跳转至

On the Compressed-Oracle Technique, and Post-Quantum Security of Proofs of Sequential Work

Notation

Operators and Their Norms

\(\mathcal{H}\): 有限维复 Hilbert 空间,默认 \(\mathcal{H} = \mathbb{C}^d\).

如果算子 \(\rho \in \mathcal{L}(\mathcal{H})\) 的支撑落在 \(\mathcal{H}_o \subset \mathcal{H}\) 上,那么混合态 \(\rho\) 被称为被子空间 \(\mathcal{H}_o\) 支撑.

\(\lVert A \rVert\): \(A \in \mathcal{L}(\mathcal{H}, \mathcal{H'})\) 的算子范数,被 Frobenius 范数约束上界.

Basic Properties: 如果 \(\mathcal{H}\) 可以被分解为若干互相正交子空间的直和 \(\mathcal{H} = \bigoplus_i \mathcal{H}_i\),且 \(A \in \mathcal{L}({\mathcal{H}})\)\(\mathcal{H}_i\) 上的限制为 \(B_i \in \mathcal{L}(\mathcal{H}_i)\),那么

\[ \lVert A \rVert = \max_{1 \leq i \leq m} \lVert B_i \rVert. \]

The Computational and the Fourier Basis

\(\mathcal{Y}\) 为一个基数为 \(M\) 的 Abelian 群,\(\{\ket{y}\}_{y \in \mathcal{Y}}\)\(\mathcal{H} = \mathbb{C}^m\) 的一组(正交)基,其由 \(\mathcal{Y}\) 中元素索引,称为计算基,这里的 \(\mathcal{H}\) 也记作 \(\mathbb{C}[\mathcal{Y}]\). 设 \(\mathcal{Y}\) 的对偶群为 \(\mathcal{\hat{Y}}\),其中的元素为所有群同态 \(\mathcal{Y} \to \{w \in \mathbb{C} \mid \lvert w \rvert = 1\}\),考虑其为加法群,单位元为 \(\hat{0}\). Fourier 基 \(\{\ket{\hat{y}}\}_{\hat{y} \in \mathcal{\hat{Y}}}\) 由如下基本变换定义:

\[ \ket{\hat{y}} = \frac{1}{\sqrt{M}} \sum_{y} \hat{y}(y)^* \ket{y} \quad \text{and} \quad \ket{y} = \frac{1}{\sqrt{M}} \sum_{\hat{y}} \hat{y}(y) \ket{\hat{y}}, \]

其中 \((\cdot)^*\) 为复共轭. 从而有 \(\mathbb{C}[\mathcal{Y}] = \mathbb{C}[\mathcal{\hat{Y}}] = \mathcal{H}\). 设 \(U \in \mathcal{L}(\mathbb{C}[\mathcal{Y}] \otimes \mathbb{C}[\mathcal{Y}])\) 定义为 \(U\ket{y} \ket{y'} \to \ket{y + y'}\ket{y'}, y, y' \in \mathcal{Y}\). 那么对于 \(\hat{y}, \hat{y}' \in \mathcal{\hat{Y}}\),其作用为 \(U\ket{\hat{y}}\ket{\hat{y}'} \to \ket{\hat{y}}\ket{\hat{y}' - \hat{y}}\).

proof
\[\begin{align*} U \ket{\hat{y}} \ket{\hat{y}'} &= U \left( \frac{1}{\sqrt{M}} \sum_{y} \hat{y}(y)^* \ket{y} \right) \left( \frac{1}{\sqrt{M}} \sum_{y'} \hat{y}'(y')^* \ket{y'} \right) \\ &= \frac{1}{M} \sum_{y,y'} \hat{y}(y)^* \hat{y}'(y')^* U \ket{y} \ket{y'} \\ &= \frac{1}{M} \sum_{y,y'} \hat{y}(y)^* \hat{y}'(y')^* \ket{y + y'} \ket{y'}. \end{align*}\]

\(z = y + y'\),则 \(y = z - y'\).

