institutefordiseasemodeling/idm-cms

IDM compartmental modeling software (CMS)

IDM Compartmental Modeling Software

This software has the following features:

• Multiple solvers
• Multiple output formats
• Simple JSON configuration
• Text based model specification

Here is a sample configuration: 👍

{
    "duration" : 3650,
    "runs" : 1,
    "samples" : 3650,
    "solver" : "R",
    "output" : {
        "headers" : true
    },
    "tau-leaping" : {
        "epsilon" : 0.01
    },
    "r-leaping" : {}
}

Here is a sample model file:

; simplemodel

(import (rnrs) (emodl cmslib))

(start-model "seir.emodl")

(species S 990)
(species E)
(species I 10)
(species R)

(observe susceptible S)
(observe exposed     E)
(observe infectious  I)
(observe recovered   R)

(param Ki 0.0005)
(param Kl 0.2)
(param Kr (/ 1 7))
(param Kw (/ 1 135))

(reaction exposure   (S I) (E I) (* Ki S I))
(reaction infection  (E)   (I)   (* Kl E))
(reaction recovery   (I)   (R)   (* Kr I))
(reaction waning     (R)   (S)   (* Kw R))

(end-model)

You can find additional documentation here:

https://docs.idmod.org/projects/cms/en/latest/

idm-cms's Issues

ODE solver(s) for deterministic execution

It would have been nice to have one deterministic solver provided by CMS, in order to sanity check or anchor results.

Per @clorton, some solvers:

• https://en.wikipedia.org/wiki/Numerical_methods_for_ordinary_differential_equations
• https://blogs.mathworks.com/loren/2015/09/23/ode-solver-selection-in-matlab/
• http://docs.juliadiffeq.org/latest/solvers/ode_solve.html

Some implementations:

• https://www.microsoft.com/en-us/research/project/open-solving-library-for-odes/
• https://github.com/OlegJakushkin/CSharpOdeLibrary
• maybe - https://www.mathdotnet.com/

Update Python2.7 scripts to Python3.6

Just realized that the scripts in the regression folder still use Python 2.7.

Upgrade Versioning to Include Git Commit Hash and Status

https://github.com/jeromerg/NGitVersion

Leaping (tau-leaping, r-leaping) may try to send a species population negative.

These solvers should detect this situation before trying to execute reactions and fall back to SSA.

Tau-leaping solver sensitive to parameter name case, consider making configuration case-insensitive.

Jaline et al. at Northwestern ran into an issue where following the documentation for configuring tau-leaping did not work. The documentation did not match the casing expected by the code. Initial fix is to update documentation. Better fix is to make the configuration case-insensitive. There is, for example, no need to differentiate between "Epsilon" and "epsilon" or "SSAruns" and "ssaruns".

Move docs build/hosting to Read the Docs

• Add RTD yaml config file, remove Tipue search and beautifulsoup, remove internal conditionalization tag and make commands

• Configure and test the RTD build in a private fork

• Move to IDM repo

• Configure subdomain and proxy from idmod.org

• Decide which prior versions of docs to build

• Push RTD docs public

• Add redirects from gh-pages

• Add redirects from idmod.org

Use mathdotnet for PRNG

Use mathdotnet MersenneTwister for PRNG.

https://numerics.mathdotnet.com/Random.html

Linux version?

Hi,

Are there any plans to release a Linux version of this?

Thanks!

Warn on NaN propensity

ewenger "commented out" some reactions in a model by setting a parameter to 0. This led to a divide by zero while calculating the reaction rate which resulted in a NaN propensity. Due to this, no reactions were chosen and the model did not progress. Unfortunately, no warning or error was produced.

infection_duration.emodl.txt
infection_duration_config.json.txt

CI build, test, and package

CI build, test, and package, delivering to IDM artifactory nuget-staging.

Add component tests for uniform sampling from PRNG.

Implement simple tests of uniformity for each PRNG. Consider order of magnitude buckets and samples. E.g., 0-0.5-1 and 0-0.25-0.5-0.75-1 etc.

Implement modulo operator

The modulo operator is useful for simulating a lookup table.
lookup.emodl.txt
config.json.txt

Add try catch around ExecuteReaction(s) in B-Leaping.

B-Leaping should give helpful comment about negative species counts since this is more likely with the algorithm.