Exploring notMNIST Dataset

The data-set has been used to learn different ML and NN models and classifiers.

If you want to kickstart in the same, following is the link for the same.
PDF File -> https://drive.google.com/file/d/18SIGbga5VzXJrfoMAsFOzDUhayhPken6/view

For better understanding of machine learning concepts I used this data-set to implement different machine learning models.

The implementation is as follows:

1. Logistic Regression - accuracy = 89.67%
   
2. Neural Network with L2 Regularization - test set accuracy = 10.0
   
3. Neural Network with Dropout Regularization - accuracy = 86.0%
   
4. Neural Network with Multi Layers = 92.1%
   

After implementing the above four models the conclusion says that with this data-set the maximum accuracy we are getting on our test set is from “Multi-Layer Neural Network”.

Note: Most of the code is used from the https://www.ritchieng.com/