SBPerceptron—working with MLPs in Cocoa
January 31, 2007
My attempts at creating a good Connect4 AI player using a neural network has had an unforeseen side effect: I’ve implemented a powerful MLP (multilayer perceptron) class in Objective-C, and I’ve decided to extract into its own project. Enter SBPerceptron:
SBPerceptron lets you conveniently create and train multi-layer perceptrons (MLPs) and start working with them immediately. Just specify how many hidden layers you want and how many nodes should be in each layer and you’re ready to create history.
The perceptrons can be trained using iterative, semi- or fully batched backpropagation of error, or, if you prefer, evolved using a genetic algorithm.
Online documentation is available. You can also download a disk image containing an embedded framework or grab the source from Subversion.
Leave a Reply