卷積

by 不卷王

卷積

by 不卷王

今天沒有偷簡報環節了吧(?

今天會講到

  • FFT
  • NTT
  • 迭代式FFT/NTT
  • 任意模數FFT
  • FWHT

 

講師不像你們那麼卷,我最近才初學一點點卷積

講很爛很不嚴謹請見諒

然後因為我進度燒雞,所以沒有放例題,這堂課就是來看我假裝會卷積

如果覺得我的code太醜的話,我只能說學長對不起

野獸 先備

矩陣

$$\begin{bmatrix}a&b\\c&d\end{bmatrix}\times \begin{bmatrix}e&f\\g&h\end{bmatrix}=\begin{bmatrix}ae+bg&af+bh\\ce+dg&cf+dh\end{bmatrix}$$

複數

$$i^2=-1$$

$$(a+bi)\times (c+di)=ac+(ad+bc)i-bd$$

歐拉公式

棣美弗公式

  • \( (cos\theta + i\ sin\theta)n = cos(n\theta) + i\ sin(n\theta)\)
  • 用歐拉公式來看的話根本是廢話 : \((e^{i\theta})^n=e^{in\theta}\)

單位根

  • \((\omega_{n})^n=1\)
  • \(x^n=1\)的根分別為\(1,\omega_{n},\cdots,\omega_{n}^{n-1}\)

$$\omega_6^3$$

$$\omega_6^2$$

$$\omega_6$$

$$\omega_6^0$$

$$\omega_6^4$$

$$\omega_6^4$$

$$\omega_6^5$$

.

原根

其實我不會ㄟ

$$(M)$$

原根

質數\(P\)以下的原根\(g\)的\(1\)次方到\(P-1\)次方,恰好對到\(1\sim P-1\)的每個數

 

模\(7\)之下,\(3\)是,\(2\)不是

$$3^1\equiv 3, 3^2 \equiv 2,3^3 \equiv 6,3^4\equiv 4,3^5\equiv5,3^6=1$$

$$2^1\equiv 2, 2^2 \equiv 4,2^3 \equiv 1$$

約定

  • 多項式\(A\)、\(B\)、\(C=A\times B\)皆為整係數多項式
  • \(C\)有\(n\)項(最高次數為\(n-1\))
  • \(A,B\)的最高次數\(<\frac{n}{2}\),自動補0在後面到\(n項\)
  • \(n\)為\(2^k\),\(k\)為整數

約定

  • 多項式\(A\)、\(B\)、\(C=A\times B\)皆為整係數多項式
  • \(C\)有\(n\)項(最高次數為\(n-1\))
  • \(A,B\)的最高次數\(<\frac{n}{2}\),自動補0在後面到\(n項\)
  • \(n\)為\(2^k\),\(k\)為整數

$$a_0 x_0 + a_1 x_1+\cdots+a_{\frac{n}{2}-1}x_{\frac{n}{2}-1}$$

 

約定

  • 多項式\(A\)、\(B\)、\(C=A\times B\)皆為整係數多項式
  • \(C\)有\(n\)項(最高次數為\(n-1\))
  • \(A,B\)的最高次數\(<\frac{n}{2}\),自動補0在後面到\(n項\)
  • \(n\)為\(2^k\),\(k\)為整數

$$a_0 x_0 + a_1 x_1+\cdots+a_{\frac{n}{2}-1}x_{\frac{n}{2}-1}+0+0+\cdots+0$$

 

$$\frac{n}{2}個$$

$$a_0 x_0 + a_1 x_1+\cdots+a_{\frac{n}{2}-1}x_{\frac{n}{2}-1}$$

 

FFT

FFT想要解決什麼?

  • 快速地把兩個多項式乘起來
  • 直接乘起來\(O(n^2)\)

FFT想要解決什麼?

  • 快速地把兩個多項式乘起來
  • 直接乘起來\(O(n^2)\)
  • 如果帶入固定的\(n\)個\(x\)取樣的話,乘法就變\(O(n)\),

FFT想要解決什麼?

  • 快速地把兩個多項式乘起來
  • 直接乘起來\(O(n^2)\)
  • 如果用取樣的話,乘法就變\(O(n)\)
  • 要怎麼快速把\(n\)個點帶入? 暴力\(O(n^2)\)
  • 要怎麼轉換回來? 拉格朗日插值要\(O(n^2)\)

是否可以找到一組有特別性質的\(x\)來加速帶入,與轉回多項式?

是否可以找到一組有特別性質的\(x\)來加速帶入,與轉回多項式?

可以!用\(\omega_n^{0\sim n-1}\)(在這裡指的是\(x^n=1\)的\(n\)個解)試試看

\begin{bmatrix} \omega^{0 \cdot 0} & \omega^{0 \cdot 1} & \omega^{0 \cdot 2} & \omega^{0 \cdot 3} & \omega^{0 \cdot 4} & \omega^{0 \cdot 5} & \omega^{0 \cdot 6} & \omega^{0 \cdot 7} \\ \omega^{1 \cdot 0} & \omega^{1 \cdot 1} & \omega^{1 \cdot 2} & \omega^{1 \cdot 3} & \omega^{1 \cdot 4} & \omega^{1 \cdot 5} & \omega^{1 \cdot 6} & \omega^{1 \cdot 7} \\ \omega^{2 \cdot 0} & \omega^{2 \cdot 1} & \omega^{2 \cdot 2} & \omega^{2 \cdot 3} & \omega^{2 \cdot 4} & \omega^{2 \cdot 5} & \omega^{2 \cdot 6} & \omega^{2 \cdot 7} \\ \omega^{3 \cdot 0} & \omega^{3 \cdot 1} & \omega^{3 \cdot 2} & \omega^{3 \cdot 3} & \omega^{3 \cdot 4} & \omega^{3 \cdot 5} & \omega^{3 \cdot 6} & \omega^{3 \cdot 7} \\ \omega^{4 \cdot 0} & \omega^{4 \cdot 1} & \omega^{4 \cdot 2} & \omega^{4 \cdot 3} & \omega^{4 \cdot 4} & \omega^{4 \cdot 5} & \omega^{4 \cdot 6} & \omega^{4 \cdot 7} \\ \omega^{5 \cdot 0} & \omega^{5 \cdot 1} & \omega^{5 \cdot 2} & \omega^{5 \cdot 3} & \omega^{5 \cdot 4} & \omega^{5 \cdot 5} & \omega^{5 \cdot 6} & \omega^{5 \cdot 7} \\ \omega^{6 \cdot 0} & \omega^{6 \cdot 1} & \omega^{6 \cdot 2} & \omega^{6 \cdot 3} & \omega^{6 \cdot 4} & \omega^{6 \cdot 5} & \omega^{6 \cdot 6} & \omega^{6 \cdot 7} \\ \omega^{7 \cdot 0} & \omega^{7 \cdot 1} & \omega^{7 \cdot 2} & \omega^{7 \cdot 3} & \omega^{7 \cdot 4} & \omega^{7 \cdot 5} & \omega^{7 \cdot 6} & \omega^{7 \cdot 7} \\ \end{bmatrix}
\begin{bmatrix} a_0\\ a_1\\ a_2\\ a_3\\ a_4\\ a_5\\ a_6\\ a_7 \end{bmatrix}
= \begin{bmatrix} y_0\\ y_1\\ y_2\\ y_3\\ y_4\\ y_5\\ y_6\\ y_7 \end{bmatrix}

這裡用\(\omega\)代替\(\omega_8\)

\begin{bmatrix} \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} \\ \omega^{0} & \omega^{1} & \omega^{2} & \omega^{3} & \omega^{4} & \omega^{5} & \omega^{6} & \omega^{7} \\ \omega^{0} & \omega^{2} & \omega^{4} & \omega^{6} & \omega^{0} & \omega^{2} & \omega^{4} & \omega^{6} \\ \omega^{0} & \omega^{3} & \omega^{6} & \omega^{1} & \omega^{4} & \omega^{7} & \omega^{2} & \omega^{5} \\ \omega^{0} & \omega^{4} & \omega^{0} & \omega^{4} & \omega^{0} & \omega^{4} & \omega^{0} & \omega^{4} \\ \omega^{0} & \omega^{5} & \omega^{2} & \omega^{7} & \omega^{4} & \omega^{1} & \omega^{6} & \omega^{3} \\ \omega^{0} & \omega^{6} & \omega^{4} & \omega^{2} & \omega^{0} & \omega^{6} & \omega^{4} & \omega^{2} \\ \omega^{0} & \omega^{7} & \omega^{6} & \omega^{5} & \omega^{4} & \omega^{3} & \omega^{2} & \omega^{1} \end{bmatrix}

這裡用\(\omega\)代替\(\omega_8\)

\begin{bmatrix} a_0\\ a_1\\ a_2\\ a_3\\ a_4\\ a_5\\ a_6\\ a_7 \end{bmatrix}
= \begin{bmatrix} y_0\\ y_1\\ y_2\\ y_3\\ y_4\\ y_5\\ y_6\\ y_7 \end{bmatrix}
\begin{bmatrix} \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} \\ \omega^{0} & \omega^{1} & \omega^{2} & \omega^{3} & \omega^{4} & \omega^{5} & \omega^{6} & \omega^{7} \\ \omega^{0} & \omega^{2} & \omega^{4} & \omega^{6} & \omega^{0} & \omega^{2} & \omega^{4} & \omega^{6} \\ \omega^{0} & \omega^{3} & \omega^{6} & \omega^{1} & \omega^{4} & \omega^{7} & \omega^{2} & \omega^{5} \\ \omega^{0} & \omega^{4} & \omega^{0} & \omega^{4} & \omega^{0} & \omega^{4} & \omega^{0} & \omega^{4} \\ \omega^{0} & \omega^{5} & \omega^{2} & \omega^{7} & \omega^{4} & \omega^{1} & \omega^{6} & \omega^{3} \\ \omega^{0} & \omega^{6} & \omega^{4} & \omega^{2} & \omega^{0} & \omega^{6} & \omega^{4} & \omega^{2} \\ \omega^{0} & \omega^{7} & \omega^{6} & \omega^{5} & \omega^{4} & \omega^{3} & \omega^{2} & \omega^{1} \end{bmatrix}

