Amazon made news in August when it announced that its Machine Learning University (MLU) was offering three new on-demand courses and supporting materials designed to give participants a running start at a career in the booming field of machine learning.
Now MLU is offering a new class that takes it in the direction of more comprehensive, in-depth machine-learning instruction. This class, on Decision Trees and Ensemble Methods (DTE), is being released through the MLU YouTube channel. Beginning Monday, Dec. 21, one day’s worth of lectures will be released on YouTube each Monday for five weeks. This repository on GitHub contains slides, notebooks and data sets for the course.
Decision trees are a common tool used by machine learning developers — a type of machine-learning algorithm that attempts to make a prediction by asking a sequence of questions. Any decision tree begins by answering a root question (“animal or mineral?”), then branches into a series of questions until an answer is reached (“mouse!”).
In machine learning, decision trees are often useful when dealing with tabular data, or in areas where an understanding of how a prediction was derived is needed, such as when identifying fraud.
As with other MLU classes, the new course is taught by Amazon scientists. In this instance, the decision trees course is taught by Cem Sazara, an MLU applied scientist. “We really tried to make this class state-of-the art,” he says. “It covers the foundations, but then goes deeper. We go into some of the boosting techniques, such as LightGBM (an open-source framework that supports performance and scalability in decision trees) and CatBoost (an open source library for gradient boosting on decision trees).”
The Decision trees course is the first of MLU’s second tier of more advanced classes. These will build on existing accelerated classes to provide in-depth discussion of the most popular techniques used across a wide range of machine learning domains.
As with previous courses, interest in the new content is high.
“This is definitely something people are interested in,” says Sazara. “Classes on topics like decision trees — and anything else related to deep learning — are in demand.”
MLU’s new class is available now. More classes will be released in 2021 which will provide a comprehensive understanding of deep learning and reinforcement learning.
To learn more, visit the MLU website.