从文件加载矩阵 - Python

Question

我在将矩阵从文本文件加载到 Python 时遇到问题。我有一个包含如下矩阵的 txt 文件：

Matrix # 1[6 x 8]
[[0.0322202  0.09434484 0.25002295 0.2528428  0.10415784 0.24142196
0.010517   0.25487851]
[0.18335361 0.1277445  0.06118253 0.18690019 0.20396079 0.28628478
0.23368012 0.07932863]
[0.01586491 0.08351546 0.09179019 0.19086502 0.26448857 0.00341661
0.0076354  0.14970549]
[0.05323177 0.01803223 0.26651485 0.249316   0.00885857 0.28183164
0.06242965 0.10416661]
[0.05393575 0.2685312  0.0928845  0.0165103  0.19793575 0.18197242
0.10990779 0.11711208]
[0.11764279 0.23854231 0.14737164 0.15334971 0.26638431 0.04492217 
0.12121334 0.0157779 ]]
-------------------------
Matrix # 2[8 x 8]
[[0.03673123 0.04280058 0.13064546 0.0483846  0.06306662 0.04791767
0.20136789 0.06709436]
[0.00859638 0.2915551  0.1329647  0.00975984 0.12029034 0.2637118
0.12587069 0.11391991]
[0.05633049 0.08800232 0.03203959 0.02466364 0.10011332 0.15201659
0.22264326 0.0558971 ]
[0.05053821 0.04099701 0.27159803 0.08778437 0.20792823 0.2030534
0.25036928 0.16582882]
[0.16867485 0.03230341 0.19495864 0.10821256 0.12185273 0.05480103
0.22728856 0.25456569]
[0.02218817 0.23359441 0.15457978 0.0275037  0.06745245 0.12328887
0.16972525 0.02161821]
[0.28029231 0.16327778 0.27735648 0.20591421 0.21236012 0.17597595
0.20992926 0.01747133]
[0.10150612 0.22284606 0.11146442 0.18066627 0.12760146 0.10264632
0.24329665 0.26529221]]
-------------------------
Matrix # 3[8 x 5]
[[0.0044761  0.26838514 0.22510378 0.22843132 0.07689473]
[0.16908802 0.15970796 0.25875775 0.04569838 0.04147033]
[0.08524995 0.04703752 0.05619528 0.14943606 0.24411115]
[0.0667661  0.13352421 0.19563742 0.11554089 0.10493734]
[0.14797975 0.06908592 0.06823431 0.04430664 0.09185596]
[0.02574791 0.0367757  0.23516482 0.1551992  0.27722899]
[0.14542998 0.01641985 0.24688273 0.21755754 0.09459343]
[0.26374249 0.12827675 0.1170908  0.004356   0.08593468]]
-------------------------

还有一千个类似的矩阵。

我写了一段 Python 代码，如下所示：

import csv

matrices = []

if __name__ == "__main__":

with open('matrices-2-2.txt', 'r') as file:
    matrix_reader = csv.reader(file, delimiter =" ")
    current_matrix = [];

    for row in matrix_reader:
        if row == "-------------------------":
            matrices.append(current_matrix)
            current_matrix = []
        elif row == "Matrix":
            self()
        else:
            current_matrix.append(list(map(float, row)))

print(matrices)

我设法加载了矩阵。目前，不幸的是，我在尝试将它们相乘时遇到了问题。我创建了以下代码：

import multiprocessing
from multiprocessing import Pool
from time import time
import numpy
import re


def matrix_multiplication(list1, list2):
    A = numpy.matrix(list1)
    B = numpy.matrix(list2)
    return A * B


def counting(dane):
    left_matrix = matrices[0]

    for matrix in matrices[1:]:
        left_matrix = numpy.matrix(left_matrix)
        matrix = numpy.matrix(matrix)
        left_matrix = matrix_multiplication(left_matrix, matrix)

if __name__ == "__main__":

