mramshaw / gophernet

A simple from-scratch neural net written in Go

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Building a Neural Net From Scratch with Go

This is a simple neural network built in Go. It is further discussed here and in Machine Learning with Go.

Results

According to the original article, this should give results in the 97% range.

$ ./gophernet

Accuracy = 0.97

$

However, my results were nowhere close to this:

$ ./gophernet 

Accuracy = 0.77

$

[My first result was 87% - so the results vary greatly.]

To Do

  • Twelve-factor everything
  • Allow for a seed value so that results are reproducible and not overly dependent on initial conditions
  • Add code to segment data into training and test datasets rather than rely on pre-segmented data
  • Investigate pickling in Go

About

A simple from-scratch neural net written in Go


Languages

Language:Go 93.0%Language:Makefile 7.0%