Machine Learning for the Subsurface

Timing for Americas, 6–9 December 2021

Agile’s famous Machine Learning for Subsurface class is fully online, so you can take it from anywhere. This class has timing best for North and South America. The class is also shortened to Monday to Thursday to make it easier for anyone to attend. And it helps us keep the price down!

This class assumes you have a sound knowledge of Python, or have taken our Intro to Geocomputing class. In particular, if you know how to:

• Write a function in Python.

• Load a CSV into Pandas.

• Make a plot in matplotlib.

…then you should have no trouble in this class!

The class starts off with a look at scikit-learn for training and evaluating ‘shallow’ learning models. Then we’ll move into deep learning and learn how neural networks can sometimes beat shallow methods. By the end of the class, you’ll be designing machine learning pipelines and making predictions of your own.

Day 1     We'll explore Scikit Learn for simple machine learning tasks using geoscience data. After this day, students will have a good overview of how to load and interrogate data sets, and train a predictive model.

Day 2 We look in more detail at classification and regression tasks in Scikit Learn. In doing so, we discuss some things that make machine learning with natural datasets difficult, and talk about ways to deal with them.

Day 3 We build a neural network from scratch, at first using only NumPy. This lets us look inside what is sometimes a bit of a black box, and answer questions about how a neural network learns.

Day 4 We look at an image classification task in Scikit-Learn, TensorFlow/keras, PyTorch/skorch, and fastai. We’ll also see how to use a pre-trained model to quickly build a powerful image classifier.

During the class, we may also have the chance to look at unsupervised learning and sequence learning. Finally, we try to spend some time helping students figure out their own machine learning tasks to continue with after the class is over.

We have taught this class more than 40 times since 2016, training over 600 geoscientists and engineers in the fundamentals of machine learning. You can do it!

Regular price: USD 995 — purchase below, or we can invoice you.
Student price: USD 249 — purchase below.

If you are paying your own way (i.e. institution or employer is not paying), then please email us and we can work out a price for you. Please also fill out the form below so that we know to register you.

Location: online

Approx 0830 CST (1430 UTC) to 1230 CST (1830 UTC), exact times to be confirmed.

The class takes place in Zoom and Slack.

Full details to follow after registration.