/DNN_from_scratch

This repo is dedicated to create a "DNN" from scratch to recognize "MNIST" dataset without using any deep learning libraries!

Primary LanguageJupyter Notebook

DNN_from_scratch

This repo is dedicated to create a "DNN" from scratch to recognize "MNIST" dataset without using any deep learning libraries!

References: All the credits goes to Samsong Zhang. Thanks for your beautiful explanation.

The architecture is in the following shape:

first layer: 28*28 by m array.
second layer: 10 by m array.
third layer: 10 by m array.
final layer: 1 by m array. (a one-hot layer)