Code generation utility for neural network models implemented in the NineML (http://www.nineml.net) language.

In order to install the Chicken 9ML toolkit, you must first install Chicken Scheme and then use the chicken-install program to download and compile the 9ML toolkit.

You will need to obtain and install the Chicken Scheme compiler and programming environment. The latest official release is available from the following URL:

http://code.call-cc.org/releases/4.13.0/chicken-4.13.0.tar.gz

On Unix and Mac OS X there are no additional prerequisites, but on Windows, you will need to first install either one of the MinGW or Cygwin packages, because Chicken must be compiled with the gcc compiler.

First, you will need to download the Chicken source archive from the above location, then unpack it and change to the directory extracted from the archive:

   tar zxf chicken-4.13.0.tar.gz
   cd chicken-4.13.0

Then, you need to compile and install:

On Linux:

make PLATFORM=linux PREFIX=$HOME/chicken install

On Mac OS X (64-bit):

make PLATFORM=macosx ARCH=x86-64 PREFIX=$HOME/chicken install

On Mac OS X (32-bit):

make PLATFORM=macosx PREFIX=$HOME/chicken install

On MinGW or Cygwin:

make PLATFORM=mingw PREFIX=C:\chicken install
make PLATFORM=cygwin PREFIX=C:\chicken install

After Chicken Scheme has been successfully installed, the current released version of 9ML toolkit can be installed using the chicken-install program:

  ~/chicken/bin/chicken-install 9ML-toolkit

(On Windows, the tilde must be replaced with C:\ or whichever drive you have installed Chicken on).

The most recent development version of the 9ML toolkit code can be installed from this source repository by calling chicken-install program in the directory where the source tree has been cloned:

  git clone https://github.com/iraikov/9ML-toolkit
  cd 9ML-toolkit
  ~/chicken/bin/chicken-install 

Note that chicken-install will fetch and install additional Chicken Scheme libraries required by 9ML-toolkit, and thus it will require network access during the installation.

In order to compile the code generated by 9ML toolkit, you will also need the MLton compiler, which is available here:

http://sourceforge.net/projects/mlton/files/mlton/

After you download, build and install MLton, please make sure that the directory where it is installed is included in your PATH.

For example, if MLton resides in /usr/local/mlton, then add /usr/local/mlton/bin to your PATH:

PATH=/usr/local/mlton/bin:$PATH

export PATH

9ML-network operand1... [options...]

Where operands are NineML user layer files containing network description.

The following options are recognized:

• --platform=PLATFORM: simulation platform (one of mlton, mlton/c)

The 9ML-network program will create an executable named Sim_{model name} in the current directory, which accepts the following arguments:

• -d, --duration=VALUE: simulation duration in milliseconds
• --abstol=VALUE: absolute error tolerance of solver
• --reltol=VALUE: relative error tolerance of solver
• --timestep=VALUE: global time step of network simulation
• --logperiod=VALUE: write out spike times and state information every N ms of simulated time
• -s, --spikerecord=POPULATION: name of population or population set for spike recording
• --statesample=VALUE: sample size of neurons for state recording
• --extsample=VALUE: sample size of neurons for external input recording
• --prjrecord: record projection connectivity edges to files
• --cell-randomseeds=SEED1,...,SEEDN: random seeds for spike generation per each population
• --prj-randomseeds=SEED1,...,SEEDN: random seeds for connectivity construction per each projection
• --spikeout=PATH: path to file name used for spike recording
• --stateprefix=PATH: prefix for file names used for state recording
• --eventprefix=PATH: prefix for file names used for event recording
• --extprefix=PATH: prefix for file name used for external input recording
• --prjprefix=PREFIX: prefix for file name used for printing projections
• -v, --verbose: prints detailed information about the internal representation of the model during the code generation process

9ML-singlecell operand1... [options...]

Where operands are NineML user layer files that instantiate a single cell.

The following options are recognized:

• --platform=PLATFORM: simulation platform (one of mlton, mlton/c)

The 9ML-singlecell program will create an executable named Sim_{model name} in the current directory which accepts the following arguments:

• -d, --duration=VALUE: simulation duration in milliseconds
• --abstol=VALUE: absolute error tolerance of solver
• --reltol=VALUE: relative error tolerance of solver
• -v, --verbose: prints detailed information about the internal representation of the model during the code generation process