Neural Network Hyperparameters

One of my goals for February was to revisit artificial neural networks, a topic that I had touched on briefly during the IBM Data Science Professional Certificate program. Happily, I am currently reading Deep Learning Illustrated by Jon Krohn, so I have plenty of relevant material to review and implement.

I decided to do a deep dive into neural network hyperparameters, namely:

• Batch Size
• Learning Rate
• Activation Functions
• Number of Layers

My goal was to see how adjustments to these hyperparameters impacted the network’s ability to learn. Exploration was done in Keras using the classic MNIST dataset.