Apache Spark Deep Learning Cookbook

How it works...

This section explains the process of transfer learning when applied to the MNIST dataset.

1. We are definitely taking a shortcut approach with transfer learning as we are either limited in resources, time, or both as we are taking prior work that has already been done and hoping that it will help us solve something new. 
2. Since we are dealing with an image classification problem, we should use a pre-trained model that has worked with classifying common images in the past. There are many common ones out there but two that stand out are:
   1. The ResNet model developed at Microsoft.
   2. The Inception model developed at Google.
3. Both models are useful for image classification because both Microsoft and Google have a wide spectrum of images ...