首页 > 其他 > 详细

[Knowledge-based AI] {ud409} Lesson 6: 06 - Production Systems

时间:2019-06-06 09:19:34      阅读:143      评论:0      收藏:0      [点我收藏+]

 技术分享图片

 

 

 技术分享图片

 

 

 

技术分享图片

This example is adapted from the following paper:

Lehman, J. F., Laird, J. E., & Rosenbloom, P. S. (1996). A gentle introduction to Soar, an architecture for human cognition. Invitation to Cognitive Science4, 212-249.

技术分享图片 

 David is using his knowledge to make a decision.

  • How is he using his knowledge to make a decision?
  • What is the architecture?
  • what is the reasoning that leads him to make that specific decision?

 

 

 

 Function of a Cognitive Architecture

技术分享图片

Production Systems:
Winston Chapter 7, pages 119-137 can be found at:http://courses.csail.mit.edu/6.034f/ai3/rest.pdf 

 

 

 

 Levels of Cognitive Architectures

 技术分享图片

 

 

 

技术分享图片

 

 

 

 Assumptions of Cognitive Architectures

 技术分享图片

This list is adapted from the following paper:

Lehman, J. F., Laird, J. E., & Rosenbloom, P. S. (1996). A gentle introduction to Soar, an architecture for human cognition. Invitation to Cognitive Science4, 212-249.

 

 

 

 

Architecture + Content = Behavior 

 技术分享图片

技术分享图片

 

 

 

 

 A Cognitive Architecture for Production Systems

技术分享图片

A general architecture of Cognitive system

 

Next is a specific kind of Cognitive system: SOAR

 技术分享图片

This example is adapted from the following paper:

Lehman, J. F., Laird, J. E., & Rosenbloom, P. S. (1996). A gentle introduction to Soar, an architecture for human cognition. Invitation to Cognitive Science4, 212-249.

 

Here we first only focus on the deliberation of SOAR => long-term memory and working memory

Thre are three parts of knowledge in long-term memory:

  • Procedural: how to do certain things
  • Sematic: generalization in the form of concepts and models of the world
  • Episodic: events, eg. what happened yesterday

 

 

 

 Return to the Pitcher

 技术分享图片

 

 

 Action Selection

 技术分享图片

技术分享图片

decision making => find a path from S0 to S101 => havent perform any action, just thinking

 (comment: model predictive control?)

 

 

 Putting Content in the Architecture

 技术分享图片

simple representation of information

 

 

 

 Bringing in Memory

 技术分享图片

 production rules:

 技术分享图片

 

 

 Exercise: Production System in Action I

left: situation; right: knowledge (rules)

 技术分享图片

r2 => end

技术分享图片

r2 => r4 => r5 => end

 技术分享图片

r2 => r4 => r5 & r6 => conflict 

 

 

 

 Chunking

 技术分享图片

r2 => r4 => r5 & r6 => conflict => learn to choice r5 or r6 here

 技术分享图片

here, reasoning, learning and memory are closely connected.

 

 

 

 

 技术分享图片

 

 

 

 

Fundamentals of Learning

 

 reason first, if reach an impasse, turn back to learning

We are trying to build a unified theory of reasoning, memory and learning, where the demands of memory and reasoning constrain the processing of learning 

 

 

 

 技术分享图片

 

 

 

 

技术分享图片

 

[Knowledge-based AI] {ud409} Lesson 6: 06 - Production Systems

原文:https://www.cnblogs.com/ecoflex/p/10982896.html

(0)
(0)
   
举报
评论 一句话评论(0
关于我们 - 联系我们 - 留言反馈 - 联系我们:wmxa8@hotmail.com
© 2014 bubuko.com 版权所有
打开技术之扣,分享程序人生!