\[\begin{align*} U \ket{\hat{y}} \ket{\hat{y}'} &= \frac{1}{M} \sum_{z,y'} \hat{y}(z - y')^* \hat{y}'(y')^* \ket{z} \ket{y'} \\ &= \frac{1}{M} \sum_{z,y'} \hat{y}(z)^* \hat{y}(-y')^* \hat{y}'(y')^* \ket{z} \ket{y'} \\ &= \frac{1}{M} \sum_{z,y'} \hat{y}(z)^* (\hat{y} - \hat{y}')(y') \ket{z} \ket{y'} \\ &= \frac{1}{\sqrt{M}} \sum_{z} \hat{y}(z)^* \ket{z} \otimes \frac{1}{\sqrt{M}} \sum_{y'} (\hat{y}' - \hat{y})(y')^* \ket{y'} \\ &= \ket{\hat{y}} \ket{\hat{y}' - \hat{y}}. \end{align*}\]

可以通过添加 \(\perp\) 来扩展 \(\mathcal{Y}\)\(\mathcal{\hat{Y}}\). 选择一个归一化的 \(\ket{\perp} \in \mathbb{C}^{M+1}\),其和 \(\mathbb{C}[\mathcal{Y}] = \mathbb{C}[\mathcal{\hat{Y}}]\) 正交. \(\mathbb{C}^{M+1}\) 便可以记作 \(\mathbb{C}[\mathcal{Y} \cup \{\perp\}] = \mathbb{C}[\mathcal{\hat{Y}} \cup \{\perp\}]\).

Functions and Their (Quantum) Representations

对任意固定非空集合 \(\mathcal{X}\)\(H: \mathcal{X} \to \mathcal{Y}\) 构成的集合记作 \(\mathfrak{H}\). \(\mathfrak{\hat{H}} = \{\hat{H}: \mathcal{\hat{X}} \to \mathcal{\hat{Y}}\}\). 考虑 \(\ket{H} = \bigotimes_{x \in \mathcal{X}} \ket{H(x)}\) 作为 \(H\)量子表示,由所有的向量 \(\ket{H}\) 张成的空间 \(\bigotimes_x \mathbb{C}[\mathcal{Y}]\) 记作 \(\mathbb{C}[\mathfrak{H}]\). \(\ket{\hat{H}}\)\(\mathbb{C}[\mathfrak{\hat{H}}]\) 的定义类似,同样有 \(\mathbb{C}[\mathfrak{H}] = \mathbb{C}[\mathfrak{\hat{H}}]\).

\(\mathcal{Y}\) 扩展为 \(\mathcal{\bar{Y}} = \mathcal{Y} \cup \{\perp\}\),函数 \(D: \mathcal{X} \to \mathcal{\bar{Y}}\) 构成的集合记作 \(\mathfrak{D}\). \(\mathfrak{\hat{D}}, \ket{D}\) 等的定义类似,且有 \(\mathbb{C}[\mathfrak{D}] = \mathbb{C}[\mathfrak{\hat{D}}]\).

\(D \in \mathfrak{D}\) 以及 \(\mathbf{x} = (x_1, \ldots, x_k) \in \mathcal{X}^k\),记 \(D(\mathbf{x}) = (D(x_1), \ldots, D(x_k)) \in \mathcal{\bar{Y}}^k\)\(\mathcal{H} \in \mathfrak{H}\) 类似. 而如果 \(\mathbf{x}\) 的分量两两互异,且 \(\mathbf{r} = (r_1, \ldots, r_k) \in \mathcal{\bar{Y}}^k\),则可以定义 \(D[\mathbf{x} \mapsto \mathbf{r}] \in \mathfrak{D}\)

\[ D[\mathbf{x} \mapsto \mathbf{r}] (x_i) = r_i \quad \text{and} \quad D[\mathbf{x} \mapsto \mathbf{r}] (\bar{x}) = D(\bar{x}) \ \forall \bar{x} \not \in \{x_1, \ldots, x_k\}. \]

Zhandry's Compressed Oracle - Refurbished

The Compressed Oracle

核心是考虑所有 \(H \in \mathfrak{H}\) 的一个叠加态 \(\sum_{H} \ket{H}\),然后在 Fourier 基下分析:

\[ \ket{\Pi_0} = \sum_{H} \ket{H} = \bigotimes_x (\sum_{y} \ket{y}) = \sqrt{M} \bigotimes_x \ket{\hat{0}} = \sqrt{M} \ket{\mathbf{\hat{0}}} \in \mathbb{C}[\mathfrak{H}] \]

设谕示机请求中包含的酉变换在计算基下为

\[ \mathsf{O} = \ket{x}\ket{y} \otimes \ket{H} \mapsto \ket{x} \ket{y + H(x)} \otimes \ket{H}, \]

