This post provides an introductionary documentation for getting started with NumPy in Python.

# importing numpy into this python script import numpy # building some lists and adding them with Python base library a = [1,2,3] b = [1,2,3] a+b

[1, 2, 3, 1, 2, 3]

# declaring an array with numpy a_np = numpy.array([1,2,3]) # checking the type fo the array to confirm it is a numpy array type(a_np)

numpy.ndarray

# numpy arrays have an attribute .dtype, allowing you to check the numpy data type of the array a_np.dtype

dtype('int32')

# a_np is of type integer; let try to assign a float a_np[0] = 13.33 a_np

array([13, 2, 3])

# numpy arrays have a .ndim attribute, allowing you to see the number of dimensions of the array a_np.ndim

1

# numpy arrays also have a .size attribute; # .size is allowing you to see the number of elements contained by the array a_np.size

3

# math becomes possible element-wise, when using numpy arrays # -- creating a second numpy array to test some element-wise mathematical operations b_np = numpy.array([2,3,4]) # addition a_np + b_np

array([11, -1, -1])

# subtraction a_np - b_np

array([11, -1, -1])

# multiplication a_np * b_np

array([26, 6, 12])

# one to the power of the other a_np ** b_np

array([169, 8, 81], dtype=int32)

# divison a_np / b_np

array([6.5 , 0.66666667, 0.75 ])

# modulo; testing a_np modulo 2 a_np % numpy.array([2,2,2])

array([1, 0, 1], dtype=int32)

# modulo; testing b_np modulo 2 b_np % numpy.array([2,2,2])

array([0, 1, 0], dtype=int32)

# numpy allows for using "universal functions"; # universal functions work element-wise, with numpy arrays # one example is .sin numpy.sin(a_np)

array([0.42016704, 0.90929743, 0.14112001])

# another example is .sqrt numpy.sqrt(a_np)

array([3.60555128, 1.41421356, 1.73205081])

# building a 2D data table using numpy c_np = numpy.array([[1,2,3], [4,5,6]]) c_np

array([[1, 2, 3], [4, 5, 6]])

# indexing, i.e. obtaining the value stored in a defined cell; # here: cell in first row, first colum c_np[0][0]

1

# here: cell in last row, last colum c_np[-1][-1]

6

# here: cell in second row, first column c_np[1][0]

4

# an alternative way of doing the above c_np[1,0]

4

# slicing through simple 1d numpy array # here: get second and third value in d_np d_np = numpy.array([7,9,11]) d_np[1:3]

array([ 9, 11])

# get first and third element d_np[0:3:2] # from:to:step

array([ 7, 11])

# here: getting every other element in d_np d_np[::2]

array([ 7, 11])

# here: getting every third element d_np[::3]

array([7])

# we can create a 1D array using .arange, a method from the numpy module e_np = numpy.arange(0,26) #param1: from; param2: to (last value excluded) e_np

array([ 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25])

# let's check the type of e_np type(e_np)

numpy.ndarray

# numpy arrays can be re-shaped, using the numpy .reshape method f_np = e_np.reshape(2,13) #param1: number of rows; param2: number of columns f_np

array([[ 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12], [13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25]])

# I can use .reshape not just for creating 2D out of 1D; # e.g. see this g_np = numpy.array([1,2,3,4,5,6,7,8,9,10,11,12]).reshape(3,2,2) g_np

array([[[ 1, 2], [ 3, 4]], [[ 5, 6], [ 7, 8]], [[ 9, 10], [11, 12]]])

# g_np is nor 3D; let slice some dimensions out of this data cube # all values in first directions of second dimension g_np[:,0,:]

array([[ 1, 2], [ 5, 6], [ 9, 10]])

# one specific cell; first in all dimensions g_np[0,0,0]

1

# one specific cell; last in all dimensions g_np[-1,-1,-1]

12

# one specific cell; SECOND last in all dimensions g_np[-2,-2,-2]

5

# let's complete this first tutorial on numpy with another example; # I create a 2d array and show it to you example = numpy.array([[1,2,3,4,5], [1.1,2.2,3.3,4.4,5.5], [1.11,2.22,3.33,4.44,5.55], [1.111,2.222,3.333,4.444,5.555], [1.1111,2.2222,3.3333,4.4444,5.5555]]) example

array([[1. , 2. , 3. , 4. , 5. ], [1.1 , 2.2 , 3.3 , 4.4 , 5.5 ], [1.11 , 2.22 , 3.33 , 4.44 , 5.55 ], [1.111 , 2.222 , 3.333 , 4.444 , 5.555 ], [1.1111, 2.2222, 3.3333, 4.4444, 5.5555]])

# now lets access the values 2.2, 4.4, 2.222, and 4.444 - in just one line of code example[1::2,1::2]

array([[2.2 , 4.4 ], [2.222, 4.444]])

# now some examples what happens when you # do copying by reference a_np = numpy.array([1,2,3]) b_np = a_np[:1] b_np

array([1])

# changing b_np element b_np[0] = 10000 b_np

array([10000])

# let's npw check a_np a_np

array([10000, 2, 3])

# above is the effect of copying by reference; # this happens when we slice through numpy arrays; # does it happen with standard python list? a = [1,2,3] b = a[:1] b[0] = 1000 a

[1, 2, 3]

# copying by reference does not happen when slicing default lists in pyhton; # if we do not want to copy by reference then we can use .copy() method, as part of numpy c_np = numpy.array([1,2,3,4,5]) d_np = c_np.copy() d_np[0] = 10000 c_np[0] = 99999 # let's check c_np now print("this is c_np:" + str(c_np))

this is c_np:[99999 2 3 4 5]

# what about d_np? d_np

array([10000, 2, 3, 4, 5])

# copying with .copy() allows to work with two objects, with independent references;

Data scientist focusing on simulation, optimization and modeling in R, SQL, VBA and Python

## Leave a Reply