Handwritten Digits Classification Using ConvNets

Welcome to this chapter on using convolution neural networks (ConvNets) for the classification of handwritten digits. In Chapter 2, Training NN for Prediction Using Regression, we built a simple neural network for classifying handwritten digits. This was 87% accurate, but we were not happy with its performance. In this chapter, we will understand what convolution is and build a ConvNet for classifying the handwritten digits to help the restaurant chain become more accurate in sending text messages to the right person. If you have not been through Chapter 2, Training NN for Prediction Using Regression, please go through it once so that you can get an understanding of the use case.

The following topics will be covered in this chapter:

• Convolution
• Pooling
• Dropout
• Training the model
• Testing the model
• Building deeper models

It would...