那么在 Fourier 基下,我们有

\[ \mathsf{O} = \ket{x}\ket{\hat{y}} \otimes \ket{\hat{H}} \mapsto \ket{x} \ket{\hat{y}} \otimes \mathsf{O}_{x \hat{y}} \ket{\hat{H}} = \ket{x} \ket{\hat{y}} \otimes \ket{\hat{H} - \hat{y} \cdot \delta_x}. \]

等式给出了 \(\mathsf{O}_{x \hat{y}}\) 的定义. \(\mathsf{O}_{x \hat{y}}\) 只对 \(x\) 寄存器作用,并且有 \(\mathsf{O}_{x \hat{y}} \mathbf{O}_{x \hat{y}'} = \mathsf{O}_{x, \hat{y} + \hat{y}'}\),所以 \(\mathbf{O}_{x \hat{y}}\)\(\mathbf{O}_{x' \hat{y}'}\) 均可交换. 因而 \(q\) 次查询后,谕示机内部的状态由 \(\ket{\hat{H}} = \ket{\hat{y}_1 \delta_{x_1} + \cdots + \hat{y}_q \delta_{x_q}}\) 支撑.

实际的压缩谕示机通过对所有的寄存器 \(x\) 施加如下的同构获得:

\[ \op{Comp}_x = \ket{\perp} \bra{\hat{0}} + \sum_{\hat{z} \neq \hat{0}} \ket{\hat{z}} \bra{\hat{z}}: \mathbb{C}[\mathcal{Y}] \to \mathbb{C}[\mathcal{\bar{Y}}], \ket{\hat{y}} \to \begin{cases} \ket{\perp}, \text{ if } \hat{y} = \hat{0} \\ \ket{\hat{y}}, \text{ if } \hat{y} \neq \hat{0} \end{cases} \]

整个压缩算子 \(\op{Comp} = \bigotimes_x \op{Comp}_x: \mathbb{C}[\mathfrak{H}] \to \mathbb{C}[\mathfrak{D}]\)\(\ket{\Pi_0}\) 映射为

\[ \op{Comp} \ket{\Pi_0} = \left(\bigotimes_x \op{Comp}_x \right) \left(\bigotimes_x \ket{\hat{0}} \right) = \bigotimes_x \op{Comp}_x \ket{\hat{0}} = \bigotimes_x \ket{\perp} = \ket{\mathbf{\perp}}. \]

推而广之,对任意的 \(\hat{H} \in \hat{\mathfrak{H}}\)\(\op{Comp} \ket{\hat{H}} = \ket{\hat{D}}\),其中 \(\hat{D} \in \hat{\mathfrak{D}}\),若 \(\hat{H}(x) \neq 0\)\(\hat{D}(x) = \hat{H}(x)\),否则 \(\hat{D}(x) = \perp\).

Linking the Compressed and the Original Oracle

Lemma

考虑任意(归一化)的量子态 \(\ket{\Pi} \in \mathbb{C}[\mathfrak{H}]\)\(\ket{\Delta} = \op{Comp} \ket{\Pi} \in \mathbb{C}[\mathfrak{D}]\) 为对应的“压缩数据库”. 设 \(\mathbf{x} = (x_1, \ldots, x_l)\) 包含的都是两两互异的 \(x_i \in \mathcal{X}\)\(\mathbf{y} = (y_1, \ldots, y_l) \in \mathcal{Y}^l\)\(P_\mathbf{x} = \ket{y_1}\bra{y_1} \otimes \cdots \otimes \ket{y_l}\bra{y_l}\),其中 \(\ket{y_i}\bra{y_i}\) 作用在寄存器 \(x_i\) 上,那么

\[ \lVert P_\mathbf{x} \ket{\Pi} \rVert \leq \lVert P_\mathbf{x} \ket{\Delta} \rVert + \sqrt{\frac{l}{M}}. \]
Proof

