In this assignment, you will develop a deep learning system for the classification of images containing decimal digits into three categories:

  1. A-digit 0
  2. B-digit 1
  3. C-others (e.g. any digit in 2,3,4,5,6,7,8,9)

For this purpose, you will use the MNIST dataset. (Download Here) (Links to an external site.) 

You will develop your solution within a transfer learning paradigm. Rather the designing and training a deep learning system from the scratch, you can use the Internet to find solutions to start from and/or augment. (HINT: There are many models in Python that can be found online to solve this problem.)


  1. Download and arrange a subset of the MNIST data in the three categories above, performing the proper partition into training, validation and testing
  2. Design, train, validate and test a model
  3. Compare and interpret the performance of each model (accuracy, behavior of the validation versus training, etc).

Note: Ensure you explain the rational and the methodology for tasks 1 and 2.

This is meant to be fun, challenging, and a stretch assignment.  Effort is what matters, not the perfect solution!!!