Jeffrey C. Schlimmer

ICML Conference 1993 Conference Paper

Efficiently Inducing Determinations: A Complete and Systematic Search Algorithm that Uses Optimal Pruning

• Jeffrey C. Schlimmer

AAAI Conference 1991 Conference Paper

Learning Meta Knowledge for Database Checking

• Jeffrey C. Schlimmer

Building a large-scale system often involves creating a large knowledge store, and as these grow and are maintained by a number of individuals, errors are inevitable. Exploring databases as a specialization of knowledge stores, this paper studies the hypothesis that descriptive, learned models can be prescriptively used to find errors. To that end, it describes an implemented system called CARPER. Applying CARPER to a real-world database demonstrates the viability of the approach and establishes a baseline of performance for future research.

AAAI Conference 1987 Conference Paper

Learning and Representation Change

• Jeffrey C. Schlimmer

To remain effective without human interaction, intelligent systems must be able to adapt to their environment. One useful form of adaptation is to incrementally form concepts from examples for the purposes of inference and problem-solving. A number of systems have been constructed for this task, yet their capability is limited by the language used to represent concepts. This paper presents an extension to the concept acquisition system STAGGER that allows it to utilize continuously valued attributes. The combination of methods employed is able to dynamically acquire appropriate representations, thereby minimizing the impact of initial representational bias decisions. Of additional interest is the distinction between the computational flavor of the learning methods, for one is similar to connectionist approaches while the other two are of a more symbolic nature.

AAAI Conference 1986 Conference Paper

A Case Study of Incremental Concept Induction

• Jeffrey C. Schlimmer

Applicat, ioli of niactiirie inductiori l, cx-tlrliques in corriplcx doltlailis promises to push the computational limits of nofiincrc~rnerila. 1, search ilitensive induction methods. 1, earning t~ff; ~ctiverlcss in complex domains requires the developrrlerlt of iricrorrierital, cost effective methods. However, discussion 01 ciinierisioris fi>r corriparirig I, he utilily of diffcrirrg increrr~e~~tal melhods teas beers lacking. 111 this paper we intro- (iuc(~: j dimensions for ctiaracterixirig incremental concept iriducLior1 syst, enis wllicti relate to the cost anti qualily of ltlarriifig. ‘ t’ he dimensions are used lo compare lhe respec- 1ivtl Irlt: rit, s of 4 iricrerrleritat variants of Quirilari’ s learning Irol~~ t>x; llrlples program, IlI3. This coniparisori indicates Itlat, cost effert, ive iriductiori car1 be obtained, without sigrlific-; Llktly det, racting frorrl the quality of induct4 knowt- CTlgt~.