\(\op{Comp}_{\mathbf{x}}\) 为作用在寄存器 \(x_1, \ldots, x_l\) 上的压缩算子,\(I_\bar{\mathbf{x}}\) 为作用在其余寄存器上的恒等算子. 有

\[\begin{align*} \lVert P_\mathbf{x} \ket{\Pi} \rVert - \lVert P_\mathbf{x} \op{Comp} \ket{\Pi} \rVert & = \lVert P_\mathbf{x} \ket{\Pi} \rVert - \lVert P_\mathbf{x} \op{Comp}_{\mathbf{x}} \ket{\Pi} \rVert \\ & \leq \lVert (P_\mathbf{x} - P_\mathbf{x} \op{Comp}_{\mathbf{x}}) \ket{\Pi} \rVert \\ & \leq \lVert (P_\mathbf{x} - P_\mathbf{x} \op{Comp}_{\mathbf{x}}) \otimes I_\bar{\mathbf{x}} \rVert \\ & = \lVert (P_\mathbf{x} - P_\mathbf{x} \op{Comp}_{\mathbf{x}}) \rVert \end{align*}\]

考虑 Fourier 下的 \(P_\mathbf{x}\),因为

\[\begin{align*} \ket{y_i} \bra{y_i} &= \left(\frac{1}{\sqrt{M}} \sum_{\hat{z} \in \mathcal{\hat{Y}}} \hat{z}(y_i) \ket{\hat{z}} \right) \left(\frac{1}{\sqrt{M}} \sum_{\hat{y} \in \mathcal{\hat{Y}}} \hat{y}(y_i)^* \bra{\hat{y}} \right) \\ &= \frac{1}{M} \sum_{\hat{z}, \hat{y} \in \mathcal{\hat{Y}}} \hat{z}(y_i) \hat{y}(y_i)^* \ket{\hat{z}} \bra{\hat{y}}, \end{align*}\]

引入相位因子 \(\omega_{\hat{z}/\hat{y}} (y_i) = \hat{z}(y_i) \hat{y}(y_i)^*\),便有

\[ P_\mathbf{x} = \bigotimes_i \frac{1}{M} \sum_{\hat{z}, \hat{y} \in \mathcal{\hat{Y}}} \omega_{\hat{z}/\hat{y}}(y_i) \ket{\hat{z}} \bra{\hat{y}}. \]

所以有

\[ P_\mathbf{x} \op{Comp}_\mathbf{x} = \bigotimes_i \frac{1}{M} \sum_{\hat{z}, \hat{0} \neq \hat{y} \in \mathcal{\hat{Y}}} \omega_{\hat{z}/\hat{y}}(y_i) \ket{\hat{z}} \bra{\hat{y}}. \]

张量积展开得到

\[\begin{align*} P_\mathbf{x} - P_\mathbf{x} \op{Comp}_\mathbf{x} & = \frac{1}{M^l} \sum_{\hat{y}_1, \ldots, \hat{z}_l \in \mathcal{\hat{Y}} \text{ and } \exists i: \hat{y}_i = \hat{0}} \omega_{\hat{z}_i/\hat{y}_i}(y_i) \ket{\hat{z}_i} \bra{\hat{y}_i} \\ & = \frac{1}{M^l} \sum_{\hat{\mathbf{y}}, \hat{\mathbf{z}} \text{ and } \exists i: \hat{y}_i = \hat{0}} \omega_{\hat{\mathbf{z}}/\hat{\mathbf{y}}} \ket{\hat{\mathbf{z}}} \bra{\hat{\mathbf{y}}}. \end{align*}\]

其中 \(\hat{\mathbf{y}} = (\hat{y}_1, \ldots, \hat{y}_l), \hat{\mathbf{z}} = (\hat{z}_1, \ldots, \hat{z}_l) \in \mathcal{\hat{Y}}^l\). \(\omega_{\hat{\mathbf{z}}/\hat{\mathbf{y}}} = \prod_i \omega_{\hat{z}_i/\hat{y}_i}\).

使用 Frobenius 范数约束算子范数,有

\[\begin{align*} \lVert P_\mathbf{x} - P_\mathbf{x} \op{Comp}_\mathbf{x} \rVert^2 & \leq \sum_{\hat{\mathbf{y}}, \hat{\mathbf{z}}} \lvert \bra{\hat{\mathbf{z}} (P_\mathbf{x} - P_\mathbf{x} \op{Comp}_\mathbf{x})} \ket{\hat{\mathbf{y}}} \rvert^2 \\ & = \frac{1}{M^{2l}} \sum_{\hat{\mathbf{y}}, \hat{\mathbf{z}} \text{ and } \exists i: \hat{y}_i = \hat{0}} \lvert \omega_{\hat{\mathbf{z}}/\hat{\mathbf{y}}} \rvert^2 \\ & \leq \frac{1}{M^{2l}} l M^{2l-1} = \frac{l}{M}. \end{align*}\]

最后一行是简单的计数估计,为 \(0\) 的位置有 \(l\) 种,并且对应的有 \(M^{l-1}\) 种选择. 尽管有所重复,但实际上是放的更松了.

