A single / multi layer / recurrent neural network written in Golang.
A single / multi level perceptron classifier with weights estimated from sonar training data set using stochastic gradient descent. Recently I added back propagation algorithm over multilayer perceptron network. The implementation is in dev. Planned features:
2017-10-04: Introduced Recurrent Neural Network (Elman Network) with "learn to sum integer" task. Big refactoring in code (working on)
2017-08-08: Introduced multi layer perceptron network definition with parametric number of hidden layer and neurons. Back propagation algorithm with different transfer function actived - I wanna thank you dakk because I was truly inspired by your code.
2017-08-01: Introduced validation package and k-fold cross validation.
2017-07-31: I started working on mlp
branch for MLP + back prop. It doens't work yet but...I push first commit after some exp in dev. I delete dev
because of some structs optimization.
2017-07-31: we started working on k-fold validation.
To run a simple test just open a shell and run the following:
git clone https://github.com/made2591/go-perceptron-go
cd go-perceptron-go
go get github.com/sirupsen/logrus
go run main.go
You can setup a MultiLayerPerceptron using PrepareMLPNet
. The first parameter, a simple []int
, define the entire network struct. Example: