eSpinn

eSpinn (Evolving SPIking Neural Networks) is a c++ package that simulates Spiking Neural Networks (SNNs) for control problems.

Quick Start

git clone https://github.com/hnqiu/eSpinn.git
cd eSpinn
mkdir build && cd build
cmake ..
make

cmake options:

cmake -DCMAKE_BUILD_TYPE=Debug ..
cmake -DNDEBUG=ON .. # if run in Ndebug mode
cmake -DESPINN_VERBOSE=ON .. # if run in verbose debug mode
cmake -DUSE_IN_MATLAB=ON .. # if package is used in MATLAB
cmake -DPYTHON_EXECUTABLE="$HOME/anaconda3/bin/python" .. # if package is compiled with conda env, i.e., the pybind module will be interpreted by the specified Python instance

Prerequisite

• Boost serialization

  # should have already been available on Ubuntu, otherwise:
  sudo apt install libboost-all-dev

• pybind11

  conda install -c conda-forge pybind11

  Refer to this post for more details on installing pybind11.

Example: Running Flappy Bird

• setup
• python tasks/sim_flappy.py -t

flappybird.mp4

Citation

To use this work, please cite:

@InProceedings{Qiu2020,
  author    = {Qiu, Huanneng and Garratt, Matthew and Howard, David and Anavatti, Sreenatha},
  booktitle = {Proceedings of the 2020 Genetic and Evolutionary Computation Conference},
  title     = {Towards Crossing the Reality Gap with Evolved Plastic Neurocontrollers},
  year      = {2020},
  address   = {New York, NY, USA},
  pages     = {130-138},
  publisher = {Association for Computing Machinery},
  series    = {GECCO '20},
  doi       = {10.1145/3377930.3389843},
  isbn      = {9781450371285},
  keywords  = {hebbian plasticity, neuroevolution, spiking neural networks, UAV control, evolutionary robotics},
  location  = {Canc\'{u}n, Mexico},
  numpages  = {9},
  url       = {https://doi.org/10.1145/3377930.3389843},
}

Author

• Huanneng Qiu