@article{mercer1909functions,
title = {Functions of positive and negative type, and their connection with the theory of integral equations},
author = {J. Mercer},
journal = {Philosophical Transactions of the Royal Society, London},
pages = {415--446},
volume = 209,
year = 1909,
biburl = {http://www.bibsonomy.org/bibtex/2e58f01606a71fae2cda3de0a0358d62f/grahl},
keywords = {machine-learning svm imported AI}
}
@book{genesereth87,
title = {Logical foundations of artificial intelligence},
address = {San Francisco},
author = {Michael R. Genesereth and Nils J. Nilsson},
publisher = {Morgan Kaufmann},
year = 1987,
id = {121895},
priority = {1},
comment = {textbook on AI with strong symbolicist approach},
description = {sdasda},
biburl = {http://www.bibsonomy.org/bibtex/2508c1fa46547e1a54be12117cd0f2027/grahl},
keywords = {AI logic}
}
@article{newell82,
title = {The {K}nowledge {L}evel},
author = {Alan Newell},
journal = {Journal of Artificial Intelligence},
volume = 18,
year = 1982,
id = {121818},
priority = {0},
comment = {p. 90 - how to define knowledge, p 91 -- special issue on kr 2 key areas of Ai: representation and learning How do we 'represent' a certain problem in such a way that an intelligent agent might understand it? There is no agreement on KR. Knowledge as a further level of CS on top of programming ? Agents + BDI Logic just doesn't cut it for certain ccass of problems which humans have trouble comprehending principle of rationality - given knowledge that an action A will satisfy goal G, the agent will select that action. Yet knowledge cannot remdily be uisualized =simply too much knowledge in the world -> Chaos theory may be useful does knowing P, and knowing P->Q, lead to knowing Q? Logic is a tool for analysing knowledge, not for reasoning by intelligent agents. p.100 knowledge (under the knowledge-level hypothesis) is a competence-like notion, being a potential for generating action" knowledge is intimately linked with rationality representations are at the symbol level realizing a body of knowledge at the knowledge level knowledge serves as the specification of what a symbol structure should be able to do. Newell takes the functional, non-structural, cybernetic, black-box approach. knowledge is a "state-variable" in this perspective - from http://ksi.cpsc.ucalgary.ca/AIKM97/gaines/KMKL.html p.107 why isn't the collection of symbols S the knowledge in an agent? - short answer is that knowledge of the world cannot be captured in a finite structure. p. 109 solutions to the problem of representing knowledge - i.e what can we do at the symbol level to implement this black box? solutions are ways to say things about the environment, not about the internal structure. logics are good candidates - refined means for saying things about environments, but not the only one if we give the agent a set of logical expressions L, to say that the agent has knowledge K is to say the agent knows all that can be inferred from the conjunction of K p.110 but how can this be true if humans cannot know all the implications of L? knoweldge level is an approximation p. 114 representation = knowledge + access p. 115 is the agent actually intelligent? not once we start to delve into the symbol level p 117 distinction between epistemological adequacy (does there exist an adequate explicit representation of some knowledge) and heuristic adequacy p. 121 logic is a tool to analyze knowledge - to determine what exactly it contains and characterize it p. 122 knowledge is justified true belief in philosophical terms},
description = {sdasda},
biburl = {http://www.bibsonomy.org/bibtex/2cc95d3869d544f9d5eabde1760311e81/grahl},
keywords = {AI}
}
@book{musen88,
title = {Graphical {S}pecification of {P}rocedural {K}nowledge for an {E}xpert {S}ystem},
address = {Norwood, NJ},
author = {M. A. Musen and Lawrence M. Fagan and Edward Shortliffe},
editor = {Ben Shneiderman},
pages = {15--35},
publisher = {Ablex Publishing},
series = {Expert Systems: The User Interface},
year = 1988,
id = {121809},
priority = {0},
comment = {KA is a bottleneck - see Buchanan et al 83 uses a graphical env to do KA of clinical trials protocols makes case for using graphical interfaces for formalized, highly stereotypic, and constrained knowledge},
description = {sdasda},
biburl = {http://www.bibsonomy.org/bibtex/263eefbbd3baa32ebe2e781421ebf4e26/grahl},
keywords = {information visualisation reasoning AI}
}
@book{minsky75,
title = {A framework for representing knowledge},
address = {New York},
author = {Marvin Minsky},
pages = {211--277},
publisher = {McGraw-Hill},
series = {The Psychology of Computer Vision},
year = 1975,
id = {121798},
priority = {0},
