This is a C++ library written for developing experiments for energy-based Hopfield networks. I wrote this library primarily because python is slow to simulate experiments for modern versions of Hopfield Networks that have very high memory capacity (~ a million memories). The library is written using Pytorch C++ for GPU and CPU acceleration.
The following libraries are required prior to build EDEN.
cxxopts- for commant line argument parsingprogressbar- for displaying progressbarlibtorch- for hardware accelerationmatplotplusplus- for plotting diagnostics
make sure matplotplusplus is downloaded to the main folder prior to building.
The library can be configured and build by running the following commands from the build directory.
cmake -DCMAKE_PREFIX_PATH=<libtorch_path> ..
cmake --build . --parallel 8
The experiment binaries will be written to the bin folder in the build directory.