This is a simple backpropagation neural net library for C++.
nnbase.cc/.h: Basic functions for creating in-memory structures representing a neural network. Includes some general learning algorithm tweaks, such as counterweights, delta-bar-delta and weight decay. In this implementation, the algorithm used for learning is a function of the neurode. One can mix and match as needed. nnbase defines classes for defining networks, with ability to stream the definition to stdout or read a definition from stdin. nnif.cc/.h: A functional wrapper over nnbase, which allows for loading or saving to files, iterating over values, analysing, reporting and manipulating the structure programatically. Each nnif instance embeds a single neural network, and it is possible to build some complex structures by chaining nnif instances.