Corollary

设关系 \(R \subset \mathcal{X}^l \times \mathbf{Y}^l\)\(\mathcal{A}\) 为输出 \(\mathbf{x} \in \mathcal{X}^l\)\(\mathbf{y} \in \mathcal{Y}^l\) 的量子谕示机算法. 设 \(p\) 为在 \(\mathcal{A}\) 与用均匀随机函数 \(H\) 初始化的标准随机谕示机交互的条件下满足 \(\mathbf{y} = H(\mathbf{x})\)\((\mathbf{x}, \mathbf{y}) \in R\) 的概率,而 \(p'\) 为在 \(\mathcal{A}\) 与压缩谕示机交互且 \(D\) 为测量压缩谕示机内部状态获得的条件下满足 \(\mathbf{y} = D(\mathbf{x})\)\((\mathbf{x}, \mathbf{y}) \in R\) 的概率,则有

\[ \sqrt{p} \leq \sqrt{p'} + \sqrt{\frac{l}{M}}. \]
Proof

考虑 \(\mathcal{A}\) 和纯化谕示机交互时的执行. 假定测量 \(\mathbf{x}, \mathbf{y}\) 后,\(\mathcal{A}\) 还会在计算基下测量其内部状态得到字符串 \(w\),并且输出. 设 \(q_{\mathbf{x}, \mathbf{y}, w}\)\(\mathcal{A}\) 输出三元组 \(\mathbf{x}, \mathbf{y}, w\) 的概率, \(p_{\mathbf{x}, \mathbf{y}, w}\) 为给定 \(\mathcal{A}\) 的输出的条件下 \(\mathbf{y} = H(\mathbf{x})\) 的概率,\(p_{\mathbf{x}, \mathbf{y}, w}'\) 为同样条件下 \(\mathbf{y} = D(\mathbf{x})\) 的概率,首先观察到:

\[\begin{align*} p &= \sum_{\mathbf{x}, \mathbf{y}, w \text{ and } (x, y) \in R} q_{\mathbf{x}, \mathbf{y}, w} \ p_{\mathbf{x}, \mathbf{y}, w}, \\ p' &= \sum_{\mathbf{x}, \mathbf{y}, w \text{ and } (x, y) \in R} q_{\mathbf{x}, \mathbf{y}, w} \ p_{\mathbf{x}, \mathbf{y}, w}'. \end{align*}\]

\(p_{\mathbf{x}, \mathbf{y}, w} = \lVert P_\mathbf{x} \ket{\Pi} \rVert^2\)\(\ket{\Pi}\) 是根据 \(\mathbf{x}, \mathbf{y}\)\(w\) 后选择的纯化谕示机的内部状态. 同理 \(p_{\mathbf{x}, \mathbf{y}, w}' = \lVert P_\mathbf{x} \op{Comp} \ket{\Pi} \rVert^2\). 依据引理再平方有

\[ p_{\mathbf{x}, \mathbf{y}, w} \leq (\sqrt{p_{\mathbf{x}, \mathbf{y}, w}'}) + \sqrt{\frac{l}{M}})^2 = p_{\mathbf{x}, \mathbf{y}, w}' + 2\sqrt{p_{\mathbf{x}, \mathbf{y}, w}'}\sqrt{\frac{l}{M}} + \frac{l}{M}. \]

\(Q = \sum_{\mathbf{x}, \mathbf{y}, w \text{ and } (x, y) \in R} q_{\mathbf{x}, \mathbf{y}, w}\),两侧用 \(q_{\mathbf{x}, \mathbf{y}, w}\) 加权平均并应用 Jensen 不等式可以得到

\[\begin{align*} p & \leq p' + 2 \sum_{\mathbf{x}, \mathbf{y}, w \text{ and } (x, y) \in R} q_{\mathbf{x}, \mathbf{y}, w} \sqrt{p_{\mathbf{x}, \mathbf{y}, w}'} \sqrt{ \frac{l}{M} } + Q \cdot \frac{l}{M} \\ & \leq p' + 2 \sqrt{ \frac{l}{M} } Q \cdot \sqrt{ \frac{\sum_{\mathbf{x}, \mathbf{y}, w \text{ and } (x, y) \in R} q_{\mathbf{x}, \mathbf{y}, w} p_{\mathbf{x}, \mathbf{y}, w}'}{Q} } + \frac{l}{M} \\ & \leq p' + 2 \sqrt{ \frac{l}{M} } \cdot \sqrt{Q \cdot p'} + \frac{l}{M} \\ & \leq (\sqrt{p'} + \sqrt{\frac{l}{M}})^2. \end{align*}\]

从而有

\[ \sqrt{p} \leq \sqrt{p'} + \sqrt{\frac{l}{M}}. \]

Working Out the Transition Matrix

扩展 \(\op{Comp}_x\) 的定义域到 \(\mathbb{C}[\mathcal{\bar{Y}}]\),定义 \(\op{Comp}_x \ket{\perp} = \ket{\hat{0}}\).

定义:

\[ \mathsf{cO} = \op{Comp} \circ \mathsf{O} \circ \op{Comp}^\dagger \in \mathcal{L}(\mathbb{C}[\mathcal{X}] \otimes \mathbb{C}[\mathcal{Y}] \otimes \mathbb{C}[\mathfrak{D}]) \]

对任意 \(D \in \mathfrak{D}\),其将 \(\ket{x} \ket{\hat{y}} \otimes \ket{D}\) 映射为 \(\ket{x} \ket{\hat{y}} \otimes \mathsf{cO}_{x\hat{y}} \ket{D}\),其中 \(\mathsf{cO}_{x\hat{y}} = \op{Comp}_x \circ \mathsf{O}_{x\hat{y}} \circ \op{Comp}_x^\dagger \in \mathcal{L}(\mathbb{C}[\mathcal{\bar{Y}}])\) 只作用在寄存器 \(x\) 上.

有交换图:

Lemma

对任意 \(\hat{y} \neq \hat{0}\)\(\mathsf{cO}_{x \hat{y}}\) 在计算基下的矩阵表示由下表给出,也即对任意 \(r, u \in \mathcal{\bar{Y}}\)\(\bra{u} \mathsf{cO}_{x \hat{y}} \ket{r} = \gamma^{\hat{y}}_{u, r}\). 此外 \(\mathsf{cO}_{x, \hat{0}} = I\).

\(\perp\) \(r \in \mathcal{Y}\)
\(\perp\) \(\gamma^{\hat{y}}_{\perp, \perp} = 0\) \(\gamma^{\hat{y}}_{\perp, r} = \frac{\hat{y}(r)}{\sqrt{M}}\)
\(u \in \mathcal{Y}\) \(\gamma^{\hat{y}}_{u, \perp} = \frac{\hat{y}(u)}{\sqrt{M}}\) \(\gamma^{\hat{y}}_{u, r} = \begin{cases} (1 - \frac{2}{M})\hat{y}(u) + \frac{1}{M} & \text{if } u = r \\ \frac{1 - \hat{y}(r) - \hat{y}(u)}{M} & \text{if } u \neq r \end{cases}\)
Proof

对任意 \(r \neq \perp\) 以及 \(\hat{y} \neq \hat{0}\)

\[ \sqrt{M}\ket{r} = \sum_{\hat{r}} \hat{r}(r) \ket{\hat{r}} = \ket{\hat{0}} + \sum_{\hat{r} \neq \hat{0}} \hat{r}(r) \ket{\hat{r}}, \]

经过 \(\op{Comp}^\dagger\) 的作用后映射为 \(\ket{\perp} + \sum_{\hat{r} \neq \hat{0}} \hat{r}(r) \ket{\hat{r}}\). 再经 \(\mathsf{O}_{x\hat{y}}\) 的作用后映射为

\[\begin{align*} & \mapsto \ket{\perp} + \sum_{\hat{r} \neq \hat{0}} \hat{r}(r) \ket{\hat{r} - \hat{y}} \\ & = \ket{\perp} - \ket{-\hat{y}} + \sum_{\hat{r}} \hat{r}(r) \ket{\hat{r} - \hat{y}} \\ & = \ket{\perp} - \ket{-\hat{y}} + \hat{y}(r) \sum_{\hat{r}} \hat{r}(r) \ket{\hat{r}} \\ & = \ket{\perp} - \ket{-\hat{y}} + \hat{y}(r) \ket{\hat{0}} + \hat{y}(r) \sum_{\hat{r} \neq \hat{0}} \hat{r}(r) \ket{\hat{r}}. \end{align*}\]