comment = {p.211 move from representing it with small fragments -notion of default reasoning ? difference with Gestalt theories p.214 ? focus on how knowledge is structured,psychology is too parsimonious with theories p216 ? but someone must develop these frames for mople to use ? p226 people's vision system are often poorer than they imagine pyramid ?seems to suggest that frames cannot deal with meta-thoughts ?p244- is world regular enough to permit precompiled frames to be useful? ?p251 - to avoid being too specific, frames allow for generalization up the hierarchy -classes should not be seen as inclusion-based. Concepts can have different meanings at different times e.q. Generator is both mechanical and electrical ?p258 - logic stores no intelligence about how a decision was made,what the problem space was. ?p260 - similarity to Kuhn's ideas ?p262 - notion of frames having a satisfaction level or fault tolerance ?p275- a number (derived procedurally ) cannot reflect the considerations that formed it.},
description = {sdasda},
biburl = {http://www.bibsonomy.org/bibtex/209bf0b4c3add8d13a22f9d7494d4bf41/grahl},
keywords = {AI frames ontology}
}
@book{minsky85,
title = {The society of mind},
address = {New York},
author = {Marvin Minsky},
publisher = {Simon \& Schuster},
year = 1985,
id = {121797},
priority = {0},
comment = {-many minds can produce a whole greater than the sum of its parts -but, if we keep subdividing a task, the new question must then be, When do we stop? For this subdivision can be shown to be fractal, can it not? -how do we encode common-sense in a computer? -I don't believe Minsky is making any claim that 'knowledge' implies 'consciousness' or other lofty claims. Mechanistic like francis bacon ?idea of BDI p.75 ?the incredible power of everyday action vs. Genius p.80 ?reinforcing agents which peform well-how to ?is Minsky right with his 'state of mind' theory? ?learning is at least two-pronged: learn from success and/or mistakes ?Piaget thought experiment p.100 ?Minsky seems to approach things very simplistically ?Padert's principle: mental growth comes not just from new skills, but new ways of using existing skills. ?the best definitions bridge between the 'purposeful' and the 'structural' p.131 ?investment principle: oldest ideas have unfair advantages->larger mass of skills p.146 ?consciousness=meta-thought ?memory is a process which makes our brain-agents act in the same way as they did in the past. P.154 ?idea of studying human minds using babies ?what is an emotion? Early eaotions signify need p.172 ? rapid change is dangerous - looks like spurts when adopted - must machines be logical? p. 186 what is logic used for? it is used for evaluating decisions we have already made. Logic as grammar is to sentence structure - can tell us whether our sentences are properly formed but not what new sentences to form. How does this idea fit in with Godot's theorem that in any system there are elements which are undecidable and unprovable - "pointless truths" - "many have assumed that logical necessity lies at the heart of our reasoning" p. 187 - chains of reasoning as ways hymans think - logic is a way of making a chain - one support for every link p. 189 - but common sense reasoning asks whether, what the chain looks like so far is in accord with our experience - once we have created these chains logic helps us pick out the most essential steps -deciding what group of reasons to use: "strength from magnitude" and "strength from multitude" p. 191 - idea of frames p. 243 what happens if our frame of how something looks ends up being invalidated - what is to be done in that case? how flexible are frames? - the idea that cognition is really about being 'reminded' of something - a trigger - slots == terminals p. 245 - how do we mentally recognize scenes that are incomplete - say half a face - its by filling in the 'default' slots},
description = {sdasda},
biburl = {http://www.bibsonomy.org/bibtex/2e927b0ca1977b520a4029990358b93e3/grahl},
keywords = {AI frames logic}
}
@inbook{mcdermott81,
title = {Artificial intelligence meets natural stupidity},
address = {Cambridge, Maine},
author = {Drew Mcdermott},
chapter = 5,
editor = {J. Haugeland},
pages = {143--160},
publisher = {MIT Press},
series = {Mind Design: Philosophy, Psychology, Artifici},
type = {Book},
year = 1981,
id = {121788},
priority = {2},
description = {sdasda},
biburl = {http://www.bibsonomy.org/bibtex/2f4a180ec13e5b660fa1768ee3ecfd528/grahl},
keywords = {AI}
}
@book{lehner88,
title = {Cognitive {I}mpacts of the {U}ser {I}nterface},
address = {Norwood, NJ},
author = {Paul E. Lehner and Mary M. Kralj},
editor = {Ben Shneiderman},
pages = {307--318},
publisher = {Ablex Publishing},
series = {Expert Systems: The User Interface},
year = 1988,
id = {121762},
priority = {0},
