公开(公告)号：US06654743B1

公开(公告)日：2003-11-25

申请号：US09711479

申请日：2000-11-13

申请人： Tad H. Hogg ,  Rajan M. Lukose

发明人： Tad H. Hogg ,  Rajan M. Lukose

IPC分类号： G06F1730

CPC分类号： G06F17/3071 ,  Y10S707/99937

摘要： A method and apparatus for clustering documents from a set of documents includes issuing a request from an electronic device for documents which are relevant to the request. The documents believed to meet the request requirements are identified and the documents are thereafter supplied to a document clustering system. The document clustering system forms clusters of the documents by operation of the clustering system. A robustness criterion, by analogy with self-assembled physical systems, is applied to select the best, most robust, clustering. Robust clusters have a higher designability feature than other clusters which are generated where designability refers to a number of ways a particular cluster can be formed.

摘要翻译： 用于从一组文档聚类文档的方法和装置包括从电子设备发出与请求相关的文档的请求。 确认满足请求要求的文件，然后将文档提供给文档集群系统。 文档聚类系统通过聚类系统的操作形成文档的集群。 应用自组装物理系统类比的鲁棒性标准来选择最佳，最强大的聚类。 强大的集群具有比其他集群更高的可设计性特征，这些集群在可设计性是指可以形成特定集群的多种方式时产生的。

公开(公告)号：US6027112A

公开(公告)日：2000-02-22

申请号：US33389

申请日：1998-03-02

申请人： Oliver Guenther ,  Tad H. Hogg ,  Bernardo A. Huberman

发明人： Oliver Guenther ,  Tad H. Hogg ,  Bernardo A. Huberman

IPC分类号： B65G51/03 ,  B65H5/22 ,  B65H7/20 ,  B65H29/24 ,  G01K15/00

CPC分类号： B65H5/228 ,  B65G51/03 ,  B65H7/20 ,  B65H2511/20 ,  B65H2515/212 ,  B65H2557/24 ,  B65H2557/33

摘要： A multi-agent control system controls a transport assembly for moving objects. The transport assembly is formed using sensors and actuators that are proximately coupled in physical space. The multi-agent control system includes a learning mechanism which takes advantage of the proximate coupling between the sensors and actuators. The learning mechanism improves system performance by making iterative changes to an interaction matrix that represents the organizational structure of the multi-agent control system. In operation, the learning mechanism makes iterative changes to several of the elements a.sub.ij, of the interaction matrix at one time, around a randomly chosen location (i,j) in the matrix. Changes to the interaction matrix continue to be made so long as the changes result in improved performance of the transport assembly. Advantageously, the learning mechanism enables the multi-agent control system to control the transport assembly without requiring knowledge of specific operating characteristics of the transport assembly.

摘要翻译： 多代理控制系统控制用于移动物体的运输组件。 运输组件使用近似耦合在物理空间中的传感器和致动器形成。 多代理控制系统包括利用传感器和致动器之间的紧密耦合的学习机构。 学习机制通过对表示多代理控制系统的组织结构的交互矩阵进行迭代更改来提高系统性能。 在操作中，学习机制在矩阵中围绕随机选择的位置（i，j），一次对交互矩阵的几个元素aij进行迭代改变。 只要更改导致传输组件的性能得到改进，继续改变交互矩阵。 有利地，学习机构使得多代理控制系统能够控制传输组件，而不需要知道传输组件的特定操作特性。