CISE Research Instrumentation: Active Learning for Text, Scene, and Biosequence Analysis

CISE 研究仪器：用于文本、场景和生物序列分析的主动学习

基本信息

批准号：
9617307
负责人：
Garrison Cottrell
金额：
$ 4万
依托单位：
University of California-San Diego
依托单位国家：
美国
项目类别：
Standard Grant
财政年份：
1997
资助国家：
美国
起止时间：
1997-02-01 至 1999-01-31
项目状态：
已结题

来源：
https://www.nsf.gov/awardsearch/showAward?AWD_ID=9617307&HistoricalAwards=false
关键词：
CISE Research Instrumentation Active Learning

项目摘要

9617307 Cottrell, Garrison W. Charles Elkan, University of California, San Diego CISE Research Instrumentation: Active Learning for Text, Scene, and Biosequence Analysis Adaptation to changing environments and changing goals through autonomous learning is a central theme of research in the artificial intelligence laboratory at UCSD. The workstations, robots, and neural network hardware acquired through this grant will enable recently developed learning methods to be tested in four large-scale, real-world applications.-Learning Semantic Representations for Information Retrieval-Embedded Virtual Agents-MEME: A New Software Tool for Sequence Analysis-Real Time Face and Object Recognition in Embedded Agents. The first subproject will apply new algorithms for document rerepresentation and response to human feedback to gigabytes of text in the national TREC competition. Another project will port to the worldwide web software agents that learn and interact now in a simulated artificial life environment. A third project will scale up new software for finding patterns in DNA and protein sequences into a tool capable of autonomously analyzing entire genomes. Finally, a fourth project will validate new active vision methods for recognizing and tracking objects by implementing these methods in mobile robots with controllable cameras.

9617307 Cottrell，Garrison W. Charles Elkan，加州大学圣地亚哥大学CISE研究仪器：对文本，场景和Biosequence分析的主动学习适应对不断变化的环境和通过自主学习的改变是UCSD人工智能实验室研究的中心主题。通过这笔赠款获得的工作站，机器人和神经网络硬件将使最近开发的学习方法能够在四个大规模的，现实世界中的应用程序中进行测试。用于信息检索 - 回件 - 嵌入式虚拟代理的学习语义表示：序列分析的新软件工具，用于序列分析时间和对象识别embedded Agents中。第一个子项目将在国家TREC竞赛中对文档的陈述和对人类对千兆字节的反馈的反应应用新算法。另一个项目将移植到全球Web软件代理商中，该代理商现在在模拟的人造生活环境中学习和互动。第三个项目将扩展新软件以在DNA和蛋白质序列中找到模式，以自主分析整个基因组的工具。最后，第四个项目将通过在具有可控摄像机的移动机器人中实现这些方法来验证新的主动视觉方法，以识别和跟踪对象。