matexp

This program solves systems of differential equations for the NEURON simulator using the approximate matrix exponential method of integration. This is a new method of integration. The solution is faster and more accurate than NEURONs built in "sparse" solver. This method is only applicable to systems which are linear and time-invariant, such as Markov kinetic models. This method is also limited to systems with one or two inputs.

This program uses the NMODL file format (".mod" files). The input kinetic model is an NMODL file, and the solution is written to a new NMODL file.

Installation

Prerequisites:

• Linux
• The g++ compiler

$ pip install matexp

Usage

$ matexp --help
usage: matexp [-h] [-v] [--plot] -t TIME_STEP [-c CELSIUS] [-e ERROR]
              [-i NAME MIN MAX] [--log [INPUT]] [--target {host,cuda}]
              [-f {32,64}]
              INPUT_PATH OUTPUT_PATH

positional arguments:
  INPUT_PATH            input path for unsolved NMODL file
  OUTPUT_PATH           output path for solved NMODL file

options:
  -h, --help            show this help message and exit
  -v, --verbose         print diagnostic information, give twice for trace mode

simulation parameters:
  --dt DT, --time_step DT
                        milliseconds, default: 0.025
  -t TEMPERATURE, --temperature TEMPERATURE
                        degrees celsius, default: 37
  -e ERROR, --error ERROR
                        maximum absolute error per millisecond. default: 10^-3

input specification:
  -i NAME MIN MAX, --input NAME MIN MAX
  --log [INPUT]         scale input logarithmically, for chemical concentrations

computer specification:
  --target {host,cuda}  default: host