Davis, Randall and Douglas B. Lenat. 1982. Knowledge-Based Systems in Artificial Intelligence. New York City, NY: McGraw-Hill.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....natural intelligence and building integrated intelligent systems. To conclude my thoughts on prospects for artificial intelligence, I will say: The light is on the way, but far away. 1. 2 Knowledge Based System There are three major research issues concerning the knowledge based paradigm [6]: ffl Knowledge Representation: In which way should knowledge be represented as symbolic data structures for computer use ffl Knowledge Utilization: What designs are available for the inference procedure ffl Knowledge Acquisition: How can we develop a systematic way for computers to access the ....

R. Davis and D. B. Lenat. Knowledge-Based Systems in Artificial Intelligence. McGraw-Hill International Book Company, 1982.

....knowledge, K, that is the data used in the problem solving process. Of these, goal knowledge has received the least attention. Early research focused on making explicit domain knowledge, K. It produced systems like MYCIN, which captured K using production rules, separate from the rule interpreter [10]. In these systems, however, some strategic knowledge, S, for controlling the order of processing rules was often compiled into the physical order of rule clauses and of rules in K. Consequently, K and S were neither accessible nor interpretable for purposes of explanation, reuse, etc. Later ....

.... 1 # 1 # A # # # . Here, if only the first two sentences apply to , conjunctively they imply that has a 0.79 CF (i.e. 0.7 0.3 (1 0. 7) Having mentioned CFs, ACE SSM uses these to manage uncertainty like in conventional RBSs (e.g. see [10]) For example, consider the above hypothetical network that links diseases and finding. Given the unsatisfied node chain , a K operator that addresses this kind of node chain would return: 1 1 , based on which ....

Davis, R., and Lenat, D., Knowledge-Based Systems in Artificial Intelligence, McGraw-Hill, 1982.

....building of rule based expert systems have been developed in the past. Teiresias was an experimental tool that assisted in the refinement of rule based expert systems by interacting with the user in the analysis of the conclusions concerning single cases, applying meta knowledge about the rulebase [4]. Although such an analysis is certainly useful, an approach as embodied by Teiresias does not give information about how well the system performs over a database of cases. Seek is a system that automatically suggests generalizations and specialization of production rules, based on the analysis of ....

R. Davis and D.B. Lenat, Knowledge-based Systems in Artificial Intelligence (McGrawHill, New York, 1982).

....or, at least, be recorded and flagged of interest. In this example, the conflict between the rule and the data is exactly the reflective equilibrium managed by experts noted above. The meta attributes generated through anomaly detection rules may be used as a basis for machine learning, as Lenat (Davis Lenat 1982) has done in AM (terming anomalies interesting ) However, they are also a basis for effective integration between expert and expert system, with the human expert continuing to have the role of managing the inductive process. 5 Socio Cognitive Foundations of Knowledge Processes The extended ....

Davis, R. & Lenat, D.B. (1982). Knowledge-Based Systems in Artificial Intelligence. New York: McGraw-Hill.

....early stages. It is reasonable to suppose that the level above the representation and processing of knowledge in the computer is that of its acquisition, breakthroughs in machine learning and inductive systems. Two breakthroughs in this area have been Lenat s AM learning mathematics by discovery (Davis and Lenat, 1982) and Michalski s inductive inference of expert rules for plant disease diagnosis (Michalski and Chilausky, 1980) In the fifth generation era machine learning became a highly active research area in its replication phase (Michalski and Carbonell, 1983; Michalski, Carbonell and Mitchell, 1986) ....

Davis, R. and Lenat, D.B. (1982). Knowledge-Based Systems in Artificial Intelligence. New York, McGraw-Hill.

....HR program which works mainly with finite algebras. For clarity, no other programs are discussed. 2. 1 The AM Program The AM program, written by Douglas Lenat, performed concept formation and conjecture making in elementary set theory and elementary number theory, as described in Lenat (1976) and Davis and Lenat (1982). Starting with 115 elementary concepts such as sets and bags, AM would re invent set theory concepts like subsets and disjoint sets, and number theory concepts such as prime numbers and highly composite numbers (with more divisors than any smaller integer) AM would also spot some well known ....

....than the reasons why the facet or action were interesting. When a heuristic was working out how interesting a concept was, it would collate and use another set of heuristics for the task. The heuristics which could measure the interestingness of any concept were recorded as heuristics 9 to 20 in Davis and Lenat (1982), and included: 9] A concept is interesting if there are some interesting conjectures about it. 13] A concept is dull if, after several attempts, only a couple of examples have been found. 15] A concept is interesting if all the examples satisfy the rarely satisfied predicate P. 20] A ....

[Article contains additional citation context not shown here]

R Davis and D Lenat. Knowledge-Based Systems in Artificial Intelligence. McGraw-Hill Advanced Computer Science Series, 1982.