先把奇偶拆開來

\begin{bmatrix} a_0\\ a_1\\ a_2\\ a_3\\ a_4\\ a_5\\ a_6\\ a_7 \end{bmatrix}
= \begin{bmatrix} y_0\\ y_1\\ y_2\\ y_3\\ y_4\\ y_5\\ y_6\\ y_7 \end{bmatrix}
\begin{bmatrix} \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} \\ \omega^{0} & \omega^{2} & \omega^{4} & \omega^{6} & \omega^{1} & \omega^{3} & \omega^{5} & \omega^{7} \\ \omega^{0} & \omega^{4} & \omega^{0} & \omega^{4} & \omega^{2} & \omega^{6} & \omega^{2} & \omega^{6} \\ \omega^{0} & \omega^{6} & \omega^{4} & \omega^{2} & \omega^{3} & \omega^{1} & \omega^{7} & \omega^{5} \\ \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} & \omega^{4} & \omega^{4} & \omega^{4} & \omega^{4} \\ \omega^{0} & \omega^{2} & \omega^{4} & \omega^{6} & \omega^{5} & \omega^{7} & \omega^{1} & \omega^{3} \\ \omega^{0} & \omega^{4} & \omega^{0} & \omega^{4} & \omega^{6} & \omega^{2} & \omega^{6} & \omega^{2} \\ \omega^{0} & \omega^{6} & \omega^{4} & \omega^{2} & \omega^{7} & \omega^{5} & \omega^{3} & \omega^{1} \end{bmatrix}

先把奇偶拆開來

\begin{bmatrix} a_0\\ a_2\\ a_4\\ a_6\\ a_1\\ a_3\\ a_5\\ a_7 \end{bmatrix}
= \begin{bmatrix} y_0\\ y_1\\ y_2\\ y_3\\ y_4\\ y_5\\ y_6\\ y_7 \end{bmatrix}
\begin{bmatrix} \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} \\ \omega^{0} & \omega^{2} & \omega^{4} & \omega^{6} & \omega^{1} & \omega^{3} & \omega^{5} & \omega^{7} \\ \omega^{0} & \omega^{4} & \omega^{0} & \omega^{4} & \omega^{2} & \omega^{6} & \omega^{2} & \omega^{6} \\ \omega^{0} & \omega^{6} & \omega^{4} & \omega^{2} & \omega^{3} & \omega^{1} & \omega^{7} & \omega^{5} \\ \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} & \omega^{4} & \omega^{4} & \omega^{4} & \omega^{4} \\ \omega^{0} & \omega^{2} & \omega^{4} & \omega^{6} & \omega^{5} & \omega^{7} & \omega^{1} & \omega^{3} \\ \omega^{0} & \omega^{4} & \omega^{0} & \omega^{4} & \omega^{6} & \omega^{2} & \omega^{6} & \omega^{2} \\ \omega^{0} & \omega^{6} & \omega^{4} & \omega^{2} & \omega^{7} & \omega^{5} & \omega^{3} & \omega^{1} \end{bmatrix}

神奇的事出現了!

\begin{bmatrix} a_0\\ a_2\\ a_4\\ a_6\\ a_1\\ a_3\\ a_5\\ a_7 \end{bmatrix}
= \begin{bmatrix} y_0\\ y_1\\ y_2\\ y_3\\ y_4\\ y_5\\ y_6\\ y_7 \end{bmatrix}
\begin{bmatrix} \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} \\ \omega^{0} & \omega^{2} & \omega^{4} & \omega^{6} & \omega^{1} & \omega^{3} & \omega^{5} & \omega^{7} \\ \omega^{0} & \omega^{4} & \omega^{0} & \omega^{4} & \omega^{2} & \omega^{6} & \omega^{2} & \omega^{6} \\ \omega^{0} & \omega^{6} & \omega^{4} & \omega^{2} & \omega^{3} & \omega^{1} & \omega^{7} & \omega^{5} \\ \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} & \omega^{4} & \omega^{4} & \omega^{4} & \omega^{4} \\ \omega^{0} & \omega^{2} & \omega^{4} & \omega^{6} & \omega^{5} & \omega^{7} & \omega^{1} & \omega^{3} \\ \omega^{0} & \omega^{4} & \omega^{0} & \omega^{4} & \omega^{6} & \omega^{2} & \omega^{6} & \omega^{2} \\ \omega^{0} & \omega^{6} & \omega^{4} & \omega^{2} & \omega^{7} & \omega^{5} & \omega^{3} & \omega^{1} \end{bmatrix}

神奇的事出現了!

\begin{bmatrix} a_0\\ a_2\\ a_4\\ a_6\\ a_1\\ a_3\\ a_5\\ a_7 \end{bmatrix}
= \begin{bmatrix} y_0\\ y_1\\ y_2\\ y_3\\ y_4\\ y_5\\ y_6\\ y_7 \end{bmatrix}
\begin{bmatrix} \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} \\ \omega^{0} & \omega^{2} & \omega^{4} & \omega^{6} & \omega^{1} & \omega^{3} & \omega^{5} & \omega^{7} \\ \omega^{0} & \omega^{4} & \omega^{0} & \omega^{4} & \omega^{2} & \omega^{6} & \omega^{2} & \omega^{6} \\ \omega^{0} & \omega^{6} & \omega^{4} & \omega^{2} & \omega^{3} & \omega^{1} & \omega^{7} & \omega^{5} \\ \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} & \omega^{4} & \omega^{4} & \omega^{4} & \omega^{4} \\ \omega^{0} & \omega^{2} & \omega^{4} & \omega^{6} & \omega^{5} & \omega^{7} & \omega^{1} & \omega^{3} \\ \omega^{0} & \omega^{4} & \omega^{0} & \omega^{4} & \omega^{6} & \omega^{2} & \omega^{6} & \omega^{2} \\ \omega^{0} & \omega^{6} & \omega^{4} & \omega^{2} & \omega^{7} & \omega^{5} & \omega^{3} & \omega^{1} \end{bmatrix}

神奇的事出現了!

\begin{bmatrix} a_0\\ a_2\\ a_4\\ a_6\\ a_1\\ a_3\\ a_5\\ a_7 \end{bmatrix}
= \begin{bmatrix} y_0\\ y_1\\ y_2\\ y_3\\ y_4\\ y_5\\ y_6\\ y_7 \end{bmatrix}
\begin{bmatrix} \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} \\ \omega^{0} & \omega^{2} & \omega^{4} & \omega^{6} & \omega^{1} & \omega^{3} & \omega^{5} & \omega^{7} \\ \omega^{0} & \omega^{4} & \omega^{0} & \omega^{4} & \omega^{2} & \omega^{6} & \omega^{2} & \omega^{6} \\ \omega^{0} & \omega^{6} & \omega^{4} & \omega^{2} & \omega^{3} & \omega^{1} & \omega^{7} & \omega^{5} \\ \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} & \omega^{4} & \omega^{4} & \omega^{4} & \omega^{4} \\ \omega^{0} & \omega^{2} & \omega^{4} & \omega^{6} & \omega^{5} & \omega^{7} & \omega^{1} & \omega^{3} \\ \omega^{0} & \omega^{4} & \omega^{0} & \omega^{4} & \omega^{6} & \omega^{2} & \omega^{6} & \omega^{2} \\ \omega^{0} & \omega^{6} & \omega^{4} & \omega^{2} & \omega^{7} & \omega^{5} & \omega^{3} & \omega^{1} \end{bmatrix}

神奇的事出現了!

\begin{bmatrix} a_0\\ a_2\\ a_4\\ a_6\\ a_1\\ a_3\\ a_5\\ a_7 \end{bmatrix}
= \begin{bmatrix} y_0\\ y_1\\ y_2\\ y_3\\ y_4\\ y_5\\ y_6\\ y_7 \end{bmatrix}
\begin{bmatrix} \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} \\ \omega^{0} & \omega^{2} & \omega^{4} & \omega^{6} & \omega^{1} & \omega^{3} & \omega^{5} & \omega^{7} \\ \omega^{0} & \omega^{4} & \omega^{0} & \omega^{4} & \omega^{2} & \omega^{6} & \omega^{2} & \omega^{6} \\ \omega^{0} & \omega^{6} & \omega^{4} & \omega^{2} & \omega^{3} & \omega^{1} & \omega^{7} & \omega^{5} \\ \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} & \omega^{4} & \omega^{4} & \omega^{4} & \omega^{4} \\ \omega^{0} & \omega^{2} & \omega^{4} & \omega^{6} & \omega^{5} & \omega^{7} & \omega^{1} & \omega^{3} \\ \omega^{0} & \omega^{4} & \omega^{0} & \omega^{4} & \omega^{6} & \omega^{2} & \omega^{6} & \omega^{2} \\ \omega^{0} & \omega^{6} & \omega^{4} & \omega^{2} & \omega^{7} & \omega^{5} & \omega^{3} & \omega^{1} \end{bmatrix}

神奇的事出現了!

\begin{bmatrix} a_0\\ a_2\\ a_4\\ a_6\\ a_1\\ a_3\\ a_5\\ a_7 \end{bmatrix}
= \begin{bmatrix} y_0\\ y_1\\ y_2\\ y_3\\ y_4\\ y_5\\ y_6\\ y_7 \end{bmatrix}
\begin{bmatrix} \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} \\ \omega^{0} & \omega^{2} & \omega^{4} & \omega^{6} & \omega^{1} & \omega^{3} & \omega^{5} & \omega^{7} \\ \omega^{0} & \omega^{4} & \omega^{0} & \omega^{4} & \omega^{2} & \omega^{6} & \omega^{2} & \omega^{6} \\ \omega^{0} & \omega^{6} & \omega^{4} & \omega^{2} & \omega^{3} & \omega^{1} & \omega^{7} & \omega^{5} \\ \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} & \omega^{4} & \omega^{4} & \omega^{4} & \omega^{4} \\ \omega^{0} & \omega^{2} & \omega^{4} & \omega^{6} & \omega^{5} & \omega^{7} & \omega^{1} & \omega^{3} \\ \omega^{0} & \omega^{4} & \omega^{0} & \omega^{4} & \omega^{6} & \omega^{2} & \omega^{6} & \omega^{2} \\ \omega^{0} & \omega^{6} & \omega^{4} & \omega^{2} & \omega^{7} & \omega^{5} & \omega^{3} & \omega^{1} \end{bmatrix}

神奇的事出現了!

\begin{bmatrix} a_0\\ a_2\\ a_4\\ a_6\\ a_1\\ a_3\\ a_5\\ a_7 \end{bmatrix}

第\(i(i<\frac{n}{2})\) row的偶數項和第\(i+\frac{n}{2}\)row 一模一樣!

