go-nn

a perceptron implementation written in Go

includes a demo with a neural network and supervised learning that tries to guess what f(x) is

every point on the graph is a randomly generated training input

green line shows f(x), for reference, and blue line shows the current guess

demo

$ go run visual_example.go

$ open localhost:3000

![Browser demo] (https://raw.githubusercontent.com/jahrlin/go-nn/master/visual_example.gif)

TODO / issues

• get rid of the casting and use same-size floats instead
• subscribe to google.visualization.events to avoid melting the client GPU
• should probably use requestAnimationFrame for that sweet 60 fps
• make it async with a websocket or something