What is a random seed in simulation, and why should we care? How was simulation used in the article this week?

A random seed is a parameter used in simulations and random number generation to ensure reproducibility. By setting a specific seed value, the same sequence of random numbers can be generated each time the code is executed, resulting in consistent outputs. Reproducibility is a vital concept in various fields, including data science. The seed function works by saving the state of the random number generator. It captures the previous value generated by the generator, allowing it to produce the same sequence of random numbers across multiple executions of the code. If there is no previous value, the seed function uses the current system time as the seed value. Random seeds are commonly employed in two main tasks:

  1. Splitting data: When dividing data into training, validation, and test sets, random seeds ensure that the data is consistently partitioned in the same way each time the code is run. This helps maintain consistency and enables accurate comparisons between different runs.

  2. Model training: Certain algorithms, such as random forest and gradient boosting, exhibit non-deterministic behavior. This means that given the same input, the output may not always be the same. To obtain reproducible results when training models using these algorithms, a random seed argument is often provided, ensuring consistent behavior across different runs. By utilizing random seeds, researchers and data scientists can achieve reproducibility, allowing for better validation, debugging, and comparison of results in simulations and other data-driven processes.

import random

# Running without seed
print("Without seed:")
## Without seed:
for i in range(3):
    print(random.randint(1, 1000))

# Running with seed
## 817
## 213
## 156
print("\nWith seed:")
## 
## With seed:
random.seed(117)
for i in range(3):
    print(random.randint(1, 1000))
## 246
## 186
## 169

In the example above, we first generate three random numbers between 1 and 1000 without setting a seed. As a result, each execution produces different random numbers. Next, we set the seed to 117 using random.seed(117) With the seed set, subsequent calls to random.randint(1, 1000) will generate the same random number, 117, three times in a row. This demonstrates the reproducibility achieved by using a specific seed value.

How was simulation used in the article this week?

In the featured article this week, Monte Carlo Simulations were employed to estimate the number of aircraft needed to meet the demand on a daily basis. The study utilized iterative sampling and adjustable parameters to explore different scenarios. A total of 250 simulation runs were conducted. The article presents the outcomes of two specific scenarios in Figure 3. Scenario 1 considered a range of 6.0 to 8.0 hours for the daily operational flight hours, while Scenario 2 maintained the same parameter but with a range of 2.5 to 8.0 hours.

Share some code from last week’s exercises that was particularly rewarding or frustrating.

The code that was demonstrated pertains to an optional exercise found in Chapter 2. It aimed to observe the fluctuation of both the “olin” and “wellesley” variables during the simulation.

The Code that is really frustating me right now has to do with Saving the resulting figure.

# Create function run_simulation

def run_simulation(p1, p2, num_steps):
    olin = TimeSeries()
    wellesley = TimeSeries()
    
    for i in range(num_steps):
        step(p1, p2)
        olin[i] = bikeshare.olin
        wellesley[i] = bikeshare.wellesley
        
    plot(olin, label='Olin')
    plot(wellesley, label='Wellesley')
    decorate(title='Wellesley-Olin Bikeshare',
             xlabel='Time step in mins', 
             ylabel='Number of bikes')