Abstract
An algebraic branching program (ABP) A can be modelled as a product expression X1amp;middot; X2… Xd, where X1 and Xd are 1 × w and w × 1 matrices, respectively, and every other Xk is a w × w matrix; the entries of these matrices are linear forms in m variables over a field F (which we assume to be either Q or a field of characteristic poly(m)). The polynomial computed by A is the entry of the 1 × 1 matrix obtained from the product ∏k=1d Xk. We say A is a full rank ABP if the w2(d − 2) + 2w linear forms occurring in the matrices X1, X2, …, Xd are F-linearly independent. Our main result is a randomized reconstruction algorithm for full rank ABPs: Given blackbox access to an m-variate polynomial f of degree at most m, the algorithm outputs a full rank ABP computing f if such an ABP exists, or outputs “no full rank ABP exists” (with high probability). The running time of the algorithm is polynomial in m and β, where β is the bit length of the coefficients of f. The algorithm works even if Xk is a wk−1 × wk matrix (with w0 = wd = 1), and w = (w1, …, wd − 1) is unknown. The result is obtained by designing a randomized polynomial time equivalence test for the family of iterated matrix multiplication polynomial IMMw, d, the (1, 1)-th entry of a product of d rectangular symbolic matrices whose dimensions are according to w ∈ Nd−1. At its core, the algorithm exploits a connection between the irreducible invariant subspaces of the Lie algebra of the group of symmetries of a polynomial f that is equivalent to IMMw, d and the “layer spaces” of a full rank ABP computing f. This connection also helps determine the group of symmetries of IMMw, d and show that IMMw, d is characterized by its group of symmetries.
- Manindra Agrawal. 2005. Proving lower bounds via pseudo-random generators. In Proceedings of the 25th International Conference on Foundations of Software Technology and Theoretical Computer Science (FSTTCS’05). Springer, Berlin, 92--105. Google Scholar
Digital Library
- Manindra Agrawal, Rohit Gurjar, Arpita Korwar, and Nitin Saxena. 2015. Hitting-sets for ROABP and sum of set-multilinear circuits. SIAM J. Comput. 44, 3 (2015), 669--697.Google Scholar
Cross Ref
- Manindra Agrawal and Nitin Saxena. 2006. Equivalence of F-algebras and cubic forms. In Proceedings of the 23rd Annual Symposium on Theoretical Aspects of Computer Science (STACS’06). Springer, Berlin, 115--126. Google Scholar
Digital Library
- Dana Angluin. 1988. Queries and concept learning. Mach. Learn. 2, 4 (1988), 319--342. Google Scholar
Digital Library
- Vikraman Arvind, Partha Mukhopadhyay, and Srikanth Srinivasan. 2008. New results on noncommutative and commutative polynomial identity testing. In Proceedings of the 23rd Annual IEEE Conference on Computational Complexity (CCC’08). 268--279. Google Scholar
Digital Library
- Amos Beimel, Francesco Bergadano, Nader H. Bshouty, Eyal Kushilevitz, and Stefano Varricchio. 2000. Learning functions represented as multiplicity automata. J. ACM 47, 3 (2000), 506--530. Google Scholar
Digital Library
- Elwyn Berlekamp. 1967. Factoring polynomials over finite fields. Bell Syst. Tech. J. 46 (1967), 1853--1859.Google Scholar
Cross Ref
- Lenore Blum, Mike Shub, and Steve Smale. 1988. On a theory of computation over the real numbers; NP completeness, recursive functions and universal machines (extended abstract). In Proceedings of the 29th Annual Symposium on Foundations of Computer Science. 387--397. Google Scholar
Digital Library
- David G. Cantor and Hans Zassenhaus. 1981. A new algorithm for factoring polynomials over finite fields. Math. Comp. 36 (1981), 587--592.Google Scholar
Cross Ref
- Enrico Carlini. 2006. Reducing the number of variables of a polynomial. In Algebraic Geometry and Geometric Modelling, Mathematics and Visualization. Springer, 237--247.Google Scholar
- Marco L. Carmosino, Russell Impagliazzo, Valentine Kabanets, and Antonina Kolokolova. 2016. Learning algorithms from natural proofs. In Proceedings of the 31st Conference on Computational Complexity (CCC’16). 10:1--10:24. Google Scholar
Digital Library
