EvoNet¶
First Steps
Step-by-step guides
About Smartpeak
Other
Overview¶
EvoNet aims to provide a machine learning framework that can optimize both network weights AND network structure simultaneously while still taking advantage of the latest hardware acceleration technology (Fig 1).
Currently, network structure is optimized using an evolutionary algorithm over network node integration and activation functions and over node connections (Fig 2), while network weights are optimized using standard backpropogation.
EvoNet is written in C++ and is optimized for hardware acceleration using native threading and CUDA GPU technology.
TODO¶
Todo
Various optimisations.
Credit¶
This project could have not existed without the excellent tools available: Boost, Eigen, Doxygen, Sphinx, and many others.
License¶
This project uses a MIT license, with the hope that it’ll be accessible to most users. If you require a different license, please raise an issue and I will consider a dual license.
The full license is available here.