mirror of
https://github.com/TheAlgorithms/Python.git
synced 2024-11-24 05:21:09 +00:00
4e0717c3cf
* Update matrix_operation.py 1. Adding error checks for integer inputs 2. Adding error checks for matrix operations where size requirements do not match up 3. Added matrix subtraction function 4. included error check so only integer is passed into identity function * Create test_matrix_operation.py * Update matrix_ops and Add Test Cases 1. Included error checks in matrix operation. There were some cases where the functions would not work correctly. 2. PEP8 changes to matrix_operations.py 3. added test cases for matrix operations using pytest. * Update pytest.ini Add carriage return to end of file
145 lines
4.4 KiB
Python
145 lines
4.4 KiB
Python
"""
|
|
function based version of matrix operations, which are just 2D arrays
|
|
"""
|
|
|
|
|
|
def add(matrix_a, matrix_b):
|
|
if _check_not_integer(matrix_a) and _check_not_integer(matrix_b):
|
|
rows, cols = _verify_matrix_sizes(matrix_a, matrix_b)
|
|
matrix_c = []
|
|
for i in range(rows[0]):
|
|
list_1 = []
|
|
for j in range(cols[0]):
|
|
val = matrix_a[i][j] + matrix_b[i][j]
|
|
list_1.append(val)
|
|
matrix_c.append(list_1)
|
|
return matrix_c
|
|
|
|
|
|
def subtract(matrix_a, matrix_b):
|
|
if _check_not_integer(matrix_a) and _check_not_integer(matrix_b):
|
|
rows, cols = _verify_matrix_sizes(matrix_a, matrix_b)
|
|
matrix_c = []
|
|
for i in range(rows[0]):
|
|
list_1 = []
|
|
for j in range(cols[0]):
|
|
val = matrix_a[i][j] - matrix_b[i][j]
|
|
list_1.append(val)
|
|
matrix_c.append(list_1)
|
|
return matrix_c
|
|
|
|
|
|
def scalar_multiply(matrix, n):
|
|
return [[x * n for x in row] for row in matrix]
|
|
|
|
|
|
def multiply(matrix_a, matrix_b):
|
|
if _check_not_integer(matrix_a) and _check_not_integer(matrix_b):
|
|
matrix_c = []
|
|
rows, cols = _verify_matrix_sizes(matrix_a, matrix_b)
|
|
|
|
if cols[0] != rows[1]:
|
|
raise ValueError(f'Cannot multiply matrix of dimensions ({rows[0]},{cols[0]}) '
|
|
f'and ({rows[1]},{cols[1]})')
|
|
for i in range(rows[0]):
|
|
list_1 = []
|
|
for j in range(cols[1]):
|
|
val = 0
|
|
for k in range(cols[1]):
|
|
val = val + matrix_a[i][k] * matrix_b[k][j]
|
|
list_1.append(val)
|
|
matrix_c.append(list_1)
|
|
return matrix_c
|
|
|
|
|
|
def identity(n):
|
|
"""
|
|
:param n: dimension for nxn matrix
|
|
:type n: int
|
|
:return: Identity matrix of shape [n, n]
|
|
"""
|
|
n = int(n)
|
|
return [[int(row == column) for column in range(n)] for row in range(n)]
|
|
|
|
|
|
def transpose(matrix, return_map=True):
|
|
if _check_not_integer(matrix):
|
|
if return_map:
|
|
return map(list, zip(*matrix))
|
|
else:
|
|
# mt = []
|
|
# for i in range(len(matrix[0])):
|
|
# mt.append([row[i] for row in matrix])
|
|
# return mt
|
|
return [[row[i] for row in matrix] for i in range(len(matrix[0]))]
|
|
|
|
|
|
def minor(matrix, row, column):
|
|
minor = matrix[:row] + matrix[row + 1:]
|
|
minor = [row[:column] + row[column + 1:] for row in minor]
|
|
return minor
|
|
|
|
|
|
def determinant(matrix):
|
|
if len(matrix) == 1:
|
|
return matrix[0][0]
|
|
|
|
res = 0
|
|
for x in range(len(matrix)):
|
|
res += matrix[0][x] * determinant(minor(matrix, 0, x)) * (-1) ** x
|
|
return res
|
|
|
|
|
|
def inverse(matrix):
|
|
det = determinant(matrix)
|
|
if det == 0:
|
|
return None
|
|
|
|
matrix_minor = [[] for _ in range(len(matrix))]
|
|
for i in range(len(matrix)):
|
|
for j in range(len(matrix)):
|
|
matrix_minor[i].append(determinant(minor(matrix, i, j)))
|
|
|
|
cofactors = [[x * (-1) ** (row + col) for col, x in enumerate(matrix_minor[row])] for row in range(len(matrix))]
|
|
adjugate = transpose(cofactors)
|
|
return scalar_multiply(adjugate, 1/det)
|
|
|
|
|
|
def _check_not_integer(matrix):
|
|
try:
|
|
rows = len(matrix)
|
|
cols = len(matrix[0])
|
|
return True
|
|
except TypeError:
|
|
raise TypeError("Cannot input an integer value, it must be a matrix")
|
|
|
|
|
|
def _shape(matrix):
|
|
return list((len(matrix), len(matrix[0])))
|
|
|
|
|
|
def _verify_matrix_sizes(matrix_a, matrix_b):
|
|
shape = _shape(matrix_a)
|
|
shape += _shape(matrix_b)
|
|
if shape[0] != shape[2] or shape[1] != shape[3]:
|
|
raise ValueError(f"operands could not be broadcast together with shape "
|
|
f"({shape[0], shape[1]}), ({shape[2], shape[3]})")
|
|
return [shape[0], shape[2]], [shape[1], shape[3]]
|
|
|
|
|
|
def main():
|
|
matrix_a = [[12, 10], [3, 9]]
|
|
matrix_b = [[3, 4], [7, 4]]
|
|
matrix_c = [[11, 12, 13, 14], [21, 22, 23, 24], [31, 32, 33, 34], [41, 42, 43, 44]]
|
|
matrix_d = [[3, 0, 2], [2, 0, -2], [0, 1, 1]]
|
|
print('Add Operation, %s + %s = %s \n' %(matrix_a, matrix_b, (add(matrix_a, matrix_b))))
|
|
print('Multiply Operation, %s * %s = %s \n' %(matrix_a, matrix_b, multiply(matrix_a, matrix_b)))
|
|
print('Identity: %s \n' %identity(5))
|
|
print('Minor of %s = %s \n' %(matrix_c, minor(matrix_c, 1, 2)))
|
|
print('Determinant of %s = %s \n' %(matrix_b, determinant(matrix_b)))
|
|
print('Inverse of %s = %s\n'%(matrix_d, inverse(matrix_d)))
|
|
|
|
|
|
if __name__ == '__main__':
|
|
main()
|