Direct Search Based on Probabilistic Descent

来源 :第八届工业与应用数学国际大会 | 被引量 : 0次 | 上传用户：yoyo88420

【摘要】

：

　　Direct-search methods are a class of derivative-free algorithms based on evaluating the objective function along directions in positive spanning sets.

【作者】

：

Gratton Serge Royer Clement Vicente Luis Nunes Zhang Zaikun

【机构】

：

Univ.of Toulouse,IRIT-CERFACS joint laboratory

【出处】

：

第八届工业与应用数学国际大会

【发表日期】

：

2015年期

下载到本地 , 更方便阅读

下载此文赞助VIP

声明 : 本文档内容版权归属内容提供方 , 如果您对本文有版权争议 , 可与客服联系进行内容授权或下架

论文部分内容阅读

　　Direct-search methods are a class of derivative-free algorithms based on evaluating the objective function along directions in positive spanning sets.

其他文献

Convergence Rate of the Consensus of Linear Multi-agent Systems with Communication Noises

　　This study mainly focuses on the convergence rate of the consensus of linear multi-agent systems.It is assumed that each agent is described by the generic l

会议

Distributed Finite-time Tracking of Multiple Euler-Lagrange Systems without Velocity Measurements

　　This paper investigates the distributed finite-time tracking problem of networked agents with Euler-Lagrange dynamics.To achieve finite-time tracking,with t

会议

Distributed control of networked agent systems

　　Many real large-scale systems can be modeled as networked agent systems,where typical examples include Internet,WWW,smart grids,etc.

会议

Node-to-node Consensus of Multi-agent Systems under Directed Topology

　　Consensus of multi-agent system has recently received considerable attention from various scientific communities due to its potential applications in engine

会议

An Expander Graph Construction and Its Implication for Compressed Sensing

　　We consider a probabilistic construction of sparse matrices with provable restricted isometry constants(RICS)using l1 norms.

会议

PCANet:A Simple Deep Learning Baseline for Image Classification?

　　We will discuss a very simple deep network for image classification which comprises only the very basic data processing components: cascaded principal compo

会议

Rank-One Tensor Updating Algorithms for Tensor Completion

　　We introduce algorithms for tensor completion which are based on rank-one tensor updating.At each iteration,a rank-one tensor is selected by computing the t

会议

Theory and Applications of Conic Integral Geometry

　　Conic integral geometry provides deep insights as well as-exactformulas for the probabilities of important events such as-success of the compressed sensing

会议

Mathematics of Information and Low Dimensional Models

　　This min-symposium considers a variety of ill-posed inverse problems associated with information theory,signal processing,and image processing.

会议

A Subspace Decomposition Framework for Nonlinear Optimization

　　We present a parallel subspace decomposition framework for nonlinear optimization,which can be regarded as an extension of the domain decomposition method f

会议

Direct Search Based on Probabilistic Descent

其他学术论文