MNIST Handwritten Digit Database - MLP (Multi-Layer Perceptron)

This is a data exploring study. You can see the descriptions along with the codes to be helpful in this regard. The dataset used is MNIST Handwritten Digit Database.

Dataset summary is explained down below.

MNIST - Mixed National Institute of Standards and Technology database.
Created by - Yaan LeCun, Corinna Cortes, Christopher Burges.
Image size - 28 x 28 pixel square (784 pixels in total).
Training images - 60,000
Testing images - 10,000
Top error rate - 0.21 (achieved by CNN)
Dataset size in Keras - 14.6 MB

This assignment is suggested to complete on Google colab to benefit from its GPU support.
I builted a simple Multi-Layer Perceptron (MLP) to recognize handwritten digit (using MNIST dataset). The details and outputs avaliable in the .ipynb file.

Test error on test set: 0.021700032949447667%

Incorrect classified test samples with predicted labels:

You can also check the addresses below to get more information and use as a resource.

[1] @basu369victor/handwritten-digits-recognition-d3d383431845"">https://medium.com/@basu369victor/handwritten-digits-recognition-d3d383431845

[2] https://gogul.dev/software/digits-recognition-mlp

[3] http://yann.lecun.com/exdb/mnist/