继续经过 \(\op{Comp}\) 映射

\[\begin{align*} & \mapsto \ket{\hat{0}} - \ket{-\hat{y}} + \hat{y}(r)\ket{\perp} + \hat{y}(r) \sum_{\hat{r} \neq \hat{0}} \hat{r}(r) \ket{\hat{r}} \\ & = \ket{\hat{0}} - \ket{-\hat{y}} + \hat{y}(r)\ket{\perp} - \hat{y}(r)\ket{\hat{0}} + \hat{y}(r) \sum_{\hat{r}} \hat{r}(r) \ket{\hat{r}} \\ & = \frac{1}{\sqrt{M}} \sum_u \ket{u} - \frac{1}{\sqrt{M}} \sum_u \hat{y}(u) \ket{u} + \hat{y}(r)\ket{\perp} - \frac{\hat{y}(r)}{\sqrt{M}} \sum_u \ket{u} + \sqrt{M} \hat{y}(r) \ket{r}. \end{align*}\]

所以有

\[ \mathsf{cO}_{x\hat{y}} \ket{r} = \frac{1}{M} \sum_u \ket{u} - \frac{1}{M} \sum_u \hat{y}(u) \ket{u} + \frac{\hat{y}(r)}{\sqrt{M}}\ket{\perp} - \frac{\hat{y}(r)}{M} \sum_u \ket{u} + \hat{y}(r) \ket{r}. \]

由此便可以得出 \(r \neq \perp\) 对应的系数 \(\gamma^{\hat{y}}_{u, r}\). 而对于 \(r = \perp\) 的情况,有

\[ \ket{\perp} \mapsto \ket{\hat{0}} \mapsto \ket{-\hat{y}} \mapsto \ket{-\hat{y}} = \frac{1}{\sqrt{M}} \sum_u \hat{y}(u) \ket{u}, \]

也就能得到 \(r = \perp\) 对应的系数 \(\gamma^{\hat{y}}_{u, \perp}\). 利用 \(\mathsf{O}_{x, \hat{0}} = I\) 便可以得到 \(\hat{y} = \hat{0}\) 的情况.

The Parallel-Query (Compressed) Oracle

\(k\)-并行查询定义为

\[ \mathsf{O}^k: \ \ket{\mathbf{x}}\ket{\mathbf{y}} \otimes \ket{H} \mapsto \ket{\mathbf{x}} \ket{\mathbf{y} + H(\mathbf{x})} \otimes \ket{H}. \]

其中 \(\mathbf{x} = (x_1, x_2, \ldots, x_k) \in \mathcal{X}^k\)\(\mathbf{y} = (y_1, y_2, \ldots, y_k) \in \mathcal{Y}^k\). \(\mathsf{cO}^k = \op{Comp} \circ \mathsf{O}^k \circ \op{Comp}^\dagger\),作用为

\[ \mathsf{cO}^k: \ \ket{\mathbf{x}}\ket{\mathbf{\hat{y}}} \otimes \ket{\Delta} \mapsto \ket{\mathbf{x}} \ket{\mathbf{\hat{y}} \otimes \mathsf{cO}_{\mathbf{x} \mathbf{\hat{y}}}} \otimes \ket{\Delta}. \]

其中 \(\mathbf{cO}_{\mathbf{x} \mathbf{\hat{y}}} = \mathsf{cO}_{x_1 \hat{y}_1} \cdots \mathsf{cO}_{x_k \hat{y}_k}\).

A Framework for Proving Quantum Query Lower Bounds

Setting Up the Framework

数据库性质:数据库 \(\mathfrak{D}\) 的一个子集 \(\mathsf{P} \subset \mathfrak{D}\).

\(\mathsf{P}(D)\) 也用来表示 \(D \in \mathsf{P}\),即 \(D\) 满足 \(\mathsf{P}\)\(\mathsf{P}\) 也被重载表示为投影算子 \(\sum_{D \in \mathsf{P}} \ket{D}\bra{D} \in \mathcal{L}(\mathbb{C}[\mathfrak{D}])\).

