Hidden linear function problem

From HandWiki

The hidden linear function problem, is a search problem that generalizes the Bernstein–Vazirani problem.[1] In the Bernstein–Vazirani problem, the hidden function is implicitly specified in an oracle; while in the 2D hidden linear function problem (2D HLF), the hidden function is explicitly specified by a matrix and a binary vector. 2D HLF can be solved exactly by a constant-depth quantum circuit restricted to a 2-dimensional grid of qubits using bounded fan-in gates but can't be solved by any sub-exponential size, constant-depth classical circuit using unbounded fan-in AND, OR, and NOT gates.[2] While Bernstein–Vazirani's problem was designed to prove an oracle separation between complexity classes BQP and BPP, 2D HLF was designed to prove an explicit separation between the circuit classes [math]\displaystyle{ QNC^{0} }[/math] and [math]\displaystyle{ NC^{0} }[/math] ([math]\displaystyle{ QNC^{0} \nsubseteq NC^{0} }[/math]).[1]

2D HLF problem statement

Given [math]\displaystyle{ A \in \mathbb{F}_2^{n \times n} }[/math](an upper-triangular binary matrix of size [math]\displaystyle{ n \times n }[/math]) and [math]\displaystyle{ b \in \mathbb{F}_2^n }[/math] (a binary vector of length [math]\displaystyle{ n }[/math]),

define a function [math]\displaystyle{ q : \mathbb{F}_2^n \to \mathbb{Z}_4 }[/math]:

[math]\displaystyle{ q(x) = (2 x^T A x + b^T x) \bmod 4 = \left(2 \sum_{i,j}A_{i,j} x_i x_j + \sum_i b_i x_i \right) \bmod 4 , }[/math]

and

[math]\displaystyle{ \mathcal{L}_q = \Big\{x \in \mathbb{F}_2^n : q(x \oplus y) = (q(x) + q(y)) \bmod 4 ~~ \forall y \in \mathbb{F}_2^n \Big\}. }[/math]

There exists a [math]\displaystyle{ z \in \mathbb{F}_2^n }[/math] such that

[math]\displaystyle{ q(x) = 2 z^T x ~~\forall x \in \mathcal{L}_q. }[/math]

Find [math]\displaystyle{ z }[/math].[1]

2D HLF algorithm

With 3 registers; the first holding [math]\displaystyle{ A }[/math], the second containing [math]\displaystyle{ b }[/math] and the third carrying an [math]\displaystyle{ n }[/math]-qubit state, the circuit has controlled gates which implement [math]\displaystyle{ U_q = \prod_{1 \lt i \lt j \lt n} CZ_{ij}^{A_{ij}} \cdot \bigotimes_{j=1}^n S_j^{b_j} }[/math] from the first two registers to the third.

This problem can be solved by a quantum circuit, [math]\displaystyle{ H^{\otimes n} U_q H^{\otimes n} \mid 0^n \rangle }[/math], where H is the Hadamard gate, S is the S gate and CZ is CZ gate. It is solved by this circuit because with [math]\displaystyle{ p(z) = \left| \langle z | H^{\otimes n} U_q H ^ {\otimes n} | 0^n \rangle \right|^2 }[/math], [math]\displaystyle{ p(z)\gt 0 }[/math] iff [math]\displaystyle{ z }[/math] is a solution.[1]

References

  1. 1.0 1.1 1.2 1.3 "Quantum advantage with shallow circuits". Science 362 (6412): 308–311. 2018-10-19. doi:10.1126/science.aar3106. PMID 30337404. Bibcode2018Sci...362..308B. 
  2. "Exponential separation between shallow quantum circuits and unbounded fan-in shallow classical circuits". Proceedings of the 51st Annual ACM SIGACT Symposium on Theory of Computing. 362. Association for Computing Machinery. June 2019. pp. 515–526. doi:10.1145/3313276.3316404. ISBN 9781450367059. 

External links