= \begin{bmatrix} y_0\\ y_1\\ y_2\\ y_3\\ y_4\\ y_5\\ y_6\\ y_7 \end{bmatrix}
\begin{bmatrix} \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} \\ \omega^{0} & \omega^{2} & \omega^{4} & \omega^{6} & \omega^{1} & \omega^{3} & \omega^{5} & \omega^{7} \\ \omega^{0} & \omega^{4} & \omega^{0} & \omega^{4} & \omega^{2} & \omega^{6} & \omega^{2} & \omega^{6} \\ \omega^{0} & \omega^{6} & \omega^{4} & \omega^{2} & \omega^{3} & \omega^{1} & \omega^{7} & \omega^{5} \\ \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} & \omega^{4} & \omega^{4} & \omega^{4} & \omega^{4} \\ \omega^{0} & \omega^{2} & \omega^{4} & \omega^{6} & \omega^{5} & \omega^{7} & \omega^{1} & \omega^{3} \\ \omega^{0} & \omega^{4} & \omega^{0} & \omega^{4} & \omega^{6} & \omega^{2} & \omega^{6} & \omega^{2} \\ \omega^{0} & \omega^{6} & \omega^{4} & \omega^{2} & \omega^{7} & \omega^{5} & \omega^{3} & \omega^{1} \end{bmatrix}

神奇的事出現了!

\begin{bmatrix} a_0\\ a_2\\ a_4\\ a_6\\ a_1\\ a_3\\ a_5\\ a_7 \end{bmatrix}
= \begin{bmatrix} y_0\\ y_1\\ y_2\\ y_3\\ y_4\\ y_5\\ y_6\\ y_7 \end{bmatrix}
\begin{bmatrix} \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} \\ \omega^{0} & \omega^{2} & \omega^{4} & \omega^{6} & \omega^{1} & \omega^{3} & \omega^{5} & \omega^{7} \\ \omega^{0} & \omega^{4} & \omega^{0} & \omega^{4} & \omega^{2} & \omega^{6} & \omega^{2} & \omega^{6} \\ \omega^{0} & \omega^{6} & \omega^{4} & \omega^{2} & \omega^{3} & \omega^{1} & \omega^{7} & \omega^{5} \\ \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} & \omega^{4} & \omega^{4} & \omega^{4} & \omega^{4} \\ \omega^{0} & \omega^{2} & \omega^{4} & \omega^{6} & \omega^{5} & \omega^{7} & \omega^{1} & \omega^{3} \\ \omega^{0} & \omega^{4} & \omega^{0} & \omega^{4} & \omega^{6} & \omega^{2} & \omega^{6} & \omega^{2} \\ \omega^{0} & \omega^{6} & \omega^{4} & \omega^{2} & \omega^{7} & \omega^{5} & \omega^{3} & \omega^{1} \end{bmatrix}

神奇的事出現了!

\begin{bmatrix} a_0\\ a_2\\ a_4\\ a_6\\ a_1\\ a_3\\ a_5\\ a_7 \end{bmatrix}

第\(i(i<4\) row的奇數項和第\(i+4\) row 差\(\omega^{4}\)!

= \begin{bmatrix} y_0\\ y_1\\ y_2\\ y_3\\ y_4\\ y_5\\ y_6\\ y_7 \end{bmatrix}
\begin{bmatrix} \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} \\ \omega^{0} & \omega^{2} & \omega^{4} & \omega^{6} & \omega^{1} & \omega^{3} & \omega^{5} & \omega^{7} \\ \omega^{0} & \omega^{4} & \omega^{0} & \omega^{4} & \omega^{2} & \omega^{6} & \omega^{2} & \omega^{6} \\ \omega^{0} & \omega^{6} & \omega^{4} & \omega^{2} & \omega^{3} & \omega^{1} & \omega^{7} & \omega^{5} \\ \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} & \omega^{4} & \omega^{4} & \omega^{4} & \omega^{4} \\ \omega^{0} & \omega^{2} & \omega^{4} & \omega^{6} & \omega^{5} & \omega^{7} & \omega^{1} & \omega^{3} \\ \omega^{0} & \omega^{4} & \omega^{0} & \omega^{4} & \omega^{6} & \omega^{2} & \omega^{6} & \omega^{2} \\ \omega^{0} & \omega^{6} & \omega^{4} & \omega^{2} & \omega^{7} & \omega^{5} & \omega^{3} & \omega^{1} \end{bmatrix}
\begin{bmatrix} a_0\\ a_2\\ a_4\\ a_6\\ a_1\\ a_3\\ a_5\\ a_7 \end{bmatrix}

試圖由上半得到下半

= \begin{bmatrix} y_0\\ y_1\\ y_2\\ y_3\\ y_4\\ y_5\\ y_6\\ y_7 \end{bmatrix}

試圖由上半得到下半

切切切!

 

\begin{bmatrix} \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} \\ \omega^{0} & \omega^{2} & \omega^{4} & \omega^{6} \\ \omega^{0} & \omega^{4} & \omega^{0} & \omega^{4} \\ \omega^{0} & \omega^{6} & \omega^{4} & \omega^{2} \end{bmatrix}
+ \begin{bmatrix} \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} \\ \omega^{1} & \omega^{3} & \omega^{5} & \omega^{7} \\ \omega^{2} & \omega^{6} & \omega^{2} & \omega^{6} \\ \omega^{3} & \omega^{1} & \omega^{7} & \omega^{5} \end{bmatrix}
\begin{bmatrix} a_1\\ a_3\\ a_5\\ a_7 \end{bmatrix}
+ \begin{bmatrix} \omega^{4} & \omega^{4} & \omega^{4} & \omega^{4} \\ \omega^{5} & \omega^{7} & \omega^{1} & \omega^{3} \\ \omega^{6} & \omega^{2} & \omega^{6} & \omega^{2} \\ \omega^{7} & \omega^{5} & \omega^{3} & \omega^{1} \end{bmatrix}
\begin{bmatrix} a_0\\ a_2\\ a_4\\ a_6 \end{bmatrix}
\begin{bmatrix} a_1\\ a_3\\ a_5\\ a_7 \end{bmatrix}
= \begin{bmatrix} y_4\\ y_5\\ y_6\\ y_7 \end{bmatrix}
\begin{bmatrix} a_0\\ a_2\\ a_4\\ a_6 \end{bmatrix}
\begin{bmatrix} \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0}\\ \omega^{0} & \omega^{2} & \omega^{4} & \omega^{6}\\ \omega^{0} & \omega^{4} & \omega^{0} & \omega^{4}\\ \omega^{0} & \omega^{6} & \omega^{4} & \omega^{2} \end{bmatrix}
= \begin{bmatrix} y_0\\ y_1\\ y_2\\ y_3 \end{bmatrix}

試圖由上半得到下半

切切切!

 

\begin{bmatrix} \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} \\ \omega^{0} & \omega^{2} & \omega^{4} & \omega^{6} \\ \omega^{0} & \omega^{4} & \omega^{0} & \omega^{4} \\ \omega^{0} & \omega^{6} & \omega^{4} & \omega^{2} \end{bmatrix}
+ \begin{bmatrix} \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} \\ \omega^{1} & \omega^{3} & \omega^{5} & \omega^{7} \\ \omega^{2} & \omega^{6} & \omega^{2} & \omega^{6} \\ \omega^{3} & \omega^{1} & \omega^{7} & \omega^{5} \end{bmatrix}
\begin{bmatrix} a_1\\ a_3\\ a_5\\ a_7 \end{bmatrix}
+ \begin{bmatrix} \omega^{4} & \omega^{4} & \omega^{4} & \omega^{4} \\ \omega^{5} & \omega^{7} & \omega^{1} & \omega^{3} \\ \omega^{6} & \omega^{2} & \omega^{6} & \omega^{2} \\ \omega^{7} & \omega^{5} & \omega^{3} & \omega^{1} \end{bmatrix}
\begin{bmatrix} a_0\\ a_2\\ a_4\\ a_6 \end{bmatrix}
\begin{bmatrix} a_1\\ a_3\\ a_5\\ a_7 \end{bmatrix}
= \begin{bmatrix} y_4\\ y_5\\ y_6\\ y_7 \end{bmatrix}
\begin{bmatrix} a_0\\ a_2\\ a_4\\ a_6 \end{bmatrix}
\begin{bmatrix} \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0}\\ \omega^{0} & \omega^{2} & \omega^{4} & \omega^{6}\\ \omega^{0} & \omega^{4} & \omega^{0} & \omega^{4}\\ \omega^{0} & \omega^{6} & \omega^{4} & \omega^{2} \end{bmatrix}
= \begin{bmatrix} y_0\\ y_1\\ y_2\\ y_3 \end{bmatrix}

上下一模一樣!

上下差\(\omega^4\)!

試圖由上半得到下半

切切切!

 

\begin{bmatrix} \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} \\ \omega^{0} & \omega^{2} & \omega^{4} & \omega^{6} \\ \omega^{0} & \omega^{4} & \omega^{0} & \omega^{4} \\ \omega^{0} & \omega^{6} & \omega^{4} & \omega^{2} \end{bmatrix}
+ \begin{bmatrix} \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} \\ \omega^{1} & \omega^{3} & \omega^{5} & \omega^{7} \\ \omega^{2} & \omega^{6} & \omega^{2} & \omega^{6} \\ \omega^{3} & \omega^{1} & \omega^{7} & \omega^{5} \end{bmatrix}
\begin{bmatrix} a_1\\ a_3\\ a_5\\ a_7 \end{bmatrix}
+\omega^{4} \begin{bmatrix} \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} \\ \omega^{1} & \omega^{3} & \omega^{5} & \omega^{7} \\ \omega^{2} & \omega^{6} & \omega^{2} & \omega^{6} \\ \omega^{3} & \omega^{1} & \omega^{7} & \omega^{5} \end{bmatrix}
\begin{bmatrix} a_0\\ a_2\\ a_4\\ a_6 \end{bmatrix}
\begin{bmatrix} a_1\\ a_3\\ a_5\\ a_7 \end{bmatrix}
= \begin{bmatrix} y_4\\ y_5\\ y_6\\ y_7 \end{bmatrix}
\begin{bmatrix} a_0\\ a_2\\ a_4\\ a_6 \end{bmatrix}
\begin{bmatrix} \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0}\\ \omega^{0} & \omega^{2} & \omega^{4} & \omega^{6}\\ \omega^{0} & \omega^{4} & \omega^{0} & \omega^{4}\\ \omega^{0} & \omega^{6} & \omega^{4} & \omega^{2} \end{bmatrix}
= \begin{bmatrix} y_0\\ y_1\\ y_2\\ y_3 \end{bmatrix}

試圖由上半得到下半

切切切!

