Neural Network in GoLang

Feature

• BackPropagation Network / RBF Network / Perceptron Network
• Parallel BackPropagation Network (each neural has its own go-routine)

Benchmark

• Dataset: MNIST Acurrency Rate : 98.2% (800 hidden nodes)
• Actually, you can get 96.9% using 100 hidden nodes in just three minutes of training

TODO

• currently, the parallel version is much slower than the tranditional one, maybe caused by the cost of context switch of threads