Integrating Quantitative and Qualitative Discovery: The ABACUS System

Most research on inductive learning has been concerned with qualitative learning that induces conceptual, logic-style descriptions from the given facts. In contrast, quantitative learning deals with discovering numerical laws characterizing empirical data. This research attempts to integrate both types of learning by combining newly developed heuristics for formulating equations with the previously developed concept learning method embodied in the inductive learning program AQ11. The resulting system, ABACUS, formulates equations that bind subsets of observed data, and derives explicit, logic-style descriptions stating the applicability conditions for these equations. In addition, several new techniques for quantitative learning are introduced. Units analysis reduces the search space of equations by examining the compatibility of variables' units. Proportionality graph search addresses the problem of identifying relevant variables that should enter equations. Suspension search focusses the search space through heuristic evaluation. The capabilities of ABACUS are demonstrated by several examples from physics and chemistry.

[1]  Thomas G. Dietterich,et al.  Inductive Learning of Structural Descriptions: Evaluation Criteria and Comparative Review of Selected Methods , 1981, Artif. Intell..

[2]  Ryszard S. Michalski,et al.  Selection of Most Representative Training Examples and Incremental Generation of VL1 Hypotheses: The Underlying Methodology and the Description of Programs ESEL and AQ11 , 1978 .

[3]  Mieczyslaw M. Kokar,et al.  Discovering Functional Formulas through Changing Representation Base , 1986, AAAI.

[4]  S. Chatterjee,et al.  Regression Analysis by Example , 1979 .

[5]  Ryszard S. Michalski,et al.  A theory and methodology of inductive learning , 1993 .

[6]  S. Chatterjee,et al.  Regression Analysis by Example , 1979 .

[7]  M. Kokar.  Determining arguments of invariant functional descriptions , 2004, Machine Learning.

[8]  Jan M. Zytkow,et al.  Combining many searches in the FAHRENHEIT discovery system , 1987 .

[9]  Kenneth D. Forbus.  Qualitative Process Theory , 1984, Artif. Intell..

[10]  Petr Hájek,et al.  The GUHA method—its aims and techniques (twenty-four questions and answers) , 1978 .

[11]  J. W. Gorman,et al.  Fitting Equations to Data. , 1973 .

[12]  Alfred V. Aho,et al.  The Design and Analysis of Computer Algorithms , 1974 .

[13]  Hugh Odishaw,et al.  Handbook of Physics (2nd edition) , 1968 .

[14]  Nayel El-Shafei.  Quantitative Discovery and Reasoning about Failure Mechanisms in Pavement , 1986 .

[15]  Pat Langley,et al.  Data-Driven Discovery of Physical Laws , 1981, Cogn. Sci..

[16]  Robert Earl Stepp,et al.  Conjunctive Conceptual Clustering: A Methodology and Experimentation , 1987 .

[17]  Brian Falkenhainer.  Proportionality Graphs, Units Analysis, and Domain Constraints: Improving the Power and Efficiency of the Scientific Discovery Process , 1985, IJCAI.

[18]  Pat Langley,et al.  Rediscovering Physics with BACON.3 , 1979, IJCAI.

[19]  J. Kleer.  Qualitative and Quantitative Knowledge in Classical Mechanics , 1975 .

[20]  Tom M. Mitchell,et al.  Generalization as Search , 2002 .

[21]  Herbert A. Simon,et al.  The Search for Regularity: Four Aspects of Scientific Discovery , 1984 .

[22]  Janusz S Kowalik,et al.  Papers from the workshop on coupling symbolic and numerical computing in expert systems, Bellevue, Washington, 27-29 August 1985 on Coupling symbolic and numerical computing in expert systems , 1986 .

[23]  H. Simon,et al.  Rediscovering Chemistry with the Bacon System , 1983 .

[24]  H. Langhaar.  Dimensional analysis and theory of models , 1951 .

[25]  W. M. Haynes.  CRC Handbook of Chemistry and Physics , 1990 .

[26]  John R. Anderson,et al.  MACHINE LEARNING An Artificial Intelligence Approach , 2009 .

[27]  Nikolay G. Zagoruiko.  Empirical Prediction Algorithm , 1975, IJCAI.

[28]  Herbert A. Simon,et al.  BACON.5: The Discovery of Conservation Laws , 1981, IJCAI.