Numerical Python

Similar to pythons list object.

• All objects in an ndarray must have same datatype
• All math operators work on each individual element in an ''ndarray'

# Properties
a = np.ndarray((2,3,4))
 
a.shape      # Tuple describing shape, e.g. (2,3,4)
a.ndim       # Number of dimensions, equal to len(a.shape), e.g. 3
a.size       # Total number of elements
a.dtype      # Datatype of the elements
a.itemsize   # Size in byte of an element

import numpy as np
 
#### np.array(content) ####
np.array(4)
# array(4)
 
np.array((1,2,3,4))
# array([1, 2, 3, 4])
 
np.array(((1,2,3),(4,5,6)))
# array([[1, 2, 3],
#        [4, 5, 6]])
 
#### np.ndarray(shape) ####
np.ndarray(4)
# array([1.11949605e-311, 1.37961913e-306, 9.42677252e-321, 1.11829178e-311])
 
np.ndarray((2,4))
# array([[0.   , 0.03 , 0.215, 0.4  ],
#        [0.586, 0.77 , 0.954, 1.   ]])

import numpy as np
 
a1 = np.zeros((4,3,2)) # Create 3-dimensional array with zeros
 
a2 = np.ones((3,4,5)) # Create 3-dimensional array with ones
 
a3 = np.arange(1,10) # Array with values 1..9
a4 = np.arange(10) # Array with values 0..9
a5 = np.arange(1,20,5) # Each 5th value between 1 to 19; [1, 6, 11, 16]

a = np.arange(10)   # a => [0, 1, 2, 3, 4, 5, 6, 7, 8, 9]
a1 = a[:6:2]        # a1 => [0,2,4]
a[:6:2] = 30        # a => [30, 1, 30, 3, 30, 5, 6, 7, 8, 9]
 
# Note that a1 holds a reference to a, changes made to a1 is reflected in a

# 2d-array
B = A[row_start: row_stop : (row_step) , col_start : col_stop : (col_step)] 
 
# Note that B holds a reference to A, changes made to B is reflected in A

# Using a list
a = np.arange(10)   # a => [0, 1, 2, 3, 4, 5, 6, 7, 8, 9]
 
# a list naming the indices
l = [0, 2, 8, 9]
a[l]
a[[0,2,8,9]]
 
# or a list of bool, masking desired elements. Must be same dimension as the array.
bl = [False]*10
bl[3] = True
a[bl]            # => [3]
 
# Note that this spawns a new ndarray, changes made to the resulting array is NOT reflected in the original array

# Using comparisons to create a masking list
a = np.arange(10)   # a => [0, 1, 2, 3, 4, 5, 6, 7, 8, 9]
 
a[a>5]           # => [6, 7, 8, 9]
 
# Note that this spawns a new ndarray, changes made to the resulting array is NOT reflected in the original array

Official documentation

Math functions

np.dot(A,B)        # Dot-product, sv skalärprodukt
np.linalg.det(A)   # Determinant
np.transpose(A)    # Transpose
np.identity(n)     # N-th identity matrix
np.linalg.inv(A)   # Inverse matrix
 
# Let v represent a polynomial
np.roots(v)        # roots to the polynom

Official documentation

Functions in the numpy library that operates on ndarrays in an element-wise manner.

import numpy as np
na = np.array([0.23, 2.3, 3.2, 4.1, 5.6])
np.cos(na) # Returns a new ndarray of same dimensions, with the result
np.log10(na)
# etc..
 
# Also arithmetic functions. Given nparray A and B of same size
elementwise_sum = A+B
elementwise_division = A/B
elementwise_square = A**B
 
elementwise_test = A != B
elementwise_test = A > B

Of these four function, only resize creates a new array whilst the rest of them returns references or modify the original array.

A = np.ndarray((3,4)) # 3x4 array
B = A.flatten()       # B is a new array
C = A.ravel()         # C is a reference to A
 
 
A = np.arange(1,16,2)
A.resize(2,4)      # Changes A to a 2x4 array
 
B = A.reshape(4,2) # B is a reference to A

np.vstack((A,B))
np.hstack((A,B))
np.concatenate((A,B), axis=1) # Similar to hstack
# All of these creates new arrays