  • \(\mathsf{PRMG} = \{D \mid \exists x: \ D(x) = 0\}\);

  • \(\mathsf{CL} = \{D \mid \exists x, x': \ D(x) = D(x') \neq \perp\}\);

  • \(\mathsf{CHN}^q = \{D \mid \exists x_0, x_1, \ldots, x_q \in \mathcal{X}: \ D(x_{i-1}) \lhd x_i \ \forall i\}\).

\(\lhd\) 表示任意关系.

对任意元素互异的 \(\mathbf{x} = (x_1, x_2, \ldots, x_k)\),以及任意 \(D: \mathcal{X} \to \mathcal{\bar{Y}}\),定义

\[ D\vert^\mathbf{x} = \{D[\mathbf{x} \mapsto \mathbf{r}] \mid \mathbf{r} \in \mathcal{\bar{Y}}^k \} \subset \mathfrak{D}, \]

其是所有除 \(\mathbf{x}\) 外其余位置的值与 \(D\) 相同的函数的集合. 对于任意数据库性质 \(\mathsf{P} \subset \mathfrak{D}\),定义

\[ \mathsf{P}\vert_{D\vert^\mathbf{x}} = \mathsf{P} \cap D\vert^\mathbf{x} \]

\(\mathsf{P}\)\(D\vert^\mathbf{x}\) 上的限制.

对于固定的 \(\mathbf{x}\)\(D\),因为存在 \(\mathbf{r} \mapsto D[\mathbf{x} \mapsto \mathbf{r}]\) 的映射,所以可以将 \(\mathcal{\bar{Y}}^k\)\(D\vert^\mathbf{x}\) 等同,此时性质 \(\mathsf{P}\vert_{D\vert^\mathbf{x}}\) 可以视为 \(\mathcal{\bar{Y}}^k\) 上的性质,即 \(\{\mathbf{r} \in \mathcal{\bar{Y}}^k \mid D[\mathbf{x} \mapsto \mathbf{r}] \in \mathsf{P}\}\).

因此,也不再区分以下两个投影算子,统一记作 \(\mathsf{P}\vert_{D\vert^\mathbf{x}}\)

\[\begin{gather*} \sum_{D' \in \mathsf{P}\vert_{D\vert^\mathbf{x}}} \ket{D'}\bra{D'} \in \mathcal{L}(\mathbb{C}[D\vert^\mathbf{x}]) \subset \mathcal{L}(\mathbb{C}[\mathfrak{D}]) \\ \sum_{\mathbf{r} \in \mathcal{\bar{Y}}^k \text{ and } D[\mathbf{x} \mapsto \mathbf{r}] \in \mathsf{P}} \ket{\mathbf{r}}\bra{\mathbf{r}} \in \mathcal{L}(\mathbb{C}[\mathcal{\bar{Y}}^k]) \end{gather*}\]

Definition

(量子传输能力) 设 \(\mathsf{P}, \mathsf{P}'\) 是两个数据库性质,那么定义(阶为 \(k\) 的)量子传输能力:

\[ \dbrack{\mathsf{P} \xrightarrow{k} \mathsf{P}'} = \max_{\mathbf{x}, \mathbf{\hat{y}}, D} \lVert \mathsf{P}'\vert_{D\vert^\mathbf{x}} \ \mathsf{cO}_{\mathbf{x} \mathbf{\hat{y}}} \ \mathsf{P}\vert_{D\vert^\mathbf{x}} \rVert. \]

\(q\) 步的量子转移能力则定义为

\[ \dbrack{\mathsf{P} \xRightarrow{k, q} \mathsf{P}'} = \sup_{U_1, \ldots, U_{q-1}} \lVert \mathsf{P}' \mathsf{cO}^k U_{q-1} \mathsf{cO}^k \cdots U_1 \mathsf{cO}^k \mathsf{P} \rVert. \]

其中 \(U_i\) 均作用于 \(\mathbb{C}[\mathcal{X}] \otimes \mathbb{C}[\mathcal{Y}] \otimes \mathbb{C}^d\)\(d\) 是任意的正整数.

Lemma

对任意性质序列 \(\mathsf{P}_0, \mathsf{P}_1, \ldots, \mathsf{P}_q\),有

\[ \dbrack{\neg \mathsf{P}_0 \xRightarrow{k, q} \mathsf{P}_q} \leq \sum_{s=1}^q \dbrack{\neg \mathsf{P}_{s-1} \xrightarrow{k} \mathsf{P}_s}. \]
Proof