Learning methods

Our module currently includes learning methods based on the Particle Swarm optimization. Future versions are supposed to include other, methods of learning, e. g. gradient-based optimization algorithms, Simulated Annealing, constrained optimization methods, ...

User defined design

Our aim is to provide a framework for arbitrarily structured Neural Networks with an emphasis on an efficient learning on massively parallel infrastructures. 

Massive parallelism

One of the main objectives of our project is to have a robust massively parallel library for a wide range of problems. Our intent is to utilize multiple parallelization technologies (OpenMP, MPI, OpenACC, vectorization).