Tensor Reconstruction Beyond Constant Rank
Abstract
We give reconstruction algorithms for subclasses of depth-3 arithmetic circuits. In particular, we obtain the first efficient algorithm for finding tensor rank, and an optimal tensor decomposition as a sum of rank-one tensors, when given black-box access to a tensor of super-constant rank. We obtain the following results: 1. A deterministic algorithm that reconstructs polynomials computed by circuits in time 2. A randomized algorithm that reconstructs polynomials computed by multilinear circuits in time 3. A randomized algorithm that reconstructs polynomials computed by set-multilinear circuits in time , where if is a finite field, and equals the maximum bit complexity of any coefficient of if is infinite. Prior to our work, polynomial time algorithms for the case when the rank, , is constant, were given by Bhargava, Saraf and Volkovich [BSV21]. Another contribution of this work is correcting an error from a paper of Karnin and Shpilka [KS09] that affected Theorem 1.6 of [BSV21]. Consequently, the results of [KS09, BSV21] continue to hold, with a slightly worse setting of parameters. For fixing the error we study the relation between syntactic and semantic ranks of circuits. We obtain our improvement by introducing a technique for learning rank preserving coordinate-subspaces. [KS09] and [BSV21] tried all choices of finding the "correct" coordinates, which led to having a fast growing function of at the exponent of . We find these spaces in time that is growing fast with , yet it is only a fixed polynomial in .
Cite
@article{arxiv.2209.04177,
title = {Tensor Reconstruction Beyond Constant Rank},
author = {Shir Peleg and Amir Shpilka and Ben Lee Volk},
journal= {arXiv preprint arXiv:2209.04177},
year = {2022}
}
Comments
Abstract shortened to meet arXiv requirements; 59 pages