/PyNN

Python neural network written to learn the mnist handwritten digits

Primary LanguagePython

PyNN

Python neural network written to learn the mnist handwritten digits

The Project

The MNIST database includes handwritten digits encoded into arrays represent pixel values. This neural network is designed to use these as inputs and produce an answer: 0-9. It does this by "feeding forward" the inputs throught the layers of the neural net. First it uses

Sig(w DOT x)

where Sig is the sigmoid function, w is the weights of that layer, and x is the inputs. Then from the hidden to output layer the process is repeated, but with the ouputs from the previous step used instead of the origin inputs. Once an answer has been produced, the backpropagation algorithm is used to update the weights according to the errors.

To Run

In the main directory of the project (where 'Main.py' is located)
$ python3 Main.py

Links

Here is a small collection of wikipedia links to explain the subjects involved