Latin Hypercube Generator (version 3.+)

Home

What this page does is simple. Here is the sequence of events:

  1. The user (that's you!) configures the running parameters of the LHC sampling tool via the interface below.
  2. The user clicks the "Submit" button.
  3. This page sends a request to the LHC API.
  4. The LHC API uses the Github Metadata API to find the latest version of the v3 LHC sampler executable.
  5. The LHC API downloads the LHC sampler executable from Github.
  6. The LHC API runs the LHC sampler executable, using the user-supplied parameters.
  7. The LHC sampler writes the output to a temporary CSV file.
  8. The LHC API reads the CSV file and returns it to this page.
  9. This page displays the output.
  10. Optionally, the user can then download the output as a CSV file.

This is intended to demonstrate the capabilities of the LHC sampler tool. If you are interested in cutting out the middleman and using the tool for yourself, please feel free to navigate back to the Home page (using the "Home" link above), click the "source" link in the CLI Latin Hypercube Point Sampler heading, and download the latest version of the executable from Github for use on your own machine! The Github repo includes documentation for the arguments that the tool will accept.

However, you might be asking yourself: What is the point of this tool? What is a "hypercube"? Why is it latin? All adequate questions.

To begin with, consider the following: You are a statistician working at Los Alamos National Laboratory in the New Mexican Jemez mountains in 1979. You're working on simulated models of mid-air nuclear detonations. You need a way of quickly understanding the general impact of such an event, but it's difficult to run a large number of simulations when you have to manually tweak the inputs between runs. There are a variety of inputs to tweak. Some examples include: height of detonation above the surface, height of detonation above sea-level, what kind of surface is below the detonation, surface topography, wind speed, wind direction, amount of nuclear material triggering the detonation, and so on. You need a way to generate lots of random, meaningful values for these inputs ahead of time so that you can queue a bunch of simulations, then analyze the results en masse. The solution you devise: a generalization of Latin squares that allows for any number of "dimensions" on the square. Each dimension equates to a parameter of the simulation, and each point of the Latin square equates to a set of values for all of the simulation's parameters.

If that explanation is not clear enough, simply click "Submit" below. This webpage will proceed with the default parameters, and you will see 1000 five-dimensional points sampled. Alternatively, my email is displayed at the bottom of the page. If you have any questions or comments, please reach out to me!

- Cameron Chrobocinski

|
|
|
(:)
Dimension Name (Default? )Slightly Randomize Values (All? )Scale Override (Base Scale for All? )
(:)
(:)
(:)
(:)
(:)