Relationless Completeness and Separations

This paper extends Valiant’s work on VP and VNP to the settings in which variables are not multiplicatively commutative and/or associative. Our main result is a theory of completeness for these algebraic worlds. We define analogs of Valiant’s classes VP and VNP, as well as of the polynomials permanent and determinant, in these worlds. We then prove that even in a completely relationless world which assumes no commutativity nor associativity, permanent remains VNP-complete, and determinant can polynomially simulate any arithmetic formula, just as in the standard commutative, associative world of Valiant. In the absence of associativity, the completeness proof gives rise to the following combinatorial problem: what is the smallest binary tree which contains as minors all binary trees with n leaves. We give an explicit construction of such a universal tree of polynomial size, a result of possibly independent interest. Given that such non-trivial reductions are possible even without commutativity and associativity, we turn to lower bounds. In the non-associative, commutative world we prove exponential circuit lower bounds on explicit polynomials, separating the non-associative commutative analogs of VP and VNP. Obtaining such lower bounds and a separation in the complementary associative, non-commutative world has been open for about 30 years.

[1]  Mark Braverman,et al.  Pebbles and Branching Programs for Tree Evaluation , 2012, TOCT.

[2]  M. S. Paterson Poceedings of the London Mathematical Society symposium on Boolean function complexity , 1992 .

[3]  Peter Bürgisser,et al.  Completeness and Reduction in Algebraic Complexity Theory , 2000, Algorithms and computation in mathematics.

[4]  Leslie G. Valiant,et al.  NP is as easy as detecting unique solutions , 1985, STOC '85.

[5]  Avi Wigderson,et al.  Non-commutative circuits and the sum-of-squares problem , 2010, Symposium on the Theory of Computing.

[6]  L. Valiant Why is Boolean complexity theory difficult , 1992 .

[7]  Leslie G. Valiant,et al.  Fast Parallel Computation of Polynomials Using Few Processors , 1983, SIAM J. Comput..

[8]  Noam Nisan,et al.  Lower bounds on arithmetic circuits via partial derivatives , 1995, Proceedings of IEEE 36th Annual Foundations of Computer Science.

[9]  Steve Chien,et al.  Algebras with Polynomial Identities and Computing the Determinant , 2007, SIAM J. Comput..

[10]  Leslie G. Valiant,et al.  General Context-Free Recognition in Less than Cubic Time , 1975, J. Comput. Syst. Sci..

[11]  Leslie G. Valiant,et al.  Completeness classes in algebra , 1979, STOC.

[12]  Guillaume Malod,et al.  Characterizing Valiant's algebraic complexity classes , 2006, J. Complex..

[13]  Christopher Umans,et al.  Group-theoretic Algorithms for Matrix Multiplication , 2005, FOCS.

[14]  Joel Friedman,et al.  Expanding graphs contain all small trees , 1987, Comb..

[15]  Don Coppersmith,et al.  Matrix multiplication via arithmetic progressions , 1987, STOC.

[16]  Steve Chien,et al.  Clifford algebras and approximating the permanent , 2003, J. Comput. Syst. Sci..

[17]  Noam Nisan,et al.  Lower bounds for non-commutative computation , 1991, STOC '91.

[18]  Laurent Hyafil On the Parallel Evaluation of Multivariate Polynomials , 1979, SIAM J. Comput..

[19]  Alexander A. Razborov,et al.  Natural Proofs , 1997, J. Comput. Syst. Sci..

[20]  M. K. Gol'dberg,et al.  On minimal universal trees , 1968 .