Ayush@Machine
/root/latent-space/topics/resume
Numpy Basics

Numpy Basics

Why Numpy over Lists?

  1. Speed of computation is faster
  2. Designed for data analysis
  3. Vector operations can be performed
  4. Concise and easy to read/write

Learn By Examples

Create a 3*3 array using predefined lists

import numpy as np

list_1 = [1,2,3] 
list_2 = [4,5,6] 
list_3 = [7,8,9]

out = np.array([list_1, list_2, list_3])

print(out)

Element Wise Multiplication

import numpy as np

list_1 = [1,2,3] 
list_2 = [4,5,6] 

arr1 = np.array(list_1)
arr2 = np.array(list_2)

mul = arr1 * arr2
print(mul)

list_mul = list(mul)
print(list_mul)

Create array of given dimension filled with given integer value

import numpy as np

rows = 5
columns = 3
value = 15

# using np.full

arr = np.full((rows, columns), value)
print(arr)

# using np.tile
arr = np.tile(value, (rows, columns))
print(arr)

# using np.ones + multiplicationm
# note that here we need to specify dtype=int as float is default for this function
arr = np.ones((rows, columns), dtype=int)*value
print(arr)

Create array filled with multiples of given number

import numpy as np

start = 5
stop = 51
step = 5

arr = np.arange(start, stop, step)
print(arr)

Create a checkerboard matrix of size n*n

import numpy as np

board_size = 5
out = np.tile([[0, 1], [1, 0]], (board_size//2, board_size//2))
print(out)

Explanation

  • np.tile can be used to repeat an array any number of times across any number of dimensions
  • Checkerboard's smallest unit is 
    [0, 1], 
[1, 0]
  • Since smallest unit is already 2x2, whatever is this required board_size, we will divide it by 2

Reference

import numpy as np
list_1 = [10,11,12,13]
list_2 = [15,12,13,14]

arr = np.array([list_1, list_2])

print(arr)
print('shape:', np.shape(arr))
print('dimension:', np.ndim(arr))
print('data type:', arr.dtype)
print('length of one array element (bytes):', arr.itemsize)

Extract all elements of 2nd column in 2D array

arr = np.arange(24).reshape(6, 4)
print(arr)
print(arr[:, 1])

Extract border row, column of numpy array

import numpy as np
array_2d = np.arange(24).reshape(6, 4)
col_first = array_2d[:, 0]
row_first = array_2d[0, :]
col_last = array_2d[:, array_2d.shape[1]-1]
row_last = array_2d[array_2d.shape[0]-1, :]

print(col_first)
print(row_first)
print(col_last)
print(row_last)

Binomial Distribution

Problem Statement

Generate a binomial distribution, tested 10 times, given the number of trials(n) and probability(p) of each trial.

Input

The input will contain seed, n and p in the same order.

Output

The output should contain a numpy array with 10 numbers representing the required binomial distribution.

import numpy as np 
# seed=int(input())
# n=int(input())
# p=float(input())
seed=10
n=10
p=0.5
np.random.seed(seed)
s = np.random.binomial(n, p, 10)
print(s)

Conclusion

  1. Numpy arrays are significantly faster than lists when it comes to mathematical operations
  2. Numpy operations are concise, and easy to read & write.
  3. Numpy arrays are built for vector mathematics so that element wise calculations can be easily carried out.
  4. While in lists, one would have to write a for-loop or list comprehension for basic things like squaring all numbers in the list, the equivalent on a numpy array will simply be "array_name * 2".

References


Tagged in pythondata-analysisnumpy