- Zeev Dvir, Rafael Mendes de Oliveira, and Amir Shpilka. 2014. Testing equivalence of polynomials under shifts. In Proceedings of the 41st International Colloquium on Automata, Languages, and Programming (ICALP’14). 417--428.Google Scholar
Cross Ref
- Michael A. Forbes and Amir Shpilka. 2013. Quasipolynomial-time identity testing of non-commutative and read-once oblivious algebraic branching programs. In 54th Annual IEEE Symposium on Foundations of Computer Science, FOCS 2013, 26-29 October, 2013, Berkeley, CA, USA. 243--252. Google Scholar
Digital Library
- Michael A. Forbes, Amir Shpilka, and Ben Lee Volk. 2017. Succinct hitting sets and barriers to proving algebraic circuits lower bounds. In Proceedings of the 49th Annual ACM SIGACT Symposium on Theory of Computing (STOC’17). 653--664. Google Scholar
Digital Library
- Fulvio Gesmundo. 2016. Gemetric aspects of iterated matrix multiplication. J. Alg. 461 (2016), 42--64.Google Scholar
Cross Ref
- Joshua A. Grochow. 2012. Symmetry and Equivalence Relations in Classical and Geometric Complexity Theory. Ph.D. Dissertation. The University of Chicago. Google Scholar
Digital Library
- Joshua A. Grochow, Mrinal Kumar, Michael E. Saks, and Shubhangi Saraf. 2017. Towards an algebraic natural proofs barrier via polynomial identity testing. CoRR abs/1701.01717 (2017).Google Scholar
- Ankit Gupta, Neeraj Kayal, and Satyanarayana V. Lokam. 2011. Efficient reconstruction of random multilinear formulas. In Proceedings of the IEEE 52nd Annual Symposium on Foundations of Computer Science (FOCS’11). 778--787. Google Scholar
Digital Library
- Ankit Gupta, Neeraj Kayal, and Satyanarayana V. Lokam. 2012. Reconstruction of depth-4 multilinear circuits with top fan-in 2. In Proceedings of the 44th Symposium on Theory of Computing Conference (STOC’12). 625--642. Google Scholar
Digital Library
- Ankit Gupta, Neeraj Kayal, and Youming Qiao. 2013. Random arithmetic formulas can be reconstructed efficiently. In Proceedings of the 28th Conference on Computational Complexity (CCC’13). 1--9.Google Scholar
Cross Ref
- Johan Håstad. 1989. Tensor rank is NP-complete. In Proceedings of the 16th International Colloquium on Automata, Languages and Programming (ICALP’89). 451--460. Google Scholar
Digital Library
- Joos Heintz and Claus-Peter Schnorr. 1980. Testing polynomials which are easy to compute (extended abstract). In Proceedings of the 12th Annual ACM Symposium on Theory of Computing. 262--272. Google Scholar
Digital Library
- Erich Kaltofen and Barry M. Trager. 1988. Computing with polynomials given by black boxes for their evaluation: Greatest common divisors, factorization, separation of numerators and denominators. In Proceedings of the 29th Annual Symposium on Foundations of Computer Science. 296--305. Google Scholar
Digital Library
- Zohar Shay Karnin and Amir Shpilka. 2009. Reconstruction of generalized depth-3 arithmetic circuits with bounded top fan-in. In Proceedings of the 24th Annual IEEE Conference on Computational Complexity (CCC’09). 274--285. Google Scholar
Digital Library
- Neeraj Kayal. 2011. Efficient algorithms for some special cases of the polynomial equivalence problem. In Proceedings of the 22nd Annual ACM-SIAM Symposium on Discrete Algorithms (SODA’11). 1409--1421. Google Scholar
Digital Library
- Neeraj Kayal. 2012. Affine projections of polynomials: Extended abstract. In Proceedings of the 44th Symposium on Theory of Computing Conference (STOC’12). 643--662. Google Scholar
Digital Library
- Neeraj Kayal. 2012. An exponential lower bound for the sum of powers of bounded degree polynomials. Proceedings of the Electronic Colloquium on Computational Complexity (ECCC’12), vol. 19, 81.Google Scholar
- Neeraj Kayal, Vineet Nair, and Chandan Saha. 2018. Average-case linear matrix factorization and reconstruction of low width algebraic branching programs. Proceedings of the Electronic Colloquium on Computational Complexity (ECCC’18), vol. 25, 29. Retrieved from https://eccc.weizmann.ac.il/report/2018/029.Google Scholar
- Adam Klivans, Pravesh Kothari, and Igor Carboni Oliveira. 2013. Constructing hard functions using learning algorithms. In Proceedings of the 28th Conference on Computational Complexity (CCC’13). 86--97.Google Scholar
Cross Ref
