Morphologica examples

This directory provides some example models demonstrating the use of Morphologica. They build as if they were external code - so they don't build as part of the morphologica build process; instead you build and install the morphologica library, then cd into one of the examples and start a fresh cmake build process there. These provide templates for developing your own programs using morphologica.

Morphologica intends to provide you with the scaffolding you need to write a C++ simulation. It gives you a way to write and load configurations, and, importantly, save the config out alongside your logs with runtime parameters (such as the date and time, the command line args used and even the git commit version of your code). It provides a nice, simple wrapper around the HDF5 library so that you can save your simulation data into files which can be opened by many different mathematical packages, including Python, MATLAB and GNU Octave. It provides you with the HexGrid class, which makes it easier to run simulations on hexagonal grids. You can apply arbitrary boundary shapes too! Finally, it has a set of classes which implement a modern OpenGL visualization scheme which is efficient enough to run alongside your simulation.

schnakenberg

This directory contains a complete example. After you have compiled and installed morphologica, you should be able to change directory into schnakenberg, create a build directory and carry out a cmake build:

cd morphologica/standalone_examples/schakenberg
ln -s ../../../morphologica # This mimics having a copy of morphologica in your code tree
mkdir build
cd build
cmake ..
make
cd ..

This program demonstrates the creation of a reaction diffusion system using the morph::RD_Base class. It is a good demonstration of the main classes that Morphologica makes available. It demonstrates:

• The use of the morph::Config JSON configuration scheme. This makes it easy to bring parameters into your simulation.
• The morph::RD_Base class (which makes use of morph::Hexgrid), and how to specialise it into the RD_Schnakenberg class.
• The morph::HdfData scheme for saving your simulation data into HDF5 files
• morph::Visual for OpenGL visualisation of the hex grid while the simulation runs, including the use of morph::ColourMap.
• The morph::Tools class, which contains a collection of static utility functions.

Run it as follows:

# The paths you use will depend on your current working directory:
./standalone_examples/build/schnakenberg ./standalone_examples/schnakenberg/schnakenberg.json

You can interact with the window using your mouse. For some help (which will appear on stdout) press 'h' (with the graphical window focussed).

neuralnet

This implements a feedforward neural network which can be trained to classify the MNIST database of handwritten numerals. It demonstrates the use of morph::RandUniform<> and morph::vVector<>. It shows how to set up a CMakeLists.txt file to use morphologica. Because it doesn't use any OpenMP pragmas for parallel execution, it shows how to write the CMakeLists.txt without OpenMP.

cd morphologica/standalone_examples/neuralnet
ln -s ../../../morphologica # This mimics having a copy of morphologica in your code tree
mkdir build
cd build
cmake ..
make
cd ..

Then to run the example:

./build/ff_mnist # see also ff_small and ff_debug for debugging

recurrentnet

A recurrent neural network example. Demonstrates the use of code in the morph::recurrentnet namespace in RecurrentNetwork.h and RecurrentNetworkModel.h.

cd morphologica/standalone_examples/recurrentnet
ln -s ../../../morphologica # This mimics having a copy of morphologica in your code tree
mkdir build
cd build
cmake ..
make
cd ..

For instructions on running the example, see the README.md file in recurrentnet.

Elman network

Elman 1990, Finding Structure in Time. An implementation of this recurrent network. See README in the elmannet directory.