Skip to content

A neural network for digit classification just using numpy.

MarcBrede/neural_network

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

10 Commits
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

neural_network

Neural network with one hidden layer, which is trained with the mnist dataset for handwritten digits. Input for the nn. is a 28x28 binary representation vector of each digit, making the input layer a size of 784 neurons. Ouput is a 1x10 vector which represents the certainty of the nn. that the current handwritten digit is the respective digit. The nn. is trained with gradient descent and mini batches.

All the other meta params can be varied. The current params seem to work best, performing with a missclassification rate of 6-10%.

The nn is also implemented with tensorflow.

Performance of the neural network

About

A neural network for digit classification just using numpy.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published