ML Classification: cloud classification

Guest speaker Sean Foley (NASA) will give an example of using neural networks for a classification task with pytorch.

Background: Cloud masking refers to the task of assigning a binary value to each pixel in a satellite image, according to whether or not that location is covered by a cloud. It is essential to many other data products. For non-atmospheric scientists, this is usually useful to know what pixels to ignore, or ‘mask out’, in processing. For atmospheric scientists, it is often the opposite: we may exclusively wish to study the clouds. In this session, we will see how to use a Simple Multi-Layer Perceptron (MLP) to perform this task.

We’ll be using pytorch. This is the most popular machine learning library among researchers as it represents a good balance between flexibility and ease-of-use.