PWS

Simple layered neural network library made in Go.

Goals

• Create a working Neural Network with Fully Connected layers
• Solve MNIST
• Create a working Convolutional Neural Network

Try it

First clone the repository

git clone https://github.com/TrizlyBear/PWS.git && cd PWS

Then run one of the testing scripts

go run testing/xor_test.go

Or try the module by importing it

go get github.com/TrizlyBear/PWS/...

package main

import (
	"github.com/TrizlyBear/PWS/sequential"
	"github.com/TrizlyBear/PWS/sequential/activation"
	"github.com/TrizlyBear/PWS/sequential/layers"
)

func main() {
	// Create a training set consisting of training values and labels
	x_train := [][][][]float64{...}
	y_train := [][][][]float64{...}
	
	// Initialize a model consisting of 2 fully connected layers and 2 activation layers
	model := &sequential.Model{Layers: []sequential.Layer{&layers.FC{Out: 10}, &activation.Tanh{}, &layers.FC{Out: 1}, &activation.Tanh{}}}
	
	// Train the model
	model.Fit(x_train, y_train, 10, 0.1)
}