Skip to content

A from scratch Neural Network, using the MNIST handwritten digit database for training

Notifications You must be signed in to change notification settings

Cory44/Neural_Network

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

6 Commits
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Neural Network

A from scratch Convolutional Neural Network class

The NerualNet class has been developed to be as flexible as possible. A user can instantiate the network with numpy arrays of varying sizes for both the inputs and outputs. I have given an example using the MNIST database of handwritten digits, which consists of a set of 60,000 images, each 784 pixels (28 x 28), and the corresponding outputs which are numpy arrays with a length of 10 (with the index representing the correct digit from 0 to 9).

With the MNIST database of handwritten data as well as tests with the MNIST Fashion database (consisting of 28 X 28 pixel images of clothing items), I have gotten the network to classify new items with an accuracy >95% at 10,000 training iterations.

Specs of the class

  • Uses numpy arrays for efficient vector calculations
  • Trains using user given inputs and outputs for a specified number of iterations
  • Instance variables for the node weights are created and are updated at each training iteration
  • Classification can be done for an individual item or in bulk using a user given numpy array of values to classify

Special Thanks

About

A from scratch Neural Network, using the MNIST handwritten digit database for training

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages