A complete neural network from scratch trained on the MNSIT data set.
After being trained, if given a 28 x 28 image from the MNIST data set it is able to predict with ~ 85% accuracy what digit is drawn on the image.
The entire neural network is created only using Numpy and Pandas for matrix manipulation and math. NO tensor flow or machine learning libraries are used, the entire network is created from scratch.
Partial Credit: Samson Zhang on YouTube