    matrices = []
    start = time()
    with open('matrices-2-2.txt', 'r') as file:
        content = file.read()
        # gets each matrix based on the pattern [[ ]]
    matrices_string = re.findall("\[([\d\s.\n\]\[]+)\]", content)
    # loops over each matrix
    for matrix_string in matrices_string:
        # parses each row of the matrix
        matrix_rows = re.findall("\[([\d\s.\n\]]+)\]", matrix_string)
        # gets all the numbers for each row of the matrix and remakes   the matrix as a list of floats
        matrices.append([list(map(float, re.findall("\d+\.\d+", row))) for row in matrix_rows])

    print('Load Matrices')
    np = multiprocessing.cpu_count()
    print('You have', np, 'processors')
    counting(matrices)
    stop = time()

    print('The Calculation took ', stop - start, 'seconds')

    matrices2 = numpy.array_split(matrices, np)
    start = time()
    pool = Pool(processes=np)
    count = pool.map(counting, matrices2)
    stop = time()

    print("Parallel Calculation took ", stop - start, 'seconds')

编译器显示以下错误：

文件“/.../Matrix1.py”，第 45 行，在 计数（矩阵） 文件“/.../Matrix1.py”，第 21 行，计数 left_matrix = matrix_multiplication(left_matrix, 矩阵) 文件“/.../Matrix1.py”，第 12 行，在 matrix_multiplication 中 返回 (A * B) mul 中的文件“/Library/Frameworks/Python.framework/Versions/3.6/lib/python3.6/site-packages/numpy/matrixlib/defmatrix.py”，第 343 行 返回 N.dot(self, asmatrix(other)) ValueError：形状（6,6）和（5,6）未对齐：6（dim 1）！= 5（dim 0） 请帮忙，我希望这个算法能正常工作。

Answer 1

如前所述，我相信正则表达式可能是解析文本文件的最佳方式。 下面的代码适用于您提供的输入

import re
from pprint import pprint
matrices = []
if __name__ == "__main__":
    with open('matrices-2-2.txt', 'r') as file:
        content = file.read()
    # gets each matrix based on the pattern [[ ]]
    matrices_string = re.findall("\[([\d\s.\n\]\[]+)\]", content)
    #loops over each matrix
    for matrix_string in matrices_string:
        # parses each row of the matrix
        matrix_rows = re.findall("\[([\d\s.\n\]]+)\]", matrix_string)
        # gets all the numbers for each row of the matrix and remakes the matrix as a list of floats
        matrices.append([list(map(float, re.findall("\d+\.\d+", row))) for row in matrix_rows]) 
    pprint(current_matrix)

一段输出：

[[[0.0322202,
   0.09434484,
   0.25002295,
   0.2528428,
   0.10415784,
   0.24142196,
   0.010517,
   0.25487851],
  [0.18335361,
   0.1277445,
   0.06118253,
   0.18690019,
   0.20396079,
   0.28628478,
   0.23368012,
   0.07932863],
  [0.01586491,

要使矩阵相乘，您应该像这样加载一个矩阵： matrix_1 = numpy.array(my_matrix_as_list_1) matrix_2 = numpy.array(my_matrix_as_list_2) 然后你可以像这样进行矩阵乘法： 结果 = numpy.dot(matrix_1, matrix_2)

Answer 2

这些条件row == "-------------------------"、row == "Matrix" 永远不会是真的，因为每一行都是一个数组。

我猜你得到了一个ValueError 异常。

因此，要修复您的代码，您需要按如下方式比较第一行元素line[0] == "matrix" 或"-" in line[0]

你的价值线上也有括号，去掉它。

由于您没有使用有效的 CSV 文件，我认为您可以更改代码以仅遍历文件，当您找到值行时，您可以使用 line.split(' ') 命令，如下例所示：

for line in open('matrix.txt', 'r'):
    if "matrix" in line or "-" in line:
        continue

    print([float(element) for element in line.split(' ')])