Learning arithmetic circuits via partial derivatives

Adam R. Klivans*, Amir Shpilka

*Corresponding author for this work

Research output: Contribution to journalConference articlepeer-review


We present a polynomial time algorithm for learning several models of algebraic computation. We show that any arithmetic circuit whose partial derivatives induce a "low"-dimensional vector space is exactly learnable from membership and equivalence queries. As a consequence we obtain the first polynomial time algorithm for learning depth three multilinear arithmetic circuits. In addition, we give the first polynomial time algorithms for learning restricted algebraic branching programs and noncommutative arithmetic formulae. Previously only restricted versions of depth-2 arithmetic circuits were known to be learnable in polynomial time. Our learning algorithms can be viewed as solving a generalization of the well known polynomial interpolation problem where the unknown polynomial has a succinct representation. We can learn representations of polynomials encoding exponentially many monomials. Our techniques combine a careful algebraic analysis of arithmetic circuits' partial derivatives with the "multiplicity automata" techniques due to Beimel et al [BBB+00].

Original languageEnglish
Pages (from-to)463-476
Number of pages14
JournalLecture Notes in Computer Science
StatePublished - 2003
Externally publishedYes
Event16th Annual Conference on Learning Theory and 7th Kernel Workshop, COLT/Kernel 2003 - Washington, DC, United States
Duration: 24 Aug 200327 Aug 2003


  • Learning with queries
  • PAC learning


Dive into the research topics of 'Learning arithmetic circuits via partial derivatives'. Together they form a unique fingerprint.

Cite this