Open Source Library for Self-Organized Learning
The NeuralGas library by Sergio Roa is a collection of algorithms based on the topology preserving self-organizing Neural Gas learning algorithm. They can be applied to density estimation/quantization/clustering tasks, i.e., to infer the underlying structure of unknown data-driven probability distributions. It is written in C++ (for now).
- For quantization, the Life long robust Growing Neural Gas (RobustGNG) algorithm is the most efficient. It can decide when to stop when the right quantization is achieved, also in the presence of noisy data sets.
- There is also an implementation of Merge Growing Neural Gas (MGNG) for temporal clustering.
- The Growing Neural Gas (GNG) algorithm is also provided for comparison purposes.
Please contact the author if you want to cite the RobustGNG algorithm in a scientific article.
You will find technical documentation in the following link: NeuralGas documentation
The library can be downloaded from the GitHub project page.