PlasmaML

Machine Learning tools for Space Weather and Plasma Physics

courtesy NASA

PlasmaML is a collection of data analysis and machine learning tools in the domain of space physics, more specifically in modelling of space plasmas & space weather prediction.

This is a multi-language project where the primary modelling is done in Scala while R is heavily leveraged for generating visualizations. The project depends on the DynaML scala machine learning library and uses model and optimization implementations in it as a starting point for extensive experiments in space physics simulations and space weather prediction.

Getting Started

PlasmaML is managed using the Simple Build Tool (sbt).

Installation

Requirements

1. Java Development Kit 8.

2. Scala

3. sbt

4. R with the following packages:

   • ggplot2
   • reshape2
   • latex2exp
   • plyr
   • gridExtra
   • reshape2
   • directlabels

Steps

After cloning the project, PlasmaML can be installed directly from the shell or by first entering the sbt shell and building the source.

From the shell

From the root directory PlasmaML run the build script (with configurable parameters).

./build.sh <heap size> <compile with gpu support> <use packaged tensorflow> <update bash env>

For example the following builds the project with 4 GB java heap and GPU support.

./build.sh 4096m true

Note that for Nvidia GPU support to work, compatible versions of CUDA and cuDNN must be installed and found in the $LD_LIBRARY_PATH environment variable see the DynaML docs for more info.

Use the last parameter <update bash env> to add the PlasmaML executable in the bash $PATH.

The following build will use 4 GB of heap, with GPU support, precompiled tensorflow binaries and adds plasmaml binary to the $PATH variable.

./build.sh 4096m true false true

From the sbt shell

Start the sbt shell with the script sbt-shell.sh having the same parameters as build.sh

./build.sh <heap size> <compile with gpu support> <use packaged tensorflow>

From the sbt shell, run

stage

After building, access the PlasmaML shell like

./target/universal/stage/bin/plasmaml

For more information on PlasmaML and its modules, refer to the scala docs below.

1. omni: Forecasting models for geomagnetic indices.

2. mag-core: API for Bayesian inference of radiation belt parameters.

3. helios: Machine learning models for solar wind and heliosperic data.

4. vanAllen: Processing of van Allen probe data.