 

\begin{bmatrix} \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} \\ \omega^{0} & \omega^{2} & \omega^{4} & \omega^{6} \\ \omega^{0} & \omega^{4} & \omega^{0} & \omega^{4} \\ \omega^{0} & \omega^{6} & \omega^{4} & \omega^{2} \end{bmatrix}
+ \begin{bmatrix} \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} \\ \omega^{1} & \omega^{3} & \omega^{5} & \omega^{7} \\ \omega^{2} & \omega^{6} & \omega^{2} & \omega^{6} \\ \omega^{3} & \omega^{1} & \omega^{7} & \omega^{5} \end{bmatrix}
\begin{bmatrix} a_1\\ a_3\\ a_5\\ a_7 \end{bmatrix}
- \begin{bmatrix} \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} \\ \omega^{1} & \omega^{3} & \omega^{5} & \omega^{7} \\ \omega^{2} & \omega^{6} & \omega^{2} & \omega^{6} \\ \omega^{3} & \omega^{1} & \omega^{7} & \omega^{5} \end{bmatrix}
\begin{bmatrix} a_0\\ a_2\\ a_4\\ a_6 \end{bmatrix}
\begin{bmatrix} a_1\\ a_3\\ a_5\\ a_7 \end{bmatrix}
= \begin{bmatrix} y_4\\ y_5\\ y_6\\ y_7 \end{bmatrix}
\begin{bmatrix} a_0\\ a_2\\ a_4\\ a_6 \end{bmatrix}
\begin{bmatrix} \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0}\\ \omega^{0} & \omega^{2} & \omega^{4} & \omega^{6}\\ \omega^{0} & \omega^{4} & \omega^{0} & \omega^{4}\\ \omega^{0} & \omega^{6} & \omega^{4} & \omega^{2} \end{bmatrix}
= \begin{bmatrix} y_0\\ y_1\\ y_2\\ y_3 \end{bmatrix}

好像快會了喔

 

\begin{bmatrix} \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} \\ \omega^{1} & \omega^{3} & \omega^{5} & \omega^{7} \\ \omega^{2} & \omega^{6} & \omega^{2} & \omega^{6} \\ \omega^{3} & \omega^{1} & \omega^{7} & \omega^{5} \end{bmatrix}
\begin{bmatrix} a_1\\ a_3\\ a_5\\ a_7 \end{bmatrix}
\begin{bmatrix} a_0\\ a_2\\ a_4\\ a_6 \end{bmatrix}
\begin{bmatrix} \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0}\\ \omega^{0} & \omega^{2} & \omega^{4} & \omega^{6}\\ \omega^{0} & \omega^{4} & \omega^{0} & \omega^{4}\\ \omega^{0} & \omega^{6} & \omega^{4} & \omega^{2} \end{bmatrix}

只要求出

就好了

好像快會了喔

\begin{bmatrix} \omega^{0+0} & \omega^{0+0} & \omega^{0+0} & \omega^{0+0} \\ \omega^{1+0} & \omega^{1+2} & \omega^{1+4} & \omega^{1+6} \\ \omega^{2+0} & \omega^{2+4} & \omega^{2+0} & \omega^{2+4} \\ \omega^{3+0} & \omega^{3+6} & \omega^{3+4} & \omega^{3+2} \end{bmatrix}
\begin{bmatrix} a_1\\ a_3\\ a_5\\ a_7 \end{bmatrix}
\begin{bmatrix} a_0\\ a_2\\ a_4\\ a_6 \end{bmatrix}
\begin{bmatrix} \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0}\\ \omega^{0} & \omega^{2} & \omega^{4} & \omega^{6}\\ \omega^{0} & \omega^{4} & \omega^{0} & \omega^{4}\\ \omega^{0} & \omega^{6} & \omega^{4} & \omega^{2} \end{bmatrix}

只要求出

就好了

發現每row都只比左邊多一個\(\omega^i\),可以先丟掉

好像快會了喔

\begin{bmatrix} \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0}\\ \omega^{0} & \omega^{2} & \omega^{4} & \omega^{6}\\ \omega^{0} & \omega^{4} & \omega^{0} & \omega^{4}\\ \omega^{0} & \omega^{6} & \omega^{4} & \omega^{2} \end{bmatrix}
\begin{bmatrix} \ a_1\\ \ a_3\\ \ a_5\\ \ a_7 \end{bmatrix}
\begin{bmatrix} a_0\\ a_2\\ a_4\\ a_6 \end{bmatrix}
\begin{bmatrix} \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0}\\ \omega^{0} & \omega^{2} & \omega^{4} & \omega^{6}\\ \omega^{0} & \omega^{4} & \omega^{0} & \omega^{4}\\ \omega^{0} & \omega^{6} & \omega^{4} & \omega^{2} \end{bmatrix}

只要求出

就好了

矩陣乘法後得到的第i項少乘了\(\omega^i\),之後要把他乘回來

然後因為\(\omega_{n}^{i} = \omega_{\frac{n}{2}}^{\frac{i}{2}}\),

會發現其實上述兩坨東西都是再做一次大小砍半的事(DFT)

好像快會了喔

\begin{bmatrix} \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0}\\ \omega^{0} & \omega^{2} & \omega^{4} & \omega^{6}\\ \omega^{0} & \omega^{4} & \omega^{0} & \omega^{4}\\ \omega^{0} & \omega^{6} & \omega^{4} & \omega^{2} \end{bmatrix}
\begin{bmatrix} a_0\\ a_2\\ a_4\\ a_6 \end{bmatrix}
\begin{bmatrix} \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0}\\ \omega^{0} & \omega^{2} & \omega^{4} & \omega^{6}\\ \omega^{0} & \omega^{4} & \omega^{0} & \omega^{4}\\ \omega^{0} & \omega^{6} & \omega^{4} & \omega^{2} \end{bmatrix}

只要求出

就好了

\begin{bmatrix} \ a_1\\ \ a_3\\ \ a_5\\ \ a_7 \end{bmatrix}

然後因為\(\omega_{n}^{i} = \omega_{\frac{n}{2}}^{\frac{i}{2}}\),

會發現其實上述兩坨東西都是再做一次大小砍半的事(DFT)

分治!

根據主定理,總時間複雜度\(O(n \log n)\)

好像快會了喔

\begin{bmatrix} \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0}\\ \omega^{0} & \omega^{2} & \omega^{4} & \omega^{6}\\ \omega^{0} & \omega^{4} & \omega^{0} & \omega^{4}\\ \omega^{0} & \omega^{6} & \omega^{4} & \omega^{2} \end{bmatrix}
\begin{bmatrix} \ 1/\omega^0\ a_1\\ \ 1/\omega^1\ a_3\\ \ 1/\omega^2\ a_5\\ \ 1/\omega^3\ a_7 \end{bmatrix}
\begin{bmatrix} a_0\\ a_2\\ a_4\\ a_6 \end{bmatrix}
\begin{bmatrix} \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0}\\ \omega^{0} & \omega^{2} & \omega^{4} & \omega^{6}\\ \omega^{0} & \omega^{4} & \omega^{0} & \omega^{4}\\ \omega^{0} & \omega^{6} & \omega^{4} & \omega^{2} \end{bmatrix}

只要求出

就好了

參考code

//CD為complex<double>
void FFT (vector <CD>& a) {
    int n = a.size();
    if (n == 1) return;
    vector <CD> even(n / 2), odd(n / 2);
    for (int i = 0; i < n / 2; i++) even[i] = a[i * 2],  odd[i] = a[i * 2 + 1];
    FFT(even); FFT(odd);
    for (int i = 0; i < n / 2; i++) {
        CD z = OMEGA(i, n);
        a[i] = even[i] + odd[i] * z;
        a[i + n / 2] = even[i] - odd[i] * z;
    }
return;}

我們現在會很快地得到多項式圖形上的\(n\)個點了!

那要怎麼很快從多個點轉回多項式?

我們現在會很快地得到多項式圖形上的\(n\)個點了!

那要怎麼很快從多個點轉回多項式?

我們剛剛帶入點是用矩陣乘法對吧,也就是\(TA\)=\(Y_A\)

然後我們會用\(Y_A,Y_B\)得到\(Y_C\)

要怎麼從\(Y_C\)得到\(C\)?

我們現在會很快地得到多項式圖形上的\(n\)個點了!

那要怎麼很快從多個點轉回多項式?

我們剛剛帶入點是用矩陣乘法對吧,也就是\(TA\)=\(Y_A\)

然後我們會用\(Y_A,Y_B\)得到\(Y_C\)

要怎麼從\(Y_C\)得到\(C\)?

再乘一個\(T^{-1}\)! 它是誰呢?

我們現在會很快地得到多項式圖形上的\(n\)個點了!

那要怎麼很快從多個點轉回多項式?

我們剛剛帶入點是用矩陣乘法對吧,也就是\(TA\)=\(Y_A\)

然後我們會用\(Y_A,Y_B\)得到\(Y_C\)

要怎麼從\(Y_C\)得到\(C\)?

再乘一個\(T^{-1}\)! 它是誰呢?

