Representation

 

• [Genesereth and Nilsson1987]\Logical Foundations of Artificial Intelligence, Chapters 1, 2.
  Chapter 1 is more motivational than anything else, although Section 1.1 does provide a nice summary of some important historical information. Take note of the Physical Symbol Systems Hypothesis on page 5. Section 2 presents the basics of predicate logic, as well as discussing other approaches to declarative representations.
• [Rich and Knight1991] Artificial Intelligence, Chapter 4,9,10.
  Summaries of Knowledge representation, conceptual dependency, semantic nets, frames, and scripts.
• [Newell1993] Reflections on the Knowledge Level.
  Understand the important concept of the knowledge level. Newell presented the idea as the 1980 AAAI presidential address. It appeared later in AI magazine and in AIJ ([Newell1981]). You may wish to skim the original article, or read the excellent summary of it in the first few pages of [Dietterich1990a].
• [Levesque1986]\ Knowledge representation and reasoning.
  An overview of the Knowledge Representation area and its interactions with work on Reasoning.
• [Forbus1996]\(*) Qualitative Reasoning.
  A good (and up to date) survay paper on qualitative Reasoning (also known as Qualitative Physics). It does lack some description of work done before the 80's on formalizing physics, and the motivation for QR is scattered all over, instead of being summerized. To compensate, read the very good summary by Lise Getoor (available on the Quail96 web-page), of [Forbus1988].
• [Halpern1995] (*) Reasoning About Knowledge: A Survey, in the Handbook of logic in AI and logic programming, pp. 1-34. A good survey of the work done on formalization of Knowledge in the past 5 years. Note that it does not deal with Belief.
• [Laird et al. 1987] Soar: an architecture for general intelligence.
  The important ideas in SOAR are problem solving as search (in the CMU tradition), universal subgoaling, and chunking. Section 2 contains a detailed description of the system -- feel free to skip sections that seem too dense. The sections on chunking (2.6 and 3.3) are important for learning.

• [Genesereth and Nilsson1987] (*) Logical Foundations of Artificial Intelligence.
  Chapter 9 (except sections 9.3, 9.6, 9.11-9.13).
  An introduction to some of the issues in formalizing knowledge and belief. Ignore the details of proofs. (* - I think it should be included although we have a replacement, since it gives some nice considerations on belief vs knowledge. It might be deferred for someone whose depth area is not close to representation/reasoning.)
• [Shoham and Goyal1988] Temporal reasoning.
  A survey of the various formalisms for temporal reasoning. A nice summary of problems with change based approaches (such as the situational calculus) is provided in the first section, as well as concise definitions of the frame, qualification and ramification problems. Skip Sections 3.2 and 3.3.
• [Brachman et al. 1985] KRYPTON: a functional approach to knowledge representation.
  The first part of this paper outlines some of the deficiencies of frame-based representations. The second part describes KRYPTON, a system that combines a frame-based language for describing terms with a logical language for making assertions.
• [Winograd1985] Frame representations and the declarative / procedural controversy.
  Read Sections 1-3 for the main issues in this controversy.
• [Marr1982] Vision, first chapter.
  Read for an understanding of Marr's three levels of description. This is not about vision but about representation.
• [] (*) CYC - TBD.