NumPy

Introduction

• NumPy stands for numerical Python and it turns data into numbers that the computer can analyze to find patterns
• We use NumPy because it’s fast and it’s optimized using C in the background

Data Types and Attributes

ndarray: n-dimensional array

• .ndim: number of dimensions
• .dtype: datatype
• .shape: tells you the shape (row, column, beyond)
• .size: how many elements in the array

import numpy as np
# 1 dimensional array 
a1 = np.array([1,2,3])
# 2 dimensional array
a2 = np.array([[1,2,3],
			   [4,5,6]])
# 3 dimensional array 
a3 = np.array([[[1,2,3],
			   [4,5,6], 
			   [7,8,9]],
			   [[10,11,12],
			   [13,14,15],
			   [16,17,18]]])

Creating arrays

• shift + tab in jupyter notebook will show you the info from the documentation
• np.ones(shape): array in the shape specified filled with ones
• np.zeros(shape): array in the shape specified filled with zeros
• np.arange(start,stop,step)
• np.random.randint(low,high,shape)
• np.random.random(shape)

Random Seed

• Note: random numbers are pseudo-random
• np.random.seed(seed=num): allows you to randomly generate things in a reproducible way

Viewing arrays and matrices

• np.unique(array): returns the unique values of the array
• You can index and slice the arrays like you would normally in Python lists
• Tip: the last number in the shape is usually the inner-most array size

Manipulating & comparing arrays

• arr1 + arr2 or np.add(arr1, arr2): adds the two arrays together
• Most standard arithmetic is compatible with arrays as well
• Important note: Not all shapes are compatible for arithmetic though so keep that in mind!
  • The term for shape compatibility is broadcasting
• Dimensions are compatible if:
  • they are equal to each other
  • one of the numbers is 1

Aggregation

• Aggregation is performing the same operation on a number of objects
• sum(arr1) vs np.sum(arr1)
  • Use the Python operation on Python datatypes and the NumPy operations on NumPy operations
• np.mean()
• np.max()
• np.min()
• np.std(): standard deviation
  • Standard deviation: measure of how spread out a group of numbers is from the mean (square root of the variance)
• np.var(): variance
  • Variance: measure of the average degree to which each number is different to the mean (ie: Higher variance = wider range of numbers)

Reshaping & transposing arrays

• .reshape(shape): reshapes the array to the desired shape
• .T: transposes the array meaning that it swaps the axises

Dot product

• np.dot(arr1, arr2): dot product aka matrix multiplication
  • Requirement: the inside numbers of the shape have to be equal
  • Result: produces a matrix that’s the shape of the outside numbers

Sorting arrays

• np.sort(arr): sorts each row
• np.argsort(arr): tells you the index the value will be when sorted
• np.argmin(arr,axis), np.argmax(arr,axis): finds the minimum or maximum value in the specified axis. Axises are “flipped”:
  • axis=0 looks at the columns
  • axis=1 looks at the rows

Practical example: turn an image into NumPy array

from matplotlib.image import imread
# turn panda image into an array 
panda = imread("file-name.png")

Other methods & functions

• np.linespace() returns evenly spaced numbers over a specified interval
• np.random.radn(size) creates a data set that has a normal distribution