通靈一下,跳結論

考慮一下\(T\=T\),這裡的\(\=T\)指的是對T的每個元素取共軛

\begin{bmatrix} \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} \\ \omega^{0} & \omega^{-1} & \omega^{-2} & \omega^{-3} & \omega^{-4} & \omega^{-5} & \omega^{-6} & \omega^{-7} \\ \omega^{0} & \omega^{-2} & \omega^{-4} & \omega^{-6} & \omega^{0} & \omega^{-2} & \omega^{-4} & \omega^{-6} \\ \omega^{0} & \omega^{-3} & \omega^{-6} & \omega^{-1} & \omega^{-4} & \omega^{-7} & \omega^{-2} & \omega^{-5} \\ \omega^{0} & \omega^{-4} & \omega^{0} & \omega^{-4} & \omega^{0} & \omega^{-4} & \omega^{0} & \omega^{-4} \\ \omega^{0} & \omega^{-5} & \omega^{-2} & \omega^{-7} & \omega^{-4} & \omega^{-1} & \omega^{-6} & \omega^{-3} \\ \omega^{0} & \omega^{-6} & \omega^{-4} & \omega^{-2} & \omega^{0} & \omega^{-6} & \omega^{-4} & \omega^{-2} \\ \omega^{0} & \omega^{-7} & \omega^{-6} & \omega^{-5} & \omega^{-4} & \omega^{-3} & \omega^{-2} & \omega^{-1} \end{bmatrix} \begin{bmatrix} \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} \\ \omega^{0} & \omega^{1} & \omega^{2} & \omega^{3} & \omega^{4} & \omega^{5} & \omega^{6} & \omega^{7} \\ \omega^{0} & \omega^{2} & \omega^{4} & \omega^{6} & \omega^{0} & \omega^{2} & \omega^{4} & \omega^{6} \\ \omega^{0} & \omega^{3} & \omega^{6} & \omega^{1} & \omega^{4} & \omega^{7} & \omega^{2} & \omega^{5} \\ \omega^{0} & \omega^{4} & \omega^{0} & \omega^{4} & \omega^{0} & \omega^{4} & \omega^{0} & \omega^{4} \\ \omega^{0} & \omega^{5} & \omega^{2} & \omega^{7} & \omega^{4} & \omega^{1} & \omega^{6} & \omega^{3} \\ \omega^{0} & \omega^{6} & \omega^{4} & \omega^{2} & \omega^{0} & \omega^{6} & \omega^{4} & \omega^{2} \\ \omega^{0} & \omega^{7} & \omega^{6} & \omega^{5} & \omega^{4} & \omega^{3} & \omega^{2} & \omega^{1} \end{bmatrix}

考慮一下\(T\=T\),這裡的\(\=T\)指的是對T的每個元素取共軛

\begin{bmatrix} \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} \\ \omega^{0} & \omega^{-1} & \omega^{-2} & \omega^{-3} & \omega^{-4} & \omega^{-5} & \omega^{-6} & \omega^{-7} \\ \omega^{0} & \omega^{-2} & \omega^{-4} & \omega^{-6} & \omega^{0} & \omega^{-2} & \omega^{-4} & \omega^{-6} \\ \omega^{0} & \omega^{-3} & \omega^{-6} & \omega^{-1} & \omega^{-4} & \omega^{-7} & \omega^{-2} & \omega^{-5} \\ \omega^{0} & \omega^{-4} & \omega^{0} & \omega^{-4} & \omega^{0} & \omega^{-4} & \omega^{0} & \omega^{-4} \\ \omega^{0} & \omega^{-5} & \omega^{-2} & \omega^{-7} & \omega^{-4} & \omega^{-1} & \omega^{-6} & \omega^{-3} \\ \omega^{0} & \omega^{-6} & \omega^{-4} & \omega^{-2} & \omega^{0} & \omega^{-6} & \omega^{-4} & \omega^{-2} \\ \omega^{0} & \omega^{-7} & \omega^{-6} & \omega^{-5} & \omega^{-4} & \omega^{-3} & \omega^{-2} & \omega^{-1} \end{bmatrix} \begin{bmatrix} \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} \\ \omega^{0} & \omega^{1} & \omega^{2} & \omega^{3} & \omega^{4} & \omega^{5} & \omega^{6} & \omega^{7} \\ \omega^{0} & \omega^{2} & \omega^{4} & \omega^{6} & \omega^{0} & \omega^{2} & \omega^{4} & \omega^{6} \\ \omega^{0} & \omega^{3} & \omega^{6} & \omega^{1} & \omega^{4} & \omega^{7} & \omega^{2} & \omega^{5} \\ \omega^{0} & \omega^{4} & \omega^{0} & \omega^{4} & \omega^{0} & \omega^{4} & \omega^{0} & \omega^{4} \\ \omega^{0} & \omega^{5} & \omega^{2} & \omega^{7} & \omega^{4} & \omega^{1} & \omega^{6} & \omega^{3} \\ \omega^{0} & \omega^{6} & \omega^{4} & \omega^{2} & \omega^{0} & \omega^{6} & \omega^{4} & \omega^{2} \\ \omega^{0} & \omega^{7} & \omega^{6} & \omega^{5} & \omega^{4} & \omega^{3} & \omega^{2} & \omega^{1} \end{bmatrix}

發現第\(i\) row和第\(i\) column其實一樣,取負號乘起來就會變\(n\)個\(\omega^0\)

考慮一下\(T\=T\),這裡的\(\=T\)指的是對T的每個元素取共軛

\begin{bmatrix} \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} \\ \omega^{0} & \omega^{-1} & \omega^{-2} & \omega^{-3} & \omega^{-4} & \omega^{-5} & \omega^{-6} & \omega^{-7} \\ \omega^{0} & \omega^{-2} & \omega^{-4} & \omega^{-6} & \omega^{0} & \omega^{-2} & \omega^{-4} & \omega^{-6} \\ \omega^{0} & \omega^{-3} & \omega^{-6} & \omega^{-1} & \omega^{-4} & \omega^{-7} & \omega^{-2} & \omega^{-5} \\ \omega^{0} & \omega^{-4} & \omega^{0} & \omega^{-4} & \omega^{0} & \omega^{-4} & \omega^{0} & \omega^{-4} \\ \omega^{0} & \omega^{-5} & \omega^{-2} & \omega^{-7} & \omega^{-4} & \omega^{-1} & \omega^{-6} & \omega^{-3} \\ \omega^{0} & \omega^{-6} & \omega^{-4} & \omega^{-2} & \omega^{0} & \omega^{-6} & \omega^{-4} & \omega^{-2} \\ \omega^{0} & \omega^{-7} & \omega^{-6} & \omega^{-5} & \omega^{-4} & \omega^{-3} & \omega^{-2} & \omega^{-1} \end{bmatrix} \begin{bmatrix} \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} \\ \omega^{0} & \omega^{1} & \omega^{2} & \omega^{3} & \omega^{4} & \omega^{5} & \omega^{6} & \omega^{7} \\ \omega^{0} & \omega^{2} & \omega^{4} & \omega^{6} & \omega^{0} & \omega^{2} & \omega^{4} & \omega^{6} \\ \omega^{0} & \omega^{3} & \omega^{6} & \omega^{1} & \omega^{4} & \omega^{7} & \omega^{2} & \omega^{5} \\ \omega^{0} & \omega^{4} & \omega^{0} & \omega^{4} & \omega^{0} & \omega^{4} & \omega^{0} & \omega^{4} \\ \omega^{0} & \omega^{5} & \omega^{2} & \omega^{7} & \omega^{4} & \omega^{1} & \omega^{6} & \omega^{3} \\ \omega^{0} & \omega^{6} & \omega^{4} & \omega^{2} & \omega^{0} & \omega^{6} & \omega^{4} & \omega^{2} \\ \omega^{0} & \omega^{7} & \omega^{6} & \omega^{5} & \omega^{4} & \omega^{3} & \omega^{2} & \omega^{1} \end{bmatrix}

發現第\(i\) row和第\(i\) column其實一樣,取負號乘起來就會變\(n\)個\(\omega^0\)

=
\begin{bmatrix} n\omega^{0} & & & & & & & \\ & n\omega^{0} & & & & & & \\ & & n\omega^{0} & & & & & \\ & & & n\omega^{0} & & & & \\ & & & & n\omega^{0} & & & \\ & & & & & n\omega^{0} & & \\ & & & & & & n\omega^{0} & \\ & & & & & & & n\omega^{0} \end{bmatrix}

考慮一下\(T\=T\),這裡的\(\=T\)指的是對T的每個元素取共軛

\begin{bmatrix} \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} \\ \omega^{0} & \omega^{-1} & \omega^{-2} & \omega^{-3} & \omega^{-4} & \omega^{-5} & \omega^{-6} & \omega^{-7} \\ \omega^{0} & \omega^{-2} & \omega^{-4} & \omega^{-6} & \omega^{0} & \omega^{-2} & \omega^{-4} & \omega^{-6} \\ \omega^{0} & \omega^{-3} & \omega^{-6} & \omega^{-1} & \omega^{-4} & \omega^{-7} & \omega^{-2} & \omega^{-5} \\ \omega^{0} & \omega^{-4} & \omega^{0} & \omega^{-4} & \omega^{0} & \omega^{-4} & \omega^{0} & \omega^{-4} \\ \omega^{0} & \omega^{-5} & \omega^{-2} & \omega^{-7} & \omega^{-4} & \omega^{-1} & \omega^{-6} & \omega^{-3} \\ \omega^{0} & \omega^{-6} & \omega^{-4} & \omega^{-2} & \omega^{0} & \omega^{-6} & \omega^{-4} & \omega^{-2} \\ \omega^{0} & \omega^{-7} & \omega^{-6} & \omega^{-5} & \omega^{-4} & \omega^{-3} & \omega^{-2} & \omega^{-1} \end{bmatrix} \begin{bmatrix} \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} & \omega^{0} \\ \omega^{0} & \omega^{1} & \omega^{2} & \omega^{3} & \omega^{4} & \omega^{5} & \omega^{6} & \omega^{7} \\ \omega^{0} & \omega^{2} & \omega^{4} & \omega^{6} & \omega^{0} & \omega^{2} & \omega^{4} & \omega^{6} \\ \omega^{0} & \omega^{3} & \omega^{6} & \omega^{1} & \omega^{4} & \omega^{7} & \omega^{2} & \omega^{5} \\ \omega^{0} & \omega^{4} & \omega^{0} & \omega^{4} & \omega^{0} & \omega^{4} & \omega^{0} & \omega^{4} \\ \omega^{0} & \omega^{5} & \omega^{2} & \omega^{7} & \omega^{4} & \omega^{1} & \omega^{6} & \omega^{3} \\ \omega^{0} & \omega^{6} & \omega^{4} & \omega^{2} & \omega^{0} & \omega^{6} & \omega^{4} & \omega^{2} \\ \omega^{0} & \omega^{7} & \omega^{6} & \omega^{5} & \omega^{4} & \omega^{3} & \omega^{2} & \omega^{1} \end{bmatrix}
=
\begin{bmatrix} n\omega^{0} & 0 & 0 & 0 & 0 & 0 & 0 & 0 \\ 0 & n\omega^{0} & 0 & 0 & 0 & 0 & 0 & 0 \\ 0 & 0 & n\omega^{0} & 0 & 0 & 0 & 0 & 0 \\ 0 & 0 & 0 & n\omega^{0} & 0 & 0 & 0 & 0 \\ 0 & 0 & 0 & 0 & n\omega^{0} & 0 & 0 & 0 \\ 0 & 0 & 0 & 0 & 0 & n\omega^{0} & 0 & 0 \\ 0 & 0 & 0 & 0 & 0 & 0 & n\omega^{0} & 0 \\ 0 & 0 & 0 & 0 & 0 & 0 & 0 & n\omega^{0} \end{bmatrix}

