Independence and port oracles for matroids, with an application to computational learning theory

Given a matroidM with distinguished elemente, aport oracie with respect toe reports whether or not a given subset contains a circuit that containse. The first main result of this paper is an algorithm for computing ane-based ear decomposition (that is, an ear decomposition every circuit of which contains elemente) of a matroid using only a polynomial number of elementary operations and port oracle calls. In the case thatM is binary, the incidence vectors of the circuits in the ear decomposition form a matrix representation forM. Thus, this algorithm solves a problem in computational learning theory; it learns the class ofbinary matroid port (BMP) functions with membership queries in polynomial time. In this context, the algorithm generalizes results of Angluin, Hellerstein, and Karpinski [1], and Raghavan and Schach [17], who showed that certain subclasses of the BMP functions are learnable in polynomial time using membership queries. The second main result of this paper is an algorithm for testing independence of a given input set of the matroidM. This algorithm, which uses the ear decomposition algorithm as a subroutine, uses only a polynomial number of elementary operations and port oracle calls. The algorithm proves a constructive version of an early theorem of Lehman [13], which states that the port of a connected matroid uniquely determines the matroid.

[1]  Leslie G. Valiant,et al.  On the learnability of Boolean formulae , 1987, STOC.

[2]  W. T. Tutte.  An algorithm for determining whether a given binary matroid is graphic. , 1960 .

[3]  James Oxley,et al.  On inequivalent representations of matroids , 1996 .

[4]  Lisa Hellerstein,et al.  Learning binary matroid ports , 1994, SODA '94.

[5]  Vijay Raghavan,et al.  Learning μ-branching programs with queries , 1993, COLT '93.

[6]  Paul D. Seymour,et al.  Recognizing graphic matroids , 1981 .

[7]  Paul D. Seymour,et al.  A note on the production of matroid minors , 1977, J. Comb. Theory, Ser. B.

[8]  Klaus Truemper,et al.  Matroid decomposition , 1992 .

[9]  Bernhard Korte,et al.  Complexity of Matroid Property Algorithms , 1982, SIAM J. Comput..

[10]  Stephen R. Schach,et al.  Learning switch configurations , 1990, COLT '90.

[11]  Marek Karpinski,et al.  Learning read-once formulas with queries , 1993, JACM.

[12]  James G. Oxley,et al.  Matroid theory , 1992 .

[13]  Marek Karpinski,et al.  An Algorithm to Learn Read-Once Threshold Formulas, and some Generic Transformations between Learning Models , 1993 .

[14]  Robert E. Bixby,et al.  Converting Linear Programs to Network Problems , 1980, Math. Oper. Res..

[15]  Jeff Kahn,et al.  On the Uniqueness of Matroid Representations Over GF(4) , 1988 .

[16]  Paul Seymour,et al.  The Forbidden Minors of Binary Clutters , 1976 .

[17]  Robert E. Bixby,et al.  An Almost Linear-Time Algorithm for Graph Realization , 1988, Math. Oper. Res..

[18]  Leslie G. Valiant,et al.  Computational limitations on learning from examples , 1988, JACM.

[19]  Alfred Lehman,et al.  A Solution of the Shannon Switching Game , 1964 .

[20]  Harvey J. Greenberg,et al.  Advanced techniques in the practice of operations research , 1982 .

[21]  Paul D. Seymour,et al.  The matroids with the max-flow min-cut property , 1977, J. Comb. Theory B.

[22]  Leslie G. Valiant,et al.  Cryptographic limitations on learning Boolean formulae and finite automata , 1994, JACM.