Python random.seed() -A Deep Dive

by

[toc]

Introduction

random is an in-built module in Python which generates pseudo-random numbers. Now, the random data generated by this module is not completely random. Instead it is pseudo-random, as mentioned previously.

Note: A “True Random Number” can be generated by a TRNG (true random number generator) while a “pseudo-random number” is generated by a PRNG (pseudorandom number generator).

⚠️ TRNG is outside the scope of the discussion in this article.

So, what is a PRNG (pseudorandom number generator)?

PRNG initially generates a random number known as seed. Then an algorithm is used to generate a pseudo-random sequence of bits based on it. In simple words, it is an algorithm that generates seemingly random numbers; however, these numbers are still reproducible.

The random module has a set of methods that help us to generate random elements (numbers). In this tutorial, we will be focusing on the seed() method of the random module.

Random seed() Method in Python

The random number generator needs a starting point, i.e., it needs a seed value to start generating a sequence of random numbers. Thus, it is the seed() method that is used to initialize the random number generator.

By default, current system time is used by the random number generator as a start-point. To customize the start number of the random number generator, you must use the seed() method.

Syntax:

Example:

import random

random.seed(10)
print(random.random())

Output:

0.5714025946899135

How To Generate The Same Random Integer Every Time?

If you set the same seed value before calling any random module function, you will get the same number repeatedly.

Example:

import random

for i in range(3):
    # setting seed value to 10
    random.seed(10)
    print("i[{}]={}".format(i,random.randint(12, 30)))

Output:

i[0]=30
i[1]=30
i[2]=30

Explanation: In the above output, we got the same number as the output because the same seed was set before using randint every time.

random.seed() and random.choice()

➥ choice() is a method of the random module that selects a random element from a specified sequence (string, range, list, tuple).

You can use a custom seed value to receive the same choice value again and again. Let’s have a look at the following example.

Example

import random

x = "PNEUMONOULTRAMICROSCOPICSILICOVOLCANOCONIOSIS"
print("Output Without Setting A Seed: ")
for i in range(3):
    print(random.choice(x))
print("Output After Setting A Seed: ")
for i in range(3):
    random.seed(5)
    print(random.choice(x))

Output:

Output Without Setting A Seed: 
R
C
N
Output After Setting A Seed: 
N
N
N

What is NumPy Random Seed?

The np.random.seed function provides a seed value, i.e., a base input value to NumPy's pseudo-random number generator in Python.

Syntax:

Example 1:

import numpy as np

for i in range(3):
    np.random.seed(101)
    print('i[{}]={}'.format(i, np.random.randint(low=1, high=10, size=10)))

Output:

i[0]=[2 7 8 9 5 9 6 1 6 9]
i[1]=[2 7 8 9 5 9 6 1 6 9]
i[2]=[2 7 8 9 5 9 6 1 6 9]

Here’s another example for you to visualize the effects of numpy.random.seed.

Example 2:

import numpy as np

print("GENERATE SAME RANDOM NUMBER WITH NUMPY.RANDOM.RANDOM")
for i in range(2):
    np.random.seed(101)
    print('i[{}]={}'.format(i, np.random.random()))

print("\nSELECT A RANDOM SAMPLE FROM AN INPUT ARRAY")
for i in range(2):
    np.random.seed(0)
    print('i[{}]={}'.format(i, np.random.choice(a=[1, 2, 3, 4, 5, 6], size=5)))

Output:

GENERATE SAME RANDOM NUMBER WITH NUMPY.RANDOM.RANDOM
i[0]=0.5163986277024462
i[1]=0.5163986277024462

SELECT A RANDOM SAMPLE FROM AN INPUT ARRAY
i[0]=[5 6 1 4 4]
i[1]=[5 6 1 4 4]

Application of numpy.random.seed

  1. Machine Learning
    • Splitting datasets into test set and training sets require random sampling. And random sampling, in turn, requires pseudo-random random numbers. Therefore if you play around with ML models, then Numpy’s random.seed() is almost a certainty!
  2. Deep Learning
    • Just like ML problems, Deep Learning problems also require splitting the dataset into test set and training set with the help of pseudo-random numbers.
  3. Random Sampling
  4. Probability and Statistics

Frequently Asked Questions

Should I Use numpy.random.seed or random.seed?

  • The answer to this question depends on whether you are using Numpy’s random generator in your code or the one in the normal random module.
  • The random generators in random and numpy.random have completely different/separate internal states. This means random.seed() won’t affect the random sequences generated by numpy.random.randn(), etc. Similarly, numpy.random.seed() will not affect the random sequences generated by random.random(), etc.
  • In case you have used both numpy.random and random in your code, then you have to separately set seeds for both.

What Number Should I Use in random.seed?

It does not matter what number you use within the numpy.random.seed() method. Using different seeds will only cause Random module (or Numpy in case of numpy.random.seed) to generate different pseudo-random numbers. Thus, the output of a random function depends on the value of random.seed() but the choice of the seed value is arbitrary.

Example:

import random

print("SEED VALUE = 10")
for i in range(3):
    # setting seed value to 10
    random.seed(10)
    print("i[{}]={}".format(i,random.randint(12, 30)))
print("SEED VALUE = 15")
for i in range(3):
    # setting seed value to 15
    random.seed(15)
    print("i[{}]={}".format(i,random.randint(12, 30)))

Output:

SEED VALUE = 10
i[0]=30
i[1]=30
i[2]=30
SEED VALUE = 15
i[0]=18
i[1]=18
i[2]=18

How Do I Get random.seed() to Use System Time?

Since time keeps changing, hence using it as a seed value to generate random numbers will ensure that the seed value keeps changing and you will get a different random sequence/number upon every execution.

Example:

import random
import time

random.seed(int(time.time()))
c = 'abcdefghijklmnopqrstuvwxyz0123456789%^*(-_=+)'
password = ''.join([c[random.randint(0, len(c) - 1)] for i in range(10)])
print("New Password: ", password)

Output:

New Password:  za2arj+hjz

Conclusion

I hope this article helped you to understand the importance and uses of random.seed in Python. Please subscribe and stay tuned for more interesting concepts. Happy coding!? 

The Complete Guide to PyCharm
  • Do you want to master the most popular Python IDE fast?
  • This course will take you from beginner to expert in PyCharm in ~90 minutes.
  • For any software developer, it is crucial to master the IDE well, to write, test and debug high-quality code with little effort.

Join the PyCharm Masterclass now, and master PyCharm by tomorrow!