被嗆說這就是角速度

每一圈加總都是0,也就是說全都是0

每一圈加總都是0,也就

\(T\=T=nI_n\)

\(T^{-1}\)其實就是\(\frac{1}{n}\=T\)

瑣事-DFT

  • 剛剛把\(\omega\)帶入方程式的動作稱為DFT
  • 把點變回方程式叫做IDFT
  • 一次FFT由兩次DFT、一次IDFT組成,\(O(n\log n)\)

瑣事-實作

  • 資料型態為complex<double>,內建四則運算
  • 先把所有的\(\omega_N\)找出來很好,這個\(N\)是多項式\(C\)的次數

完整code

using ld = long double;
using CD = complex<ld>;

CD OMEGA (int i, int n) {
    ld angle = i;
    angle = angle / n * (2 * pi);
    return CD(cos(angle), sin(angle));
}

void FFT (vector <CD>& a, bool inv = 0) {
    int n = a.size();
    if (n == 1) return;
    vector <CD> even(n / 2), odd(n / 2);
    for (int i = 0; i < n / 2; i++) even[i] = a[i * 2],  odd[i] = a[i * 2 + 1];
    FFT(even, inv); FFT(odd, inv);
    for (int i = 0; i < n / 2; i++) {
        CD z = OMEGA(i * (inv?-1:1), n);
        a[i] = even[i] + odd[i] * z;
        a[i + n / 2] = even[i] - odd[i] * z;
        if (inv) a[i] /= 2, a[i + n / 2] /= 2;
    }
return;}

NTT

為什麼FFT是好的?

為什麼FFT是好的?

因為\((\omega_n^{i})^{2j}=(\omega_n^{i+\frac{n}{2}})^{2j}\)和\(-(\omega_n^{i})^{2j+1}=(\omega_n^{i+\frac{n}{2}})^{2j+1}\)

為什麼FFT是好的?

因為\((\omega_n^{i})^{2j}=(\omega_n^{i+\frac{n}{2}})^{2j}\)和\(-(\omega_n^{i})^{2j+1}=(\omega_n^{i+\frac{n}{2}})^{2j+1}\)

什麼也有這個性質?(可以拿來取代單位根)

還記得野獸先備裡面有甚麼嗎?

為什麼FFT是好的?

因為\((\omega_n^{i})^{2j}=(\omega_n^{i+\frac{n}{2}})^{2j}\)和\(-(\omega_n^{i})^{2j+1}=(\omega_n^{i+\frac{n}{2}})^{2j+1}\)

什麼也有這個性質?(可以拿來取代單位根)

還記得野獸先備裡面有甚麼嗎?

在模質數\(M=r\times 2^k+1\)之下為\(M\)的原根\(g\)的\(r\)次方

為什麼FFT是好的?

因為\((\omega_n^{i})^{2j}=(\omega_n^{i+\frac{n}{2}})^{2j}\)和\(-(\omega_n^{i})^{2j+1}=(\omega_n^{i+\frac{n}{2}})^{2j+1}\)

什麼也有這個性質?(可以拿來取代單位根)

還記得野獸先備裡面有甚麼嗎?

在模質數\(M=r\times 2^k+1\)之下為\(M\)的原根\(g\)的\(r\)次方

\((g^r)^{n}=g^{r\times 2^k}\)\(=g^{M-1}\equiv1\)

 

為什麼FFT是好的?

因為\((\omega_n^{i})^{2j}=(\omega_n^{i+\frac{n}{2}})^{2j}\)和\(-(\omega_n^{i})^{2j+1}=(\omega_n^{i+\frac{n}{2}})^{2j+1}\)

什麼也有這個性質?(可以拿來取代單位根)

還記得野獸先備裡面有甚麼嗎?

在模質數\(M=r\times 2^k+1\)之下為\(M\)的原根\(g\)的\(r\)次方

\((g^r)^{n}=g^{r\times 2^k}\)\(=g^{M-1}\equiv 1\)

\((g^r)^{\frac{n}{2}}=g^{r\times 2^{k-1}}\)\(=g^\frac{M-1}{2}\equiv \pm1\)(正不合因為還沒繞一圈)

參考code

using ll = long long;
const int M = 998244353;
void NTT (vector <ll>& a, bool inv = 0) {
    int n = a.size();
    if (n == 1) return;
    vector <ll> even(n / 2), odd(n / 2);
    for (int i = 0; i < n / 2; i++) even[i] = a[i * 2],  odd[i] = a[i * 2 + 1];
    NTT(even, inv); 
    NTT(odd, inv);
    for (int i = 0; i < n / 2; i++) {
        ll z = OMEGA(i, n);
        a[i] = (even[i] + odd[i] * z) % M;
        a[i + n / 2] = (even[i] - odd[i] * z) % M;
        a[i] = (a[i] % M + M) % M;
        a[i + n / 2] = (a[i + n / 2] % M + M) % M;
    }
    if (inv and n == nn) {
        int qq = f(n, M-2);
        for (int i = 0; i < n; i++){
            a[i] *= qq;
            a[i] = (a[i] % M + M) % M;
        }
    }
return;}
...
...
...
omega[0] = 1,  omega[1] = f(3, (M - 1) / nn);
for (int i = 2; i <= nn; i++) omega[i] = omega[i - 1] * omega[1] % M;

常用模數

  • \(998244353=119\times2^{23}+1\),\(g=3\)
  • \(1004535809=479\times2^{21}+1\),\(g=3\)
  • \(469762049=7\times2^{26}+1\),\(g=3\)
  • 可以上網看一下NTT常用模數表
  • 原根不一定只有一個,但只要背一個就好

FFT v.s. NTT

  • FFT需要浮點數,精度可能燒雞
  • NTT的值雖然是整數,但全都模過\(M\)
  • 運行上,NTT會快一點

丟掉遞迴

丟掉手錶丟外套丟掉背包再丟嘮叨丟掉電視丟電腦丟掉大腦再丟煩惱衝啥大衝啥小

\(a_0, a_1, a_2, a_3, a_4, a_5, a_6, a_7\)

\(a_1, a_3, a_5, a_7\)

\(a_0, a_4\)

\(a_0, a_2, a_4, a_6\)

\(a_2, a_6\)

\(a_1, a_5\)

\(a_3, a_7\)

\(a_0\)

\(a_4\)

\(a_2\)

\(a_6\)

\(a_1\)

\(a_5\)

\(a_3\)

\(a_7\)

DFT架構圖

\(a_0, a_1, a_2, a_3, a_4, a_5, a_6, a_7\)

\(a_1, a_3, a_5, a_7\)

\(a_0, a_4\)

\(a_0, a_2, a_4, a_6\)

\(a_2, a_6\)

\(a_1, a_5\)

\(a_3, a_7\)

\(a_0\)

\(a_4\)

\(a_2\)

\(a_6\)

\(a_1\)

\(a_5\)

\(a_3\)

\(a_7\)

顯然可以排序成最下面那排,再一一往上合併

\(a_0, a_1, a_2, a_3, a_4, a_5, a_6, a_7\)

\(a_1, a_3, a_5, a_7\)

\(a_0, a_4\)

\(a_0, a_2, a_4, a_6\)

\(a_2, a_6\)

\(a_1, a_5\)

\(a_3, a_7\)

\(a_0\)

\(a_4\)

\(a_2\)

\(a_6\)

\(a_1\)

\(a_5\)

\(a_3\)

\(a_7\)

000

100

010

110

001

101

011

111

二進位

\(a_0, a_1, a_2, a_3, a_4, a_5, a_6, a_7\)

\(a_1, a_3, a_5, a_7\)

\(a_0, a_4\)

\(a_0, a_2, a_4, a_6\)

\(a_2, a_6\)

\(a_1, a_5\)

\(a_3, a_7\)

\(a_0\)

\(a_4\)

\(a_2\)

\(a_6\)

\(a_1\)

\(a_5\)

\(a_3\)

\(a_7\)

000

100

010

110

001

101

011

111

二進位

反二進位

000

001

010

011

100

101

110

111

其實是對reverse的二進位排序!

\(a_0, a_1, a_2, a_3, a_4, a_5, a_6, a_7\)

\(a_1, a_3, a_5, a_7\)

\(a_0, a_4\)

\(a_0, a_2, a_4, a_6\)

\(a_2, a_6\)

\(a_1, a_5\)

\(a_3, a_7\)

\(a_0\)

\(a_4\)

\(a_2\)

\(a_6\)

\(a_1\)

\(a_5\)

\(a_3\)

\(a_7\)

其實是對reverse的二進位排序!

對最後一位排序

對第二位排序

對首位排序

參考code

using ll = long long;
const int M = 998244353;
int rev[2010000];
ll omega[2010000];

void NTT (vector <ll>& a) {
	//排序
    int n = a.size();
    int k = __lg(n);
    memset(rev, 0, sizeof(rev));
    for (int i = 0; i < nn; i++) for (int j = 0; j < k; j++) 
    	rev[i] += (1 << j) * ((i >> (k - 1 - j)) & 1);
    for (int i = 0; i < nn; i++) if (rev[i] < i) swap(a[i], a[rev[i]]);

	//NTT,我這裡是用滾動的
    vector <ll> ans[2] = {a, a};
    int lst = 0, now = 1;
    for (int len = 2; len <= nn; len *= 2) {
        for (int j = 0; j < nn; j += len) {
            for (int p = j; p < j + len / 2; p++) {
                ll c = ans[lst][p], d = ans[lst][p + len / 2];
                ans[now][p] = ((c + OMEGA(p - j, len) * d) % M + M) % M; 
                ans[now][p + len / 2] = ((c + OMEGA(p - j + len / 2, len) * d) % M + M) % M; 
            }    
        }
        swap(lst, now);
    }

    a = ans[lst];
return;}

優點

  • 跑比較快

任意模數FFT

與其說是任意模數不如說是求得真值後,

再對一個任意的模數取模

3NTT

  • 做三次NTT
  • 中國剩餘定理
  • 聰明的合併、取模或大數處理

3NTT

  • 做三次NTT
  • 中國剩餘定理
  • 聰明的合併、取模或大數處理
  • 不會中國剩餘定理的請舉手

3NTT

  • 做三次NTT
  • 中國剩餘定理
  • 聰明的合併、取模或大數處理
  • 不會中國剩餘定理的請舉手
  • 記得學一下,你們現在可能不知道,但我下次看到你們的時候,你們就知道為什麼很重要了(by蔡孟宗)