....The most cited work in automated mathematical concept formation is Lenat s work on his AM 1 computer program which invented new definitions and made conjectures based on empirical evidence. This was an exploratory program designed to work in elementary set theory which actually delved 1 See [Davis Lenat 82] into elementary number theory. A notion of interestingness was used to maintain an agenda of tasks to do next, calculated using values for how interesting old concepts were and values for how worthwhile a task involving those concepts might be. The calculations are complex and based on many ....

R Davis and D Lenat. KnowledgeBased Systems in Artificial Intelligence. McGraw-Hill Advanced Computer Science Series, 1982.

....by Langley et al., the concepts formed were polynomial relations between variables in physical systems. A concept was interesting if the polynomial relation was observable in the data, and uninteresting if not. In this case, they were able to guarantee some level of interestingness by look 1 See [Davis Lenat 82] 2 See [Langley et al. 87] ing at the data first, spotting patterns and trends and forming the new relations in this data driven manner. Another example of narrowing the search space is Sims IL 3 program, which accepted specifications for an operator (for instance the multiplication ....

R Davis and D Lenat. KnowledgeBased Systems in Artificial Intelligence. McGraw-Hill Advanced Computer Science Series, 1982.

....other areas of mathematics. 3. Produce results understandable by mainstream mathematicians. 4. Keep the theory behind the concept formation clear and concise. 1. 1 Background Machine discovery in mathematics had an enthusiastic start and early programs, such as Lenat s AM, as described in [4], excited people because they re invented classically interesting mathematical concepts. For example, AM looked at elementary number theory, and re invented well known concepts such as prime numbers and square numbers, and well known conjectures such as Goldbach s conjecture (that every even ....

.... 2 G a b a b C1.1 1 1 1 C2.1 1 1 1 C2.1 2 1 2 C2.1 1 2 2 C2.1 2 2 1 C2.2 1 1 2 C2.2 2 1 1 C2.2 1 2 1 C2.2 2 2 2 conjunct = 1; 2; 3; 4] Table 4 G a b a b C1.1 1 1 1 C2.1 1 1 1 C2.1 1 2 2 C2.1 2 1 2 C2.1 2 2 1 C2.2 1 1 2 C2.2 1 2 1 C2.2 2 1 1 C2.2 2 2 2 (Note that the [1,2,3,4] parameters say that column 1 of table 0 should match column 1 of table 2, column 2 of table 0 should match column 2 of table 2, and so on) Table 4 is exactly the same as table 1 because the functional definition of the calculation producing table 4 is the following: f 4 (G) f(a; b; c) 2 G ....

R Davis and D Lenat, Knowledge-Based Systems in Artificial Intelligence, McGraw-Hill Advanced Computer Science Series, 1982.

....of creativity to whoever has to pore over the output looking for the good stuff. For example, even though they utilized powerful heuristic search strategies, Douglas Lenat s AM and Eurisko programs required a human perceiver to filter their results in order to find and remove the bad ones [Lenat, 1982][Lenat, 1983] Although AM and Eurisko discovered some mathematical theorems, they had help in high places. Another problem with blind manipulation is that it can easily lead to combinatorial explosion, as Chuck Rieger s work on syntactic, match driven inference chaining clearly shows [Rieger, ....

Lenat, D. (1982). Knowledge-Based Systems in Artificial Intelligence, chapter AM: Discovery in mathematics as heuristic search, pages 1--225. McGrawHill, New York.

....definitions. The close connection between the class algebra logical class definitions and the set of class instances is what makes this reasoning system so powerful. The use of examples and counter examples in artificial intelligence is wellknown. For example, Lenat s famous program AM [6] was able to discover the interesting concepts such as sum, product, prime numbers, and prime number pairs by looking at examples and counterexamples. 2. Related logical problems 2.1 Undecidability Previous attempts at relating logic and sets have been hindered by the undecidability of most ....

Randall Davis and Douglas B. Lenat, Knowledge-Based Systems in Artificial Intelligence, McGraw-Hill, Inc., New York, 1982.

....the easiest method for refin ing knowledge based systems . 7 These arguments are not conclusive. The refinemen t of rule bases involving changes in their structure is likely 7 to be sometimes necessary. Some relevant empirical work includes ( Davis and Lenat [12]; Rada [30] Rada [31] Politakis [33] G insberg [19] The literature on refinemen t of expert system s written in Prolog and OPS5 is h igh ly relevan t when a numeric mechan ism for uncertain reason ing is not used ( Shapiro [33] Brownston et al. 6] Select other window and hit enter to ....

....( and efficiency ) of refinemen t at the analysis and design levels or, equivalently, during earlier stages of the KBS lifecycle. Meta knowledge and records of project history may prove useful in refining an implemented expert system . This was partly shown in TEIRESIAS ( Davis and Lenat [12]) However, TEIRESIAS and most other knowledge based approaches to knowledge base refinement do not deal with problems arising from deleterious in teractions of rules. For a brief discussion con trasting knowledge based and empirical approaches see ( Valtorta [43] For additional ....

