This is a simple dimensionality reduction exercise. The 30 dimensional data is reduced to 2 dimensions using linear autoencoders with Principal Component Analysis (PCA).

What is Dimensionality Reduction?
In machine learning problems, there are often so many features and variables that goes into the training of the model. The higher the dimensions are, the harder to visualize the data. Therefore, we use dimensionality reduction to reduce the correlated features into smaller dimensions and to reduce the redundancy. It can be thought as a way of data compression