Last active
December 17, 2015 08:39
-
-
Save niklasf/5582042 to your computer and use it in GitHub Desktop.
An O(n^2.81) n-x-n-matrix multiplication algorithm.
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
#!/usr/bin/env python | |
# -*- coding: utf-8 -*- | |
import unittest | |
def multiply(A, B): | |
if len(A) == 1: | |
return [[ A[0][0] * B[0][0] ]] | |
elif len(A) % 2 == 1: | |
return shrink(multiply(expand(A), expand(B))) | |
else: | |
A00, A01, A10, A11 = split_into_blocks(A) | |
B00, B01, B10, B11 = split_into_blocks(B) | |
M1 = multiply(add(A00, A11), add(B00, B11)) | |
M2 = multiply(add(A10, A11), B00) | |
M3 = multiply(A00, add(B01, B11, -1)) | |
M4 = multiply(A11, add(B10, B00, -1)) | |
M5 = multiply(add(A00, A01), B11) | |
M6 = multiply(add(A10, A00, -1), add(B00, B01)) | |
M7 = multiply(add(A01, A11, -1), add(B10, B11)) | |
return join_blocks(add(M1, add(M4, add(M7, M5, -1))), add(M3, M5), | |
add(M2, M4), add(M1, add(M3, add(M6, M2, -1)))) | |
def add(A, B, l = 1): | |
C = [] | |
for x in range(len(B)): | |
row = [] | |
for y in range(len(B)): | |
row.append(A[x][y] + l * B[x][y]) | |
C.append(row) | |
return C | |
def split_into_blocks(A): | |
n = len(A) | |
A00 = [] | |
A01 = [] | |
A10 = [] | |
A11 = [] | |
for i in range(n / 2): | |
A00.append(A[i][0:n/2]) | |
A10.append(A[i + n/2][0:n/2]) | |
A01.append(A[i][n/2:n]) | |
A11.append(A[i + n/2][n/2:n]) | |
return (A00, A01, A10, A11) | |
def join_blocks(A00, A01, A10, A11): | |
A = [] | |
for i in range(len(A00)): | |
A.append(A00[i] + A01[i]) | |
for i in range(len(A10)): | |
A.append(A10[i] + A11[i]) | |
return A | |
def expand(A): | |
result = [] | |
for row in A: | |
result.append(row[:] + [0]) | |
result.append([0] * (len(row) + 1)) | |
return result | |
def shrink(A): | |
result = [] | |
for i in range(len(A) - 1): | |
result.append(A[i][:-1]) | |
return result | |
class MatrixTestCase(unittest.TestCase): | |
def test_add(self): | |
A = [[1, 2], | |
[3, 4]] | |
B = [[4, 3], | |
[2, 1]] | |
Five = [[5, 5], | |
[5, 5]] | |
self.assertEqual(add(B, A), Five) | |
self.assertEqual(add(A, B), add(B, A)) | |
self.assertEqual(add(A, B), Five) | |
def test_joining_and_splitting_blocks(self): | |
A00 = [[1, 2], | |
[3, 4]] | |
A01 = [[-1, -2], | |
[-3, -4]] | |
A10 = [[5, 6], | |
[7, 8]] | |
A11 = [[-5, -6], | |
[-7, -8]] | |
Joined = [[1, 2, -1, -2], | |
[3, 4, -3, -4], | |
[5, 6, -5, -6], | |
[7, 8, -7, -8]] | |
self.assertEqual(join_blocks(A00, A01, A10, A11), Joined) | |
B00, B01, B10, B11 = split_into_blocks(Joined) | |
self.assertEqual(B00, A00) | |
self.assertEqual(B01, A01) | |
self.assertEqual(B10, A10) | |
self.assertEqual(B11, A11) | |
def test_expanding_and_shrinking(self): | |
A = [[1, 2, 3], | |
[3, 2, 1], | |
[2, 1, 3]] | |
A_expanded = [[1, 2, 3, 0], | |
[3, 2, 1, 0], | |
[2, 1, 3, 0], | |
[0, 0, 0, 0]] | |
self.assertEqual(expand(A), A_expanded) | |
def test_2x2(self): | |
id_2 = [[1, 0], | |
[0, 1]] | |
A = [[1, 2], | |
[3, 4]] | |
B = [[3, 2], | |
[6, 7]] | |
AxB = [[15, 16], | |
[33, 34]] | |
self.assertEqual(multiply(id_2, A), A) | |
self.assertEqual(multiply(B, id_2), B) | |
self.assertEqual(multiply(A, B), AxB) | |
def test_3x3(self): | |
id_3 = [[1, 0, 0], | |
[0, 1, 0], | |
[0, 0, 1]] | |
A = [[1, 2, 3], | |
[4, 5, 6], | |
[7, 8, 9]] | |
AxA = [[ 30, 36, 42], | |
[ 66, 81, 96], | |
[102, 126, 150]] | |
self.assertEqual(multiply(id_3, A), A) | |
self.assertEqual(multiply(A, id_3), A) | |
self.assertEqual(multiply(A, A), AxA) | |
def test_4x4(self): | |
id_4 = [[1, 0, 0, 0], | |
[0, 1, 0, 0], | |
[0, 0, 1, 0], | |
[0, 0, 0, 1]] | |
A = [[ 1, 2, 3, 4], | |
[ 5, 6, 7, 8], | |
[ 9, 10, 11, 12], | |
[13, 14, 15, 16]] | |
self.assertEqual(multiply(id_4, id_4), id_4) | |
self.assertEqual(multiply(id_4, A), A) | |
if __name__ == '__main__': | |
unittest.main() |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment