This project is a playground for me to explore implementations of machine learning techniques. It contains:
- basic stuff like automatic differentiation, statistics stuff and a term algebra for symbolic computations
- a CGP evolutionary learning algorithm
- several ways to represent (simple backpropagation) neural networks in Scala using functional programming or symbolic computation to experiment with several learning algorithms
- an attempt at running the result on the GPU for speed with AparAPI and CuDA. (I don't think I'll continue that GPU Mini-Tensorflow stuff - as much fun and interesting that is, it's just more than I could possibly get done in my sparetime, and it obviously couldn't match the real Tensorflow - except being designed in an IMHO way better langage than Python - Scala. :-) So I'm trying out the real Tensorflow now.)
Disclaimer: this is just quick sparetime fun stuff, and not representative of my professional code quality in various ways, most notably in the amount of documentation. ;-)
Dr. Hans-Peter Störr, http://www.stoerr.net/