Source-linked AI summary

Reflections for quantum query algorithms

Ben W. Reichardt

arXiv:1005.1601v1quant-ph

TL;DR

The paper addresses whether the general adversary lower bound and span-program-based quantum algorithms can be made exactly tight without the prior logarithmic overhead. It constructs a direct dual-SDP algorithm alternating two reflections, proving optimal bounded-error query complexity and exact composition, while leaving broader classical–quantum query relationships open.

  • Problem

    The prior span-program algorithm was tight only up to a sub-logarithmic factor caused by converting a continuous-time query algorithm into a discrete-query algorithm.

  • Method

    The paper uses the dual SDP for the general adversary bound to construct a simpler algorithm alternating the input oracle with a fixed reflection.

  • Results

    The general adversary bound exactly characterizes quantum query complexity for boolean functions, and composed boolean functions achieve the product of their query complexities without logarithmic error reduction.

  • Takeaways & Limitations

    Every boolean function has an optimal bounded-error two-reflections algorithm, and span programs are equivalent to quantum query algorithms for boolean-function evaluation.

  • Takeaways & Limitations

    The relationship between classical and quantum query complexities for total functions remains open; the best stated upper bound is D(f)=O(Q(f)^6).

Abstract

from arXiv · show

We show that any boolean function can be evaluated optimally by a quantum query algorithm that alternates a certain fixed, input-independent reflection with a second reflection that coherently queries the input string. Originally introduced for solving the unstructured search problem, this two-reflections structure is therefore a universal feature of quantum algorithms. Our proof goes via the general adversary bound, a semi-definite program (SDP) that lower-bounds the quantum query complexity of a function. By a quantum algorithm for evaluating span programs, this lower bound is known to be tight up to a sub-logarithmic factor. The extra factor comes from converting a continuous-time query algorithm into a discrete-query algorithm. We give a direct and simplified quantum algorithm based on the dual SDP, with a bounded-error query complexity that matches the general adversary bound. Therefore, the general adversary lower bound is tight; it is in fact an SDP for quantum query complexity. This implies that the quantum query complexity of the composition f(g,...,g) of two boolean functions f and g matches the product of the query complexities of f and g, without a logarithmic factor for error reduction. It further shows that span programs are equivalent to quantum query algorithms.

1 Introduction

The paper proves that the general adversary bound exactly characterizes bounded-error quantum query complexity and gives a simpler algorithm alternating two reflections. This yields optimal composition results and establishes an equivalence between span programs and quantum query algorithms.

  • Motivation: The previous span-program approach incurred an unnatural log-over-log-log factor from converting a continuous-time algorithm into a discrete-query algorithm.
  • Main result: The general adversary bound characterizes quantum query complexity for every function f:D→{0,1}.
  • Composition: For boolean functions f and g, the quantum query complexity of their composition matches the product of their individual query complexities without a logarithmic error-reduction factor.
  • Algorithm: The new algorithm alternates the input oracle with a fixed reflection derived from the dual SDP for the general adversary bound.
  • Algorithm: Every boolean function can therefore be evaluated with bounded error using two alternating reflections, generalizing the structure of Grover’s search algorithm.
  • Consequences: Because optimal span-program witness size equals the general adversary bound, span programs and quantum query algorithms are equivalent computational models for evaluating boolean functions.

2 The algorithms

The algorithms are built from a feasible dual-SDP solution and evaluate the function using alternating reflections, with three increasingly simplified variants. Each achieves O(W) query complexity, while the efficient implementation of the fixed reflection remains unclear.

  • Algorithmic framework: A feasible dual-SDP solution with objective value W yields three bounded-error algorithms for evaluating f, each using O(W) queries.The SDP value also equals the witness size of a corresponding canonical span program.
  • Algorithmic framework: The graph construction uses a weighted bipartite graph derived from the SDP solution, with vertices representing F_0, the output vertex 0, and index set I.Edges are weighted by entries of the graph’s biadjacency matrix.
  • Reflection-based algorithm: The unitary U_x alternates a fixed reflection independent of x with a reflection determined by the input, and the latter uses one input-oracle call.The fixed reflection projects onto the eigenvalue-zero subspace of the graph adjacency matrix; the input-dependent reflection excludes vertices associated with complemented input bits.
  • Three algorithm variants: The second and third algorithms choose a random evolution time and apply U_x, with the third outputting 1 exactly when the measured vertex is |0⟩.The stated procedures use T uniformly from ranges scaling as 100W and 10^5W, respectively.

3 Analysis of the algorithms

The analysis studies the spectrum of a product of reflections and transfers graph-based spectral properties to the unitary used by the algorithms. These bounds establish completeness and soundness for the algorithms, including explicit success-probability gaps.

  • Spectral analysis: The analysis studies U_x = (2Π_x − 1)(2∆ − 1) through the spectrum of a product of reflections.The proof uses the eigen-decomposition of products of two projections to analyze the unitary’s invariant subspaces.
  • Spectral analysis: For f(x) = 1, U_x has an eigenvalue-one eigenvector, while for f(x) = 0, its small-phase eigenspaces have bounded overlap with |0⟩.The latter statement is obtained by translating the graph effective spectral-gap lemma into an analogous unitary statement.
  • Correctness: The first algorithm outputs 1 with probability at least 4/5 when f(x) = 1 and at most 2/5 when f(x) = 0.These bounds follow from Lemma 3.4 with the stated phase-estimation parameters.
  • Graph-to-unitary translation: The graph construction links the input-dependent graph G(x) to the fixed graph G and enables the spectral-gap argument underlying Lemma 3.4.The proof uses AG∆ = 0 and the relation AG(x)∆ = T(1 − Π_x), then restricts the accepting eigenvector to G.
  • Correctness: The third algorithm outputs 1 with probability at least 64% for f(x) = 1 and at most 61% for f(x) = 0.The analysis sets τ = ⌈10^5W⌉ and chooses Λ = π − Θ with Θ = 1/(2000W).
Loading 1005.1601v1…