scnANNlib Feature List
Version 0.1 (initial release)
- Library allows programmer to create a traditional neural network with an arbitrary number of layers, each with an arbitrary number of nodes.
- Networks can use momentum, with an arbitrary momentum coefficient. Learning rate also configurable.
- Node weight vectors randomized automatically-- range of random values specified in construction of network.
- Network code organized into a class, allowing easy creation and destruction of full networks.
- Object-oriented code allows for easy custom edits of internals.
- Well documented external API. (Internal API documentation to come in future releases.)
|
