Exercise 10

Part of the course Machine Learning for Materials and Chemistry.

Task 10.1: Setup / understand

Download the python code for a simple neural network. Install tensorflow using pip install tensorflow and run the file. Go through the code and comment on what each line is doing.

Task 10.2: Parameters

Vary some of the following: the learning rate, the layer size, the network depth, the number of training steps and the activation function. Describe and explain what you observe.

Task 10.3: Dimensionality reduction

Add another dense layer between the existing ones and reduce its dimensionality drastically. What would you expect to happen? How is it possible can ten digits be learned if the layer has dimensionality 1?