量子搜索算法基础: Quantum Amplititude Amplification

1. Introduction

You have likely heard that one of the many advantages a quantum computer has over a classical computer is its superior speed searching databases. Grover’s algorithm demonstrates this capability. This algorithm can speed up an unstructured search problem quadratically, but its uses extend beyond that; it can serve as a general trick or subroutine to obtain quadratic run time improvements for a variety of other algorithms. This is called the amplitude amplification trick.

两步走：1. 反转目标状态的振幅；2. 根据平均值再翻转。

2. Unstructured Search

Suppose you are given a large list of N items. Among these items there is one item with a unique property that we wish to locate; we will call this one the winner w. Think of each item in the list as a box of a particular color. Say all items in the list are gray except the winner w, which is purple.

To find the purple box – the marked item – using classical computation, one would have to check on average N/2 of these boxes, and in the worst case, all N of them. On a quantum computer, however, we can find the marked item in roughly \sqrt{N} steps with Grover’s amplitude amplification trick. A quadratic speedup is indeed a substantial time-saver for finding marked items in long lists. Additionally, the algorithm does not use the list’s internal structure, which makes it generic; this is why it immediately provides a quadratic quantum speed-up for many classical problems.

3. Creating an Oracle

For the examples in this textbook, our ‘database’ is comprised of all the possible computational basis states our qubits can be in. For example, if we have 3 qubits, our list is the states |000\rangle, |001\rangle, \dots |111\rangle (i.e the states |0\rangle \rightarrow |7\rangle).

Grover’s algorithm solves oracles that add a negative phase to the solution states. I.e. for any state $|x⟩$ in the computational basis:

$$U_{\omega }|x⟩=\{\begin{array}{r}|x⟩\text{if}x\ne \omega \\ -|x⟩\text{if}x=\omega \end{array}$$
This oracle will be a diagonal matrix, where the entry that correspond to the marked item will have a negative phase. For example, if we have three qubits and \omega = \text{101}, our oracle will have the matrix:

U ω = [ 1 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 − 1 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 1 ] ← ω = 101 U_\omega = \begin{bmatrix} 1 & 0 & 0 & 0 & 0 & 0 & 0 & 0 \\ 0 & 1 & 0 & 0 & 0 & 0 & 0 & 0 \\ 0 & 0 & 1 & 0 & 0 & 0 & 0 & 0 \\ 0 & 0 & 0 & 1 & 0 & 0 & 0 & 0 \\ 0 & 0 & 0 & 0 & 1 & 0 & 0 & 0 \\ 0 & 0 & 0 & 0 & 0 & -1 & 0 & 0 \\ 0 & 0 & 0 & 0 & 0 & 0 & 1 & 0 \\ 0 & 0 & 0 & 0 & 0 & 0 & 0 & 1 \\ \end{bmatrix} \begin{aligned} \\ \\ \\ \\ \\ \\ \leftarrow \omega = \text{101}\\ \\ \\ \\ \end{aligned} Uω​=⎣⎢⎢⎢⎢⎢⎢⎢⎢⎢⎢⎡​10000000​