- Adam Klivans and Amir Shpilka. 2003. Learning arithmetic circuits via partial derivatives. In Proceedings of the 16th Annual Conference on Computational Learning Theory and 7th Kernel Workshop on Computational Learning Theory and Kernel Machines (COLT/Kernel’03). 463--476.Google Scholar
Cross Ref
- Adam Klivans and Daniel A. Spielman. 2001. Randomness efficient identity testing of multivariate polynomials. In Proceedings on 33rd Annual ACM Symposium on Theory of Computing. 216--223. Google Scholar
Digital Library
- A. K. Lenstra, H. W. jun. Lenstra, and Lászlo Lovász. 1982. Factoring polynomials with rational coefficients. Math. Ann. 261 (1982), 515--534.Google Scholar
Cross Ref
- Meena Mahajan and V. Vinay. 1997. Determinant: Combinatorics, algorithms, and complexity. Chicago J. Theor. Comput. Sci. 1997, 5 (1997).Google Scholar
- Daniel Minahan and Ilya Volkovich. 2016. Complete derandomization of identity testing and reconstruction of read-once formulas. Proceedings of the Electronic Colloquium on Computational Complexity (ECCC’16), vol. 23, 171. Google Scholar
Digital Library
- Noam Nisan. 1991. Lower bounds for non-commutative computation (extended abstract). In Proceedings of the 23rd Annual ACM Symposium on Theory of Computing. 410--418. Google Scholar
Digital Library
- Jacques Patarin. 1996. Hidden fields equations (HFE) and isomorphisms of polynomials (IP): Two new families of asymmetric algorithms. In Proceedings of the International Conference on the Theory and Application of Cryptographic Techniques: Advances in Cryptology (EUROCRYPT’96). 33--48. Google Scholar
Digital Library
- Alexander A. Razborov and Steven Rudich. 1997. Natural proofs. J. Comput. Syst. Sci. 55, 1 (1997), 24--35. Google Scholar
Digital Library
- Amir Shpilka. 2007. Interpolation of depth-3 arithmetic circuits with two multiplication gates. In Proceedings of the 39th Annual ACM Symposium on Theory of Computing. 284--293. Google Scholar
Digital Library
- Amir Shpilka and Ilya Volkovich. 2009. Improved polynomial identity testing for read-once formulas. In Proceedings of the 12th International Workshop on Approximation, Randomization, and Combinatorial Optimization. Algorithms and Techniques (APPROX’09), and 13th International Workshop (RANDOM’09). 700--713. Google Scholar
Digital Library
- Amir Shpilka and Amir Yehudayoff. 2010. Arithmetic circuits: A survey of recent results and open questions. Found. Trends Theor. Comput. Sci. 5, 3-4 (2010), 207--388. Google Scholar
Digital Library
- Gaurav Sinha. 2016. Reconstruction of real depth-3 circuits with top fan-in 2. In Proceedings of the 31st Conference on Computational Complexity (CCC’16). 31:1--31:53. Google Scholar
Digital Library
- Thomas Thierauf. 1998. The isomorphism problem for read-once branching programs and arithmetic circuits. Chicago J. Theor. Comput. Sci. 1998, 1 (1998).Google Scholar
- Ilya Volkovich. 2016. A guide to learning arithmetic circuits. In Proceedings of the 29th Conference on Learning Theory (COLT’16). 1540--1561.Google Scholar
Index Terms
Reconstruction of Full Rank Algebraic Branching Programs
Recommendations
On Algebraic Branching Programs of Small Width
In 1979, Valiant showed that the complexity class VPe of families with polynomially bounded formula size is contained in the class VPs of families that have algebraic branching programs (ABPs) of polynomially bounded size. Motivated by the problem of ...
Separation Between Read-once Oblivious Algebraic Branching Programs (ROABPs) and Multilinear Depth-three Circuits
We show an exponential separation between two well-studied models of algebraic computation, namely, read-once oblivious algebraic branching programs (ROABPs) and multilinear depth-three circuits. In particular, we show the following:
(1) There exists an ...
Reconstruction of full rank algebraic branching programs
CCC '17: Proceedings of the 32nd Computational Complexity ConferenceAn algebraic branching program (ABP) A can be modelled as a product expression X1 · X2 ... Xd, where X1 and Xd are 1 X w and w X 1 matrices respectively, and every other Xk is a w × w matrix; the entries of these matrices are linear forms in m variables ...






Comments