3NTT 參考code

#include <bits/stdc++.h>
using namespace std;
#pragma GCC optimize("Ofast")
using ll = long long;
using Int = __int128;


int n, m, nn;
vector <ll> a, b, c;
vector <ll> amod[3], bmod[3], cmod[3];

int fmul (int x, int t, int M) {
    if (t == 0) return 1;
    ll y = fmul(x, t >> 1, M);
    y = y * y % M;
    if (t & 1) y = y * x % M;
    return y;
}



struct NTTblackbox{
    ll M = 998244353, G;
    int rev[2010000];
    ll omega[2010000];

    int f (int x, int t) {return fmul(x, t, M);}

    void init (int mod, int g = 3) {
        M = mod, G = g;
        omega[0] = 1,  omega[1] = f(G, (M - 1) / nn);
        for (int i = 2; i <= nn; i++) omega[i] = omega[i - 1] * omega[1] % M;
    }

    void reverseomega(){
        for (int i = 0; i < nn / 2; i++) swap(omega[i], omega[nn - i]);
    }

    ll OMEGA (ll i, int n) {
        i *= (nn / n);
        i %= M;
        return omega[i];
    }

    vector<ll> NTT (vector <ll>& oa, bool inv = 0) {
        auto a = oa;
        int kk = __lg(nn);
        memset(rev, 0, sizeof(rev));
        for (int i = 0; i < nn; i++) for (int j = 0; j < kk; j++) rev[i] += (1 << j) * ((i >> (kk - 1 - j)) & 1);
        for (int i = 0; i < nn; i++) if (rev[i] < i) swap(a[i], a[rev[i]]);

        vector <ll> ans[2] = {a, a};
        int lst = 0, now = 1;

        for (int len = 2; len <= nn; len *= 2) {
            for (int j = 0; j < nn; j += len) {
                for (int p = j; p < j + len / 2; p++) {
                    ll c = ans[lst][p], d = ans[lst][p + len / 2];
                    ans[now][p] = ((c + OMEGA(p - j, len) * d) % M + M) % M; 
                    ans[now][p + len / 2] = ((c + OMEGA(p - j + len / 2, len) * d) % M + M) % M; 
                }    
            }
            swap(lst, now);
        }
        if (inv) {
            int ninv = f(nn, M - 2);
            for (int i = 0; i < nn; i++) ans[lst][i] = ans[lst][i] * ninv % M;
        }
        return ans[lst];}

}nttg[3];



Int chineseremainder (vector<ll>M, vector<ll>a) {
    int N = M.size();
    Int res = 0, n = 1;
    for (int i = 0; i < N; i++) n *= M[i];
    for (int i = 0; i < N; i++) {
        Int ni = n / M[i];
        Int ci = ni * fmul(ni % M[i], M[i] - 2, M[i]) % n;
        res += a[i] * ci % n;
    }
    res %= (n);
    return res;
}

int main () {
    // ios_base::sync_with_stdio(false); cin.tie(0);
    cin >> n >> m;
    nn = 2 << __lg(n + m - 1);
    a = b = c = vector<ll> (nn, 0);
    for (int i = 0; i < n; i++) cin >> a[i];
    for (int i = 0; i < m; i++) cin >> b[i];

    
    nttg[0].init(998244353, 3);
    nttg[1].init(1004535809, 3);
    nttg[2].init(469762049, 3);
    for (int t = 0; t < 3; t++) {
        amod[t] = nttg[t].NTT(a);
        bmod[t] = nttg[t].NTT(b);
        cmod[t].resize(nn);
        for (int i = 0; i < nn; i++) cmod[t][i] = amod[t][i] * bmod[t][i] % nttg[t].M;
        nttg[t].reverseomega();
        cmod[t] = nttg[t].NTT(cmod[t], 1);
    }

    for (int i = 0; i < n + m - 1; i++) {
        Int res = chineseremainder ({nttg[0].M, nttg[1].M, nttg[2].M}, {cmod[0][i], cmod[1][i], cmod[2][i]});
        ll ans = res % (1000000000 + 7);
        cout << ans << ' ';
    }
    cout << '\n';
}

MTT

  • 神仙FFT
  • 把一個多項式拆成兩個來減少精度造成的影響
  • 純樸做要12次DFT,但可以優化到4次

FWHT(一般化位元卷積)

FFT 是想要快速算出$$c_k=\sum_{i+j=k}a_i\ b_j$$

FWHT 是想要快速算出$$c_k=\sum_{f(i,j)=k}a_i\ b_j$$

這個\(f(i,j)\)可以是任何二元位元運算

同樣的方法甚至可以擴展到某些不是位元運算的運算

我們先直接接受FWHT和FFT可以用一樣的方法做

  1. 算\(TA\), \(TB\)
  2. 用\(TA\), \(TB\)做出\(TC\)
  3. 算\(T^{-1}TC\),得到\(C\)

以XOR為例

\(F\)是誰?

以XOR為例子

先看2*2的矩陣\(T_2=\begin{bmatrix}w&x\\y&z\end{bmatrix}\)

我們要\(T_2\begin{bmatrix}a_0\\a_1\end{bmatrix}\cdot\) \(T_2\begin{bmatrix}b_0\\b_1\end{bmatrix}=\) \(T_2\begin{bmatrix}c_0\\c_1\end{bmatrix}=\) \(T_2\begin{bmatrix}a_0b_0+a_1b_1\\a_0b_1+a_1b_0\end{bmatrix}\)

內積(點積)

以XOR為例子

先看2*2的矩陣\(T_2=\begin{bmatrix}w&x\\y&z\end{bmatrix}\)

我們要\(T_2\begin{bmatrix}a_0\\a_1\end{bmatrix}\cdot\) \(T_2\begin{bmatrix}b_0\\b_1\end{bmatrix}=\) \(T_2\begin{bmatrix}c_0\\c_1\end{bmatrix}=\) \(T_2\begin{bmatrix}a_0b_0+a_1b_1\\a_0b_1+a_1b_0\end{bmatrix}\)

\begin{bmatrix} wa_0+xa_1\\ ya_0+za_1 \end{bmatrix} \cdot \begin{bmatrix} wb_0+xb_1\\ yb_0+zb_1 \end{bmatrix} = \begin{bmatrix} w(a_0b_0+a_1b_1)+x(a_0b_1+a_1b_0)\\ y(a_0b_0+a_1b_1)+z(a_0b_1+a_1b_0) \end{bmatrix}

以XOR為例子

先看2*2的矩陣\(T_2=\begin{bmatrix}w&x\\y&z\end{bmatrix}\)

我們要\(T_2\begin{bmatrix}a_0\\a_1\end{bmatrix}\cdot\) \(T_2\begin{bmatrix}b_0\\b_1\end{bmatrix}=\) \(T_2\begin{bmatrix}c_0\\c_1\end{bmatrix}=\) \(T_2\begin{bmatrix}a_0b_0+a_1b_1\\a_0b_1+a_1b_0\end{bmatrix}\)

\begin{bmatrix} w^2a_0b_0 + wxa_0b_1 + wxa_1b_0 + x^2a_1b_1\\ y^2a_0b_0 + yza_0b_1 + yza_1b_0 + z^2a_1b_1 \end{bmatrix} = \begin{bmatrix} wa_0b_0+xa_0b_1+xa_1b_0+xa_1b_1\\ ya_0b_0+za_0b_1+ya_1b_0+za_1b_1 \end{bmatrix}

以XOR為例子

先看2*2的矩陣\(T_2=\begin{bmatrix}w&x\\y&z\end{bmatrix}\)

我們要\(T_2\begin{bmatrix}a_0\\a_1\end{bmatrix}\cdot\) \(T_2\begin{bmatrix}b_0\\b_1\end{bmatrix}=\) \(T_2\begin{bmatrix}c_0\\c_1\end{bmatrix}=\) \(T_2\begin{bmatrix}a_0b_0+a_1b_1\\a_0b_1+a_1b_0\end{bmatrix}\)

\begin{bmatrix} w^2a_0b_0 + wxa_0b_1 + wxa_1b_0 + x^2a_1b_1\\ y^2a_0b_0 + yza_0b_1 + yza_1b_0 + z^2a_1b_1 \end{bmatrix} = \begin{bmatrix} wa_0b_0+xa_0b_1+xa_1b_0+xa_1b_1\\ ya_0b_0+za_0b_1+ya_1b_0+za_1b_1 \end{bmatrix}
\begin{cases} w^2=w\\ wx=x\\ x^2=w \end{cases}
\begin{cases} y^2=y\\ yz=z\\ z^2=y \end{cases}

得到方程組

以XOR為例子

先看2*2的矩陣\(T_2=\begin{bmatrix}w&x\\y&z\end{bmatrix}\)

我們要\(T_2\begin{bmatrix}a_0\\a_1\end{bmatrix}\cdot\) \(T_2\begin{bmatrix}b_0\\b_1\end{bmatrix}=\) \(T_2\begin{bmatrix}c_0\\c_1\end{bmatrix}=\) \(T_2\begin{bmatrix}a_0b_0+a_1b_1\\a_0b_1+a_1b_0\end{bmatrix}\)

\begin{bmatrix} w^2a_0b_0 + wxa_0b_1 + wxa_1b_0 + x^2a_1b_1\\ y^2a_0b_0 + yza_0b_1 + yza_1b_0 + z^2a_1b_1 \end{bmatrix} = \begin{bmatrix} wa_0b_0+xa_0b_1+xa_1b_0+xa_1b_1\\ ya_0b_0+za_0b_1+ya_1b_0+za_1b_1 \end{bmatrix}

\((w,x)\)和\((y,z)\)皆可為\((0,0),(1,1),(1,-1)\)

以XOR為例子

先看2*2的矩陣\(T_2=\begin{bmatrix}w&x\\y&z\end{bmatrix}\)

我們要\(T_2\begin{bmatrix}a_0\\a_1\end{bmatrix}\cdot\) \(T_2\begin{bmatrix}b_0\\b_1\end{bmatrix}=\) \(T_2\begin{bmatrix}c_0\\c_1\end{bmatrix}=\) \(T_2\begin{bmatrix}a_0b_0+a_1b_1\\a_0b_1+a_1b_0\end{bmatrix}\)

\begin{bmatrix} w^2a_0b_0 + wxa_0b_1 + wxa_1b_0 + x^2a_1b_1\\ y^2a_0b_0 + yza_0b_1 + yza_1b_0 + z^2a_1b_1 \end{bmatrix} = \begin{bmatrix} wa_0b_0+xa_0b_1+xa_1b_0+xa_1b_1\\ ya_0b_0+za_0b_1+ya_1b_0+za_1b_1 \end{bmatrix}

