Thanks to visit codestin.com
Credit goes to github.com

Skip to content

epochventures/PPNN

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

4 Commits
 
 
 
 
 
 

Repository files navigation

PPNN

Pure Python Neural Network

    The project is aimed at those who want to use the very good book on Neural Networks
that has been used by many called 'Practical Neural Network Recipies in C++'
ISBN: 978-0124790407, by Timothy Masters. This is a simple python implemention 
that tries to tie in with the book but expressly for python without recourse to 
using modules such as NUMPY or SCIPY, not that there is anything wrong in using these 
fast modules, but here simply stay faithful to the implementations within the book.

As an example (like the book) we illustrate a working manifestation of the XOR example
and also provide the training example which are generated on the fly and then used to
train the network. After training the network you are free to test this network.

In PPNN networks exist as python dictionaries, which means transporting them is very easy
as is examing each and every weight. For example a working neural network that 
performs that function of XOR looks like this:

{'output': {0: {0: -8.30648077083447, 1: 3.6378400405424904, 2: 14.006753208219276, 3: 1.895120416489752}}, 
'hidden': {0: {0: -10.513505383255971, 1: 10.913534176278933, 2: 0.24543391390819042}, 
1: {0: 2.4869136576399664, 1: -0.3450042945080538, 2: -4.722548494099791}, 
2: {0: -3.685140869118192, 1: 4.433265048621298, 2: -3.1641081562711477}}}

Caveats: This code is not designed to be super quick but as a way to get up and running
as quickly as possible with python. I'm sure there are better ways to do many of the functions
but I'm sure this will be updated with time. The point is to be able to fiddle and play
with Neural Networks without treating them as some kind of magical black box, but to be
able to delve inside, tinker and play.

About

Pure Python Neural Network

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Contributors 2

  •  
  •  

Languages