two sequences, so that only straightforward calculations are required and no optimization is involved in inference. TAP can be applied in different distance-based machine learning tasks. For supervised sequence representation learning, we show that TAP trained with various metric learning losses achieves completive performances with much faster inference speed. For few-shot action classification, we apply TAP as the distance measure in the metric learning-based episode-training paradigm. This simple strategy achieves comparable results with state-of-the-art few-shot action recognition methods. © 2022 ICLR 2022 - 10th International Conference on Learning Representationss. All rights reserved.

关键词： Forecasting

来源：评论

学校读者我要写书评

暂无评论

Performance optimization of planar photonic crystal bound states in the continuum cavities: mitigating finite-size effects

引用

Frontiers of Optoelectronics 2025年第1期18卷 53-62页

作者： Ran Hao Bilin Ye Jinhong Xu Yonggang Zou State Key Laboratory on High Power Semiconductor Lasers Changchun University of Science and TechnologyChangchun 130022China College of Optical and Electronic Technology China Jiliang UniversityHangzhou 310018China

Bound states in the continuum (BICs) offer a promising solution to achieving high-quality factor (Q factor) cavities. However, finite-size effects severely deteriorate the BIC mode in practical applications. This paper reports the experimental demonstration of an electrically pumped 940 nm laser based on optimized BIC cavity, achieving a high Q factor of up to 1.18 × 10^(4) even with finite photonic crystal footprint, which is two orders of magnitude larger than un-optimized BIC design. Two strategies have been systematically investigated to mitigate finite-size effects: reflective photonic crystal cavity design and graded photonic crystal cavity design. Both methods significantly improve the Q factor, demonstrating the effectiveness of preserving BIC characteristics in finite-sized photonic crystal cavities. In addition, the reflective boundary photonic crystal design is fabricated and experimentally characterized to demonstrate its lasing characteristics. The fabricated laser exhibits single-mode operation with a signal-to-noise ratio of 38.6 dB. These results pave the way for future designs of BICs with finite size in real applications, promoting the performance of BIC-based integrated lasers.

关键词： Bound states in the continuum High-quality factor Graded photonic crystals Electrical pumped laser

来源：评论

学校读者我要写书评

暂无评论

Multi-Agent Reinforcement Learning for Power Control in Wireless Networks via Adaptive Graphs

arXiv

引用

arXiv 2023年

作者： Amorosa, Lorenzo Mario Skocaj, Marco Verdone, Roberto Gündüz, Deniz Department of Electrical Electronic and Information Engineering University of Bologna Italy Department of Electrical and Electronic Engineering Imperial College London United Kingdom WiLab - National Wireless Communication Laboratory CNIT Italy

The ever-increasing demand for high-quality and heterogeneous wireless communication services has driven extensive research on dynamic optimization strategies in wireless networks. Among several possible approaches, multi-agent deep reinforcement learning (MADRL) has emerged as a promising method to address a wide range of complex optimization problems like power control. However, the seamless application of MADRL to a variety of network optimization problems faces several challenges related to convergence. In this paper, we present the use of graphs as communication-inducing structures among distributed agents as an effective means to mitigate these challenges. Specifically, we harness graph neural networks (GNNs) as neural architectures for policy parameterization to introduce a relational inductive bias in the collective decision-making process. Most importantly, we focus on modeling the dynamic interactions among sets of neighboring agents through the introduction of innovative methods for defining a graph-induced framework for integrated communication and learning. Finally, the superior generalization capabilities of the proposed methodology to larger networks and to networks with different user categories is verified through simulations. © 2023, CC BY-NC-ND.

关键词： Reinforcement learning

来源：评论

学校读者我要写书评

暂无评论

Optimizing Irregular Communication with Neighborhood Collectives and Locality-Aware Parallelism

Optimizing Irregular Communication with Neighborhood Collect...

引用

2023 International Conference on High Performance Computing, Network, Storage, and Analysis, SC Workshops 2023

作者： Collom, Gerald Li, Rui Peng Bienz, Amanda Department of Computer Science University of New Mexico Albuquerque United States Center for Advanced Scientific Computing Lawrence Livermore National Laboratory Livermore United States

ISBN: (纸本)9798400707858

Irregular communication often limits both the performance and scalability of parallel applications. Typically, applications individually implement irregular communication as point-to-point, and any optimizations are integrated directly into the application. As a result, these optimizations lack portability. It is difficult to optimize point-to-point messages within MPI, as the interface for single messages provides no information on the collection of all communication to be performed. However, the persistent neighbor collective API, released in the MPI 4 standard, provides an interface for portable optimizations of irregular communication within MPI libraries. This paper presents methods for implementing existing optimizations for irregular communication within neighborhood collectives, analyzes the impact of replacing point-to-point communication in existing codebases such as Hypre BoomerAMG with neighborhood collectives, and finally shows up to a 1.38x speedup on sparse matrix-vector multiplication communication within a BoomerAMG solve through the use of our optimized neighbor collectives. The authors analyze three implementations of persistent neighborhood collectives for Alltoallv: an unoptimized wrapper of standard point-to-point communication, and two locality-aware aggregating methods. The second locality-aware implementation exposes an non-standard interface to perform additional optimization, and the authors present the additional 0.07x speedup from the extended interface. All optimizations are available in an open-source codebase, MPI Advance, which sits on top of MPI, allowing for optimizations to be added into existing codebases regardless of the system MPI install. © 2023 ACM.

关键词： Application programming interfaces (API)

来源：评论

学校读者我要写书评

暂无评论

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案：

请选择收藏分类：

通借通还

建议与咨询 留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案： 新增检索档案 确定 取消

请选择收藏分类： 新增自定义分类 确定 取消

通借通还

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

请选择保存的检索档案：

请选择收藏分类：