The Complexity of Order-Finding for ROABPs

Vishwas Bhargava,Pranjal Dutta,Sumanta Ghosh,Anamay Tengse
2024-11-28
Abstract:We study the \emph{order-finding problem} for Read-once Oblivious Algebraic Branching Programs (ROABPs). Given a polynomial $f$ and a parameter $w$, the goal is to find an order $\sigma$ in which $f$ has an ROABP of \emph{width} $w$. We show that this problem is NP-hard in the worst case, even when the input is a constant degree polynomial that is given in its dense representation. We provide a reduction from CutWidth to prove these results. Owing to the exactness of our reduction, all the known results for the hardness of approximation of Cutwidth also transfer directly to the order-finding problem. Additionally, we also show that any constant-approximation algorithm for the order-finding problem would imply a polynomial time approximation scheme (PTAS) for it. On the algorithmic front, we design algorithms that solve the order-finding problem for generic ROABPs in polynomial time, when the width $w$ is polynomial in the individual degree $d$ of the polynomial $f$. That is, our algorithm is efficient for most/random ROABPs, and requires more time only on a lower-dimensional subspace (or subvariety) of ROABPs. Even when the individual degree is constant, our algorithm runs in time $n^{O(\log w)}$ for most/random ROABPs. This stands in strong contrast to the case of (Boolean) ROBPs, where only heuristic order-finding algorithms are known.
Computational Complexity
What problem does this paper attempt to address?
The problem that this paper attempts to solve is about the sequential search problem of read - once oblivious algebraic branching programs (ROABP). Specifically, given a polynomial \(f\) and a parameter \(w\), the goal is to find an order \(\sigma\) such that \(f\) has a ROABP with width \(w\) under this order. ### Specific description of the problem: 1. **Input**: A polynomial \(f(x_1,\ldots,x_n)\) and a natural number \(w\in\mathbb{N}\). 2. **Output**: Determine whether there exists a permutation \(\sigma\in S_n\) such that the ROABP width of \(f\) under this permutation does not exceed \(w\). ### Main contributions of the paper: - **NP - hard property**: The authors prove that this problem is NP - hard even under the dense representation of constant - degree polynomials. They prove this by reduction from the CutWidth problem to the ROABP sequential search problem. - **Approximation hardness**: The authors also show that any constant - approximation algorithm would imply a polynomial - time approximation scheme (PTAS), which means that it is very difficult to approximate the solution of this problem even. - **Algorithm progress**: For most random or general ROABP, the authors design a polynomial - time algorithm that can efficiently solve the problem in these cases. For some specific cases, the time complexity of the algorithm is \(n^{O(\log w)}\). ### Related formulas and concepts: - **Width of ROABP**: For a given order \(\sigma\), the width of ROABP is defined as the maximum number of vertices in each layer. - **CutWidth problem**: Given a graph \(G\), find a vertex permutation \(\sigma\) such that the maximum cut width is minimized. The cut width refers to the number of edges between the left and right parts at each position after the vertices are arranged in order. ### Formula representation: - **Width of ROABP**: \[ \text{RO - width}_{\sigma}(f)\] - **CutWidth**: \[ \text{CutWidth}_{\sigma}(G)\] ### Main conclusions: - The sequential search problem is NP - hard, even for constant - degree polynomials. - It is also very difficult to approximate the solution of this problem, because any constant - approximation algorithm would imply PTAS. - For most random or general ROABP, there are efficient polynomial - time algorithms. Through the study of these problems, the authors reveal the complexity of the ROABP sequential search problem and provide new ideas for future algorithm design.