Davis, R . and D .B. Lenat. Knowledge-Based Systems in Artificial Intelligence. New York: McG raw-H ill, 1982. Select other window and hit enter to continue... ## ## 30

....which are taught as procedures. Free associate. Reflect endlessly, in a stream of consciousness manner[15] This demonstrates cognitive vigor (see below) in its purest form. In this mode of operation, and with the right heuristics) the system could approximate the discovery behavior of AM[14]. 3.2 Ls 1 s Memories Ls 1 has three memories: Taxonomy A conventional, tangled tree hierarchy of object types, which provides the basis for a typed predicate calculus, with which to encode declarative knowledge. The types are also used to identify the domain and range of procedures known to ....

Davis, R. and Lenat, D. B. (1982). Knowledge-Based Systems in Artificial Intelligence. New York: McGraw-Hill.

....modeled as contextual interpretation of surprise, and it is possible to regard meta rules based on surprise as incorporating some of the attention directing aspects of human emotion. The meta attributes generated through anomaly detection rules may be used as a basis for machine learning, as Lenat (Davis Lenat 1982) has done in AM (terming surprise interesting ) However, they are also a basis for effective integration between expert and expert system, with the human expert continuing to have the role of managing the inductive process. 5 Conclusions The role of the expert in the knowledge processes of ....

Davis, R. & Lenat, D.B. (1982). Knowledge-Based Systems in Artificial Intelligence. New York: McGraw-Hill.

....in order for the designer to be able to recover the design object from the verbal specification. In fact, given sufficient information, the specification process might lead to implementable specifications or even to concrete solutions in which case we talk about executable specification (c.f. Davis, 1982 ] Partridge and Wilks, 1987 ] Rattray et al. 1991 ] Lenart et al. 1994 ] Usually, however, there is a huge gap between informal specification and design implementation. The author s goal is to minimize this gap and support the automation of this process. For this, they are ....

R. Davis. Knowledge-Based Systems in Artificial Intelligence, chapter Teiresias: Applications of meta-level knowledge, pages 227--490. McGraw-Hill, New York, 1982.

....AI toolkits only support the knowledge engineer during the phases of implementing and debugging the system. No support is normally provided at the knowledge acquisition stage. On the other hand, a number of efforts have been made toward the direction of automating the knowledge acquisition process [26] [10] 20] 14] For better or worse, automated knowledge acquisition hasn t yet reached the point of making the knowledge engineer redundant, and, for most applications, it is still necessary for the knowledge engineer to tackle all of the activities described in our model. To understand why this ....

Davis, R., Lenat, D. Knowledge-Based Systems in Artificial Intelligence. McGraw-Hill, 1982.

....of Edinburgh Edinburgh EH1 1HN United Kingdom Theory Formation Programs in Pure Mathematics Broadly speaking, machine learning programs are asked to identify a single concept given a set of examples and some background knowledge. Mathematical theory formation programs, such as the AM program, (Davis Lenat 1982) and the HR program, Colton, Bundy, Walsh 1999) are also given a set of examples and some background knowledge. However, they are not asked to find a single concept, but rather to explore the domain and attempt to gain some understanding of it. Because the domain is mathematics, there are a ....

Davis, R., and Lenat, D. 1982. Knowledge-Based Systems in Artificial Intelligence. McGraw-Hill Advanced Computer Science Series.

No context found.

Davis, Randall and Douglas B. Lenat. 1982. Knowledge-Based Systems in Artificial Intelligence. New York City, NY: McGraw-Hill.

No context found.

Davis, Randall and Douglas B. Lenat. 1982. Knowledge-Based Systems in Artificial Intelligence. New York City, NY: McGraw-Hill.

No context found.

R. Davis. Knowledge-Based Systems in Artificial Intelligence, chapter Teiresias: Applications of meta-level knowledge, pages 227-490. McGraw-Hill, New York.

No context found.

avis and D.B. Lenat, Knowledge-Based Systems in Artificial Intelligence, McGraw-14ill, New York, 1982.

No context found.

.B. Lenat, Knowledge-Based Systems in Artificial Intelligence, McGraw-Itill, New York, 1982.

No context found.

R. Davis and D. Lenat, Knowledge-Based Systems in Artificial Intelligence, McGraw Hill, 1982.

No context found.

B. Lenat (Ed.), Knowledge based systems in Artificial Intelligence. McGraw-Hill. Carbonell, J. G., & Gil, Y. (1987). Learning By Experimentation. In Proceedings of Forth International Workshop on Machine Learning (pp. 256-266). Irvine, California: Morgan Kaufmann.

No context found.

Davis, R., Lenat, D. Knowledge-Based Systems in Artificial Intelligence. McGraw-Hill, 1982.

First 50 documents

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC