RIEVL: Recursive induction learning in hand gesture recognition
 Author(s): Zhao, Meide; Quek, Francis K.H.; Wu, Xindong
 Author Affiliation: Univ of Illinois at Chicago
 Source: IEEE Transactions on Pattern Analysis and Machine Intelligence v 20 n 11 Nov 1998 IEEE Comp Soc Los
 Alamitos CA USA p 1174-1185 0162-8828 ITPIDJ

 Abstract: This paper presents a recursive inductive learning scheme that is able to acquire hand pose models in the form of
 disjunctive normal form expressions involving multivalued features. Based on an extended variable-valued logic, our
 rule-based induction system is able to abstract compact rule sets from any set of feature vectors describing a set of
 classifications. The rule bases which satisfy the completeness and consistency conditions are induced and refined through five
 heuristic strategies. A recursive induction learning scheme in the RIEVL algorithm is designed to escape local minima in the
 solution space. A performance comparison of RIEVL with other inductive algorithms, ID3, NewID, C4.5, CN2, and HCV,
 is given in the paper. In the experiments with hand gestures, the system produced the disjunctive normal form descriptions of
 each pose and identified the different hand poses based on the classification rules obtained by the RIEVL algorithm. RIEVL
 classified 94.4 percent of the gesture images in our testing set correctly, outperforming all other inductive algorithms. English
 (Author abstract) 31 Refs.

 Subjects: Learning systems Feature extraction; Recursive functions; Heuristic methods; Formal logic; Algorithms;
 Classification Codes: 723.4; 723.5; 721.1; 921
 Document Type: JA
 Identifiers: Recursive induction learning; Rule-based induction; Variable-valued logic

 Abstract: At present, algorithms of the ID3 family are not based on background knowledge. For that reason, most of the
 time they are neither logical nor understandable to experts. These algorithms cannot perform different types of generalization
 as others can do, nor can they can reduce the cost of classifications. The algorithm presented in this paper tries to generate
 more logical and understandable decision trees than those generated by ID3-like algorithms; it executes various types of
 generalization and at the same time reduces the classification cost by means of background knowledge. The background
 knowledge contains the ISA hierarchy and the measurement cost associated with each attribute. The user can define the
 degrees of economy and generalization. These data will influence directly the quantity of search that the algorithm must
 undertake. This algorithm, which is an attribute version of the EG2 method, has been implemented and the results appear in
 this paper comparing them with other methods. English (Author abstract) 19 Refs.

 Subjects: Computer Systems Programming Mathematical Techniques--Trees; Learning Systems; Expert
 Systems--Knowledge Bases; Computer Programming--Algorithms;
 Classification Codes: 723; 921
 Document Type: JA
 Identifiers: Decision Tree Induction; Knowledge Acquisition; Backround Knowledge

 Author(s): Baldwin, J.F.; Lawry, J.; Martin, T.P.
 Author Affiliation: Univ of Bristol
 Source: International Journal of Intelligent Systems v 12 n 7 Jul 1997 John Wiley & Sons Inc New York NY USA p
 523-552 0884-8173 IJISED

 Abstract: A mass assignment based ID3 algorithm for learning probabilistic fuzzy decision is introduced. Fuzzy partitions are
 used to continuous feature universes and to reduce complexity when universes are discrete but with large cardinalities.
 Furthermore, the fuzzy partitioning of classification universes facilitates the use of these decision trees in function
 approximation problems. Generally the incorporation of fuzzy sets into this paradigm overcomes many of the problems
 associated with the application of decision trees to real-world problems. The probabilities required for the trees are calculated
 according to mass assignment theory applied to fuzzy labels. The latter concept is introduced to overcome computational
 complexity problems associated with higher dimensional mass assignment evaluations on databases. English (Author
 abstract) 11 Refs.

 Subjects: Artificial intelligence Trees (mathematics); Fuzzy sets; Learning algorithms; Probabilistic logics; Computational
 complexity; Database systems;
 Classification Codes: 723.4; 921.4; 921; 723; 721.1; 922.1
 Document Type: JA
 Identifiers: Decision trees; Fuzzy partitioning

 Author(s): Gaga, Lena; Moustakis, Vassilis; Vlachakis, Yannis; Charissis, Giorgos
 Author Affiliation: Inst of Computer Science
 Source: Applied Artificial Intelligence v10 n2 Mar-Apr 1996 Taylor & Francis Ltd Basingstoke Engl p 79-94 0883-9514
 AAINEH

 Abstract: Learning from patient records may aid medical knowledge acquisition and decision making. Decision tree
 induction, based on ID3, is a well-known approach of learning from examples. In this article we introduce a new data
 representation formalism that extends the original ID3 algorithm. We propose a new algorithm, ID+, which adopts this
 representation scheme. ID+ provides the capability of modeling dependencies between attributes or attribute values and of
 handling multiple values per attribute. We demonstrate our work via a series of medical knowledge acquisition experiments
 that are based on a 'real-world' application of acute abdominal pain in children. In the context of these experiments, we
 compare ID+ with C4.5, Newld, and a Naive Bayesian classifier. Results demonstrate that the rules acquired via ID+
 improve decision tree clinical comprehensibility and complement explanations supported by the Naive Bayesian classifier,
 while in terms of classification, accuracy decrease is marginal. English (Author abstract) 33 Refs.

 Subjects: Knowledge acquisition Medical computing; Learning systems; Knowledge representation; Learning algorithms;
 Computer simulation; Data structures; Data handling; Decision making; Decision theory; Pattern recognition; Knowledge
 based systems;
 Classification Codes: 723.4; 723.5; 461.1; 461.6; 723.2; 723.4.1
 Document Type: JA
 Identifiers: Decision tree induction; Medical knowledge acquisition; Data representation; Naive Bayesian classifier

 Author(s): Broner, Israel; King, J. Phillip; Nevo, Amnon
 Author Affiliation: Colorado State Univ
 Source: Computers and Electronics in Agriculture v 5 n 2 Oct 1990 p 87-99 0168-1699 CEAGE6

 Abstract: This paper gives a general discussion of knowledge acquisition and formalization using structured induction, and
 illustrates the application of induction with an example. The example is taken from actual work on a knowledge based system
 for malting barley crop management, where the ID3 algorithm was used to derive rules and generate a decision tree to make
 the irrigation decision. The objective of the paper is to give the reader a working understanding of the principles of induction
 and the mechanics of the ID3 algorithm. English (Author abstract) 13 Refs
 
Subjects: Agricultural Engineering Expert Systems--Knowledge Bases; Grain; Irrigation--Management; Computer
Programming--Algorithms;
Classification Codes: 821; 723
Document Type: JA
Identifiers: Structural Induction; Knowledge Acquisition

@InProceedings{
                mans91,
        Author =
                "Y. Mansuri and P. Compton and C. Sammut",
          Title =
                "A comparison of a manual knowledge acquisition method and an inductive learning method",
      Booktitle =
                " Australian workshop on knowledge acquisition for knowledge based systems, Pokolbin",
      Publisher =
                "University of Technology, Sydney",
          Year =
                1991,
         Editor =
                "J. Boose and J. Debenham and B. Gaines and J. Quinlan",
         Pages =
                "114-132",
         Notes =
                " compared manual KA using RDR versus automatic inductive learning using ID3 [Quinlan82]. An example
                set of 9514 examples (from Garvin ES-1 [Compton89]) was divided into a test set of 1514 examples and
                a training set of 8000 examples. For each item in the test suite, an RDR tree was executed. If a human
                operator detected an incorrect classification, the tree was patched. This item was then added into an ID3
                training set. The percentage errors of the current RDR tree and the ID3 tree were compared. The
                comparison shows that in the case of RDR vs ID3 in the Garvin ES-1 domain, manual RDR KA generated
                KBs with much fewer errors than the ID3 inductive learner when only hundreds of examples were
                available. Further, only when thousands of examples are available (in this case, 5000) did manual KA and
                ML perform equally as well.",
         From =
                [Timm]}