comment = {incr complexity demands a better mapping btw user cognitive model and system's reflection of that domain experts assume that the expert system uses a similar problem solving method to their own interface can cause people to assume things are happening which aren't},
description = {sdasda},
biburl = {http://www.bibsonomy.org/bibtex/2145d9b45ed46cd51d9545696dce80fd4/grahl},
keywords = {visualisation AI information modeling cognition reasoning}
}
@article{schreiber94,
title = {Common{KADS}: {A} comprehensive methodology for {KBS} development},
author = {Guus Schreiber and Bob Wielinga and Robert d. de Hoog and Hans Akkermans and Walter v. d. van de Velde},
journal = {IEEE Expert (IEEE Intelligent Systems)},
pages = {28--38},
volume = {December},
year = 1994,
url = {http://www.cs.toronto.edu/~nernst/papers/schreiber-commonkads.pdf},
id = {121616},
priority = {0},
comment = {models the application and the knowledge of the organization previous to this the AI community was more interested in rapid prototyping using LISP and ESS move of KE from acquiring knowledge to modelling it - not just the expert knowledge, but how that knowledge is used in the org. - expertise modelling is what distinguish KADS from other software development - a suite of models: expertise, task, organization, agent, communication, design - proposes a "cyclic, risk-driven" model similar to Boehm's spiral model - place the KBS in the larger context of the organization (using task, agent, organization models), recognizing that it forms a small part of the org entirety - "first generation KBS used one relatively simple inference engine working on a knowledge base in a particular representational format, usually production rules. But such a knowledge base hides important properties of the reasoning process and knowledge structure in the application domain" (we lose the context in which certain rules were generated, so we want to expose the knowledge level model) - knowledge typingprovide an application independent view of how various knowledge components interact during the problem solving process - defines a metamodel of components of knowledge: domain knowledge (static, concepts, relations and facts needed to reason), control knowledge - two sub parts, (inference knowledge, how to use domain knowledge in inferences), task knowledge, how to decompose top-level reasoning task - an ontology is used to model these components - interaction problem states domain knowledge and control knowledge are highly interdependent (Bylander and Chandrasekaran, Knowledge Acquisition for KBS, 1988) - ontologies are linked together to define this - they use different levels of ontologies to identify the different levels of interoperability and dependency - can we think of PSMs as software patterns? PSM has a competency (goal matching) level, and acceptance criteria, to identify whether a task can be solved with the PSM -structure preserving design approach, - distinctions in the expertise model are preserved in the implementation, and design decisions are explicitly documented -claims the expressiveness of this approach is greater than ER modelling or OO analysis},
description = {sdasda},
biburl = {http://www.bibsonomy.org/bibtex/2bf949b33569e217df0fd317e7ecd4020/grahl},
keywords = {AI knowledge methodology acquisition}
}
@book{russell95,
title = {Artificial {I}ntelligence: {A} modern approach},
address = {New Jersey},
author = {Stuart Russell and Peter Norvig},
publisher = {Prentice-Hall, Inc.},
year = 1995,
id = {121612},
priority = {0},
comment = {hardcover Good AI text which is comprehensive. Some subjects are not treated in great depth but there is a good further reading section. Captures history and development of field pretty well but 1995 was a while ago. Ch 8, 9 and 10 are good introductions to the KR field.},
description = {sdasda},
biburl = {http://www.bibsonomy.org/bibtex/2062b7d92144467ad3cef7d2148def72a/grahl},
keywords = {reference text AI}
}
@book{stelzner88,
title = {The {E}volution of {I}nterface {R}equirements for {E}xpert {S}ystems},
address = {Norwood, NJ},
author = {Marilyn Stelzner and Michael D. Williams},
editor = {Ben Shneiderman},
pages = {285--306},
publisher = {Ablex Publishing},
series = {Expert Systems: The User Interface},
year = 1988,
id = {121564},
priority = {0},
comment = {list for requirements - incr scope of system and number of users req: natural idioms, mapping btw mental and system models immediate feedback recoverability (for experimentation) appropriate granularity (object, intraobject, object hier. (class/subclass, class/member, partof), collection or library of objects and classes multiple interfaces},
description = {sdasda},
biburl = {http://www.bibsonomy.org/bibtex/2d098913f48724004cfda6e149f6fecf6/grahl},
keywords = {visualisation systems AI expert information reasoning}
