Running a Connections Engine Experiment

Set Up Local Environment

1. Change into the infection_simulation directory.
2. Check out the develop branch with git checkout develop.
3. Change into the experiments/connection_engine directory.

Submitting a Job

NOTE:

• It is best to run the experiment in a TMUX window so that you can close your connection and come back to it later. See TMUX Tutorial for instructions on doing that.

Determine Parameters

• To see what parameters are available, and their respecitve flags, execute the following:

  python3.7 connections_experiment.py --help
  

• You should see the following:

usage: connections_experiment.py [-h] [-np NUM_PEOPLE [NUM_PEOPLE ...]]
                                 [-mc MEAN_CONNECTIONS [MEAN_CONNECTIONS ...]]
                                 [-sd STD] [-s SIZE] [-nr NUM_RUNS]

optional arguments:
  -h, --help            show this help message and exit
  -np NUM_PEOPLE [NUM_PEOPLE ...], --num_people NUM_PEOPLE [NUM_PEOPLE ...]
                        single value or list for number of people
  -mc MEAN_CONNECTIONS [MEAN_CONNECTIONS ...], --mean_connections MEAN_CONNECTIONS [MEAN_CONNECTIONS ...]
                        single value or list for nmean umber of connections
  -sd STD, --std STD    standard deviation for connection distribution
  -s SIZE, --size SIZE  sample size for connection distribution
  -nr NUM_RUNS, --num_runs NUM_RUNS
                        number of times to iterate over
                        (num_people,num_connections) pairs

Parameter Descriptions

NOTE:

• For each of the num_people and mean_connections parameters, you can pass a list of values. The experiment engine will automatically run all pairs.
• The max values are suggestions to keep runtimes manageable.

Run Experiment

Say you decided that you want to run an experiment for the following:

• num_people = [200,300,400]
• mean_connections = [10,20,30]
• num_runs = 2

Execute the following:

python3.7 connections_experiment.py \
--num_people 200 300 400 \
--mean_connections 10 20 30 \
--num_runs 2 

You can also use the short-hand version of the flags:

python3.7 connections_experiment.py \
-np 200 300 400 \
-mc 10 20 30 \
-nr 2