\((w,x)\)和\((y,z)\)皆可為\((0,0),(1,1),(1,-1)\)

因為\(T_2\)可逆,所以\(T_2=\begin{bmatrix}1&1\\1&-1\end{bmatrix}\),發現\(T_2^{-1}=\frac{1}{2}T_2\)

以XOR為例子-將\(T_2\)擴展

\(T_2=\begin{bmatrix}1&1\\1&-1\end{bmatrix}\)

將多項式\(A\)(長度為4)分成兩半\(A_0\),\(A_1\),個別視作單一元素

我們要\(U\begin{bmatrix}T_2 A_0\\T_2 A_1\end{bmatrix}\cdot\) \(U\begin{bmatrix}T_2 B_0\\T_2 B_1\end{bmatrix}=\) \(U\begin{bmatrix}T_2 C_0\\T_2 C_1\end{bmatrix}=\) \(U\begin{bmatrix}T_2A_0B_0+T_2A_1B_1\\T_2A_0B_1+T_2A_1B_0\end{bmatrix}\)

以XOR為例子-將\(T_2\)擴展

\(T_2=\begin{bmatrix}1&1\\1&-1\end{bmatrix}\)

將多項式\(A\)(長度為4)分成兩半\(A_0\),\(A_1\),個別視作單一元素

我們要\(U\begin{bmatrix}T_2 A_0\\T_2 A_1\end{bmatrix}\cdot\) \(U\begin{bmatrix}T_2 B_0\\T_2 B_1\end{bmatrix}=\) \(U\begin{bmatrix}T_2 C_0\\T_2 C_1\end{bmatrix}=\) \(U\begin{bmatrix}T_2A_0B_0+T_2A_1B_1\\T_2A_0B_1+T_2A_1B_0\end{bmatrix}\)

根據之前的經驗可以得到\(U=\begin{bmatrix}I_2&I_2\\I_2&-I_2\end{bmatrix}\),可以分治做事。

還可以得到滿足\(T_4A\cdot\) \(T_4B=\) \(T_4C\)的\(T_4=\begin{bmatrix}T_2&T_2\\T_2&-T_2\end{bmatrix}\)

以XOR為例子-參考code(700ms)

using ll = long long;
const int M = 998244353;
int k;
vector<ll> FWHT (vector<ll>&a) {
    int n = a.size();
    if (n == 1) return a;
    vector<ll> zero(n / 2), one(n / 2);
    for (int i = 0; i < n / 2; i++) zero[i] = a[i], one[i] = a[i + n / 2];
    zero = FWHT(zero), one = FWHT(one);
    for (int i = 0; i < n / 2; i++) a[i] = zero[i] + one[i], a[i + n / 2] = zero[i] - one[i];
    for (int i = 0; i < n; i++) a[i] = (a[i] % M + M) % M;
    return a;
}

int main () {
    cin >> k;
    int n = (1 << k);
    vector <ll> a(n), b(n), c(n);
    for (auto &x : a) cin >> x;
    for (auto &x : b) cin >> x;

    a = FWHT(a), b = FWHT(b);
    for (int i = 0; i < n; i++) c[i] = a[i] * b[i] % M;
    c = FWHT(c);
    for (int i = 0; i < n; i++) c[i] = c[i] * fmul(n, M - 2) % M;

    for (int i = 0; i < n; i++) cout << c[i] << ' ';
    cout << '\n';
}

以XOR為例子-參考code-迭代(200ms)

vector<ll> FWHT (vector<ll>&a) {
    int n = a.size();
    if (n == 1) return a;
    for (int len = 2; len <= n; len <<= 1) {
        int sml = len / 2;
        for (int i = 0; i < n / len; i++) {
            for (int j = 0; j < sml; j++) {
                int old0 = a[i * len + j], old1 = a[i * len + j + sml];
                a[i * len + j] = (old0 + old1) % M;
                a[i * len + j + sml] = (old0 - old1 + M) % M;
            }
        }
    }
    return a;
}

OR, AND之類的也可以這樣先列式處理2*2再擴展到更大

OR的矩陣是\(\begin{bmatrix}1&0\\1&1\end{bmatrix}\)

AND的矩陣是\(\begin{bmatrix}0&1\\1&1\end{bmatrix}\)

瑣事-指稱

  • 這種位元卷積的一般化解法可以叫做FWHT
  • FWHT曾專指XOR卷積

OR卷積和SOS DP

OR卷積的矩陣長得像\(\begin{bmatrix}1&0\\1&1\end{bmatrix}\)

OR卷積的矩陣長得像\(\begin{bmatrix}1&0\\1&1\end{bmatrix}\)

vector<int> FWHT (vector<int>&a) {
    int n = a.size();
    if (n == 1) return a;
    for (int len = 2; len <= n; len <<= 1) {
        int sml = len / 2;
        for (int i = 0; i < n / len; i++) {
            for (int j = 0; j < sml; j++) {
                a[i * len + j + sml] = (a[i * len + j] + a[i * len + j + sml]) % M;
            }
        }
    }
    return a;
}

code可能長的像這樣

OR卷積的矩陣長得像\(\begin{bmatrix}1&0\\1&1\end{bmatrix}\)

vector<int> FWHT (vector<int>&a) {
    int n = a.size();
    if (n == 1) return a;
    for (int len = 2; len <= n; len <<= 1) {
        int sml = len / 2;
        for (int i = 0; i < n / len; i++) {
            for (int j = 0; j < sml; j++) {
                a[i * len + j + sml] = (a[i * len + j] + a[i * len + j + sml]) % M;
            }
        }
    }
    return a;
}

code可能長的像這樣

怎麼有點像SOS ?

OR卷積的矩陣長得像\(\begin{bmatrix}1&0\\1&1\end{bmatrix}\)

vector<int> FWHT (vector<int>&a) {
    int n = a.size();
    if (n == 1) return a;
    for (int len = 2; len <= n; len <<= 1) {
        int sml = len / 2;
        for (int i = 0; i < n / len; i++) {
            for (int j = 0; j < sml; j++) {
                a[i * len + j + sml] = (a[i * len + j] + a[i * len + j + sml]) % M;
            }
        }
    }
    return a;
}

code可能長的像這樣

怎麼有點像SOS ?

因為就真的是SOS

  • OR是SOS(子集和)
  • AND也是SOS(超集合)
  • 現在只考慮OR
  • \(c_k=\sum_{i\lor j = k}a_i b_j\)
  • 現在只考慮OR
  • \(c_k=\sum_{i\lor j = k}a_i b_j\)
  • \({c_k}'=\sum_{(i\lor j)\subseteq k}a_i b_j\)
  • 現在只考慮OR
  • \(c_k=\sum_{i\lor j = k}a_i b_j\)
  • \({c_k}'=\sum_{(i\lor j)\subseteq k}a_i b_j\)
    \( =\sum_{(i\subseteq k) \land (j\subseteq k)} a_ib_j\)
  • 現在只考慮OR
  • \(c_k=\sum_{i\lor j = k}a_i b_j\)
  • \({c_k}'=\sum_{(i\lor j)\subseteq k}a_i b_j\)
    \(=\sum_{(i\subseteq k) \land (j\subseteq k)} a_ib_j\)
    \(=(\sum_{i\subseteq k}a_i)(\sum_{j\subseteq k}b_j)\)
  • 現在只考慮OR
  • \(c_k=\sum_{i\lor j = k}a_i b_j\)
  • \({c_k}'=\sum_{(i\lor j)\subseteq k}a_i b_j\)
    \(=\sum_{(i\subseteq k) \land (j\subseteq k)} a_ib_j\)
    \(=(\sum_{i\subseteq k}a_i)(\sum_{i\subseteq k}b_i)\)
  • 現在只考慮OR
  • \(c_k=\sum_{i\lor j = k}a_i b_j\)
  • \({c_k}'=\sum_{(i\lor j)\subseteq k}a_i b_j\)
    \(=\sum_{(i\subseteq k) \land (j\subseteq k)} a_ib_j\)
    \(=(\sum_{i\subseteq k}a_i)(\sum_{i\subseteq k}b_i)\)
  • 怎麼求\((\sum_{i\subseteq k}a_i)\)、\((\sum_{i\subseteq k}b_i)\)?
  • 現在只考慮OR
  • \(c_k=\sum_{i\lor j = k}a_i b_j\)
  • \({c_k}'=\sum_{(i\lor j)\subseteq k}a_i b_j\)
    \(=\sum_{(i\subseteq k) \land (j\subseteq k)} a_ib_j\)
    \(=(\sum_{i\subseteq k}a_i)(\sum_{i\subseteq k}b_i)\)
  • 怎麼求\((\sum_{i\subseteq k}a_i)\)、\((\sum_{i\subseteq k}b_i)\)? 子集和
  • 現在只考慮OR
  • \(c_k=\sum_{i\lor j = k}a_i b_j\)
  • \({c_k}'=\sum_{(i\lor j)\subseteq k}a_i b_j\)
    \(=\sum_{(i\subseteq k) \land (j\subseteq k)} a_ib_j\)
    \(=(\sum_{i\subseteq k}a_i)(\sum_{i\subseteq k}b_i)\)
  • 怎麼求\((\sum_{i\subseteq k}a_i)\)、\((\sum_{i\subseteq k}b_i)\)? 子集和
  • 怎麼把\({c_k}'\)變回\(c_k\)?
  • 現在只考慮OR
  • \(c_k=\sum_{i\lor j = k}a_i b_j\)
  • \({c_k}'=\sum_{(i\lor j)\subseteq k}a_i b_j\)
    \(=\sum_{(i\subseteq k) \land (j\subseteq k)} a_ib_j\)
    \(=(\sum_{i\subseteq k}a_i)(\sum_{i\subseteq k}b_i)\)
  • 怎麼求\((\sum_{i\subseteq k}a_i)\)、\((\sum_{i\subseteq k}b_i)\)? 子集和
  • 怎麼把\({c_k}'\)變回\(c_k\)? 用子集和的結構來扣

黃色小鴨Pro Max

總結各種經典卷積

模板題

如果聽不懂我在講什麼...

更多令人討厭的...

(想學的回家自己卷的)

🎉恭喜中国台湾竣耀同学勇夺IOI金牌

国家栋梁,保送清华、北大

2 0 2 5

Convolution

By weakweakweak

Convolution

  • 81