}
@misc{smith02,
title = {Web Ontology Language (OWL) Guide version 1.0},
author = {M. K. Smith and Deborah Mcguinness and R. Volz and C. Welty},
number = {December 18},
publisher = {World Wide Web Consortium},
volume = 2002,
year = 2002,
url = {http://www.w3.org/TR/owl-guide},
id = {121559},
priority = {0},
comment = {Fairly easy to understand description of what OWL hopes to accomplish, uses the wines example as well},
description = {sdasda},
biburl = {http://www.bibsonomy.org/bibtex/20d71c20b1458e909ef3cfad0f154ef28/grahl},
keywords = {RDF AI semantic web inference}
}
@book{zadeh86,
title = {Is {P}robability {T}heory {S}ufficient for {D}ealing {W}ith {U}ncertainty in {AI}?},
author = {Lotfi Zadeh},
pages = {103--115},
publisher = {North-Holland},
series = {Uncertainty in Artificial Intelligence},
year = 1986,
id = {120110},
priority = {0},
comment = {question: how to represent it is very likely that Mary is young (fuzzy probability and fuzzy predicate) - since prabability is two-valued, an event either happens or it does not, and there is no room for descriptors like warm, short etc. .A proposition is a description of an event. 'It will be a warm day tomorrow' can be assigned a probability of .6 say, but we ignore the nature of what a warm day is. I agree that warm is definitely fuzzy in definition, but what the Bayesians are saying is that the probability is measuring howmuch someone bleives in that statement, not this abstract value called 'warm' and to what degree that is true. I think Cheeseman answers this, but his explanation seems very complicated, resulting in probabilities of probabilities eg. what is the prob. that A beleives it will be warm tomorrow, what is the prob. that warm means >25 degrees, etc.},
description = {sdasda},
biburl = {http://www.bibsonomy.org/bibtex/263ac11d10572b0922a39800c24c45ece/grahl},
keywords = {probabilistic AI}
}
@article{fensel01,
title = {OIL: An Ontology Infrastructure for the World Wide Web},
author = {D. Fensel and F. v. van Harmelen and I. Horrocks and D. Mcguinness and Patel P. Schneider},
journal = {IEEE Intelligent Systems},
pages = {38--45},
year = 2001,
url = {http://www.cs.toronto.edu/~nernst/papers/mcguiness-daml+oil-is.pdf},
id = {111826},
priority = {0},
comment = {Question- who creates the ontology? What is the info you seek to represent isn't in the default ontology? where do we draw the line between ontology and the entire representation (ie ultimately an ontology begins to approach the level of the system itself). eg metadata in GIS poorly written paper {some text representation eg ASCII/Unicode},
description = {sdasda},
biburl = {http://www.bibsonomy.org/bibtex/2c69486297b863d6a84cca4c1871bb85c/grahl},
keywords = {RDF semantic knowledge representation AI web inference}
}
@book{buchanan84,
title = {Rule-{B}ased {E}xpert {S}ystems: {T}he {MYCIN} experiments of the {S}tanford {H}euristic {P}rogramming {P}roject},
address = {Reading, MA},
author = {B. Buchanan and E. Shortliffe},
publisher = {Addison-Wesley},
year = 1984,
id = {111767},
priority = {2},
description = {sdasda},
biburl = {http://www.bibsonomy.org/bibtex/2e44385084203c0d6ff1334ada4a77b39/grahl},
keywords = {AI medical informatics}
}
@techreport{minsky.frames,
title = {{A Framework For Representing Knowledge}},
author = {M. Minsky},
institution = {{MIT-AI Laboratory, Memo 306}},
year = 1974,
owner = {ggrimnes},
description = {My Main bibliography file},
biburl = {http://www.bibsonomy.org/bibtex/20c2ae6e3ef47ff8e1cf31e5be32f43ed/grahl},
keywords = {classic AI}
}
@book{dynamicmemory,
title = {{Dynamic Memory}},
author = {R. C. Schank},
publisher = {{Cambridge University Press}},
year = 1982,
owner = {ggrimnes},
description = {My Main bibliography file},
biburl = {http://www.bibsonomy.org/bibtex/2138c8ad0479bed7843e6238ed0d6982b/grahl},
keywords = {AI machine-learning}
}
@book{BF81,
title = {The Handbook of Artificial Intelligence},
address = {Reading, MA},
editor = {Avron Barr and {Edward A.} Feigenbaum},
publisher = {Addison-Wesley},
year = 1981,
biburl = {http://www.bibsonomy.org/bibtex/2d762cb252aca018c820d2af172a1d3f9/grahl},
keywords = {AI}
}
@book{mitchell97,
title = {Machine Learning},
author = {Tom M. Mitchell},
publisher = {McGraw-Hill},
year = 1997,
biburl = {http://www.bibsonomy.org/bibtex/23e79734ee1a6e49aee02ffd108224d1c/grahl},
keywords = {machine-learning neuralnetworks introduction iownit book AI}
}
@book{525377,
title = {Fluid Concepts and Creative Analogies: Computer Models of the Fundamental Mechanisms of Thought},
address = {New York, NY, USA},
author = {Douglas R. Hofstadter},
publisher = {Basic Books, Inc.},
year = 1996,
isbn = {0465024750},
biburl = {http://www.bibsonomy.org/bibtex/26286082182e33f4229a878f4d1421f02/grahl},
keywords = {hofstadter AI}
}