Sampling Bounds and Random Number Generators

[gdelato2]

1. User-Defined Sampling Ranges: Allow users to specify a parameter's lower and upper bounds for sampling. The feature will enable: 1. parameters to have both negative and positive values, 2. avoid non-physical system configurations and 3. better control of input space

2. Random Number Function Selection: Provide an option for users to provide or select a random number generator. Uniform (current option), Sobol, and Latin Hypercube generators will be a good start.

[gdelato2]

Proposed Changes:

1. Extend the action space yaml file to hold parameter specifications. Besides range, additional specifications can be added as required in the future (units, data type, if the parameter is somehow constrained, etc.). The specification can be parsed during input sample generation.

2. The current paradigm of creating a single random action, executing the model, and repeating may not work for Latin Hypercube and Sobol. These methods work best when they generate multiple random samples at once. LHS, in particular, requires the number of samples to ensure samples are correctly distributed. I think we need either:

• A way to store generated random numbers in an object or
• A way to store input samples in csv file and then use this file to execute simulation runs

Either way, we need a mechanism to store a large set of random numbers.