| Univ. Cote d’Azur (Nice) Prof. M. Boquien | Statistical methods: Introduction to Machine Learning in Astronomy (S2, elective, 3 ECTS) – offered from A.Y. 2024/2025 |
| Learning Outcomes: | Machine learning has been sweeping the world over the past 15 years. The revolution brought about by these new techniques has considerable ramifications across astronomy, bringing spectacular new results over a large range of applications. With the relentless growth of the volume of data in the years to come, these methods are set to develop even more and the revolution is certainly only beginning. In this course we will take the first steps into this new world and learn of different machine learning techniques and how to use them to answer our questions about the universe. |
| Knowledge and Understanding: | First and foremost, the students will become familiar with different machine learning algorithms, which they will implement themselves or use through specialized public libraries. They will learn to identify which problems are best addressed by such techniques, clearly identifying the benefits and drawbacks with respect to more classical approaches in order to make an informed decision. They will also be able to select the most appropriate machine learning algorithm to address a specific question. Finally, they will learn how to adapt and apply such algorithms to astronomical cases. |
| Applying Knowledge and Understanding: | The students will apply machine learning algorithms to real astronomical cases based on models or large surveys such as Gaia or SDSS. This will include for instance automated object classification or the detection or rare objects in large catalogs. The students will analyze their results critically, comparing them with those published in the literature. Finally, they will infer what these results tell us about the universe |
| Prerequisites | Familiarity with Python. Knowledge of object-oriented programming is an asset but is not required |
| Program | The course will cover the following topics: support vector machines, decision trees, ensemble learning and random forests, dimensionality reduction, and clustering. |
| Description of how the course is conducted | Presentations, lectures, practice of tools and codes |
| Description of the didactic methods | Lectures, exercises, and project |
| Description of the evaluation methods | Written project report and oral presentations |
| Adopted Textbooks | None |
| Recommended readings | Statistics, Data Mining, and Machine Learning in Astronomy, Ivezić. Connolly, Vanderplas & Gray, Princeton University Press, ISBN 9780691198309 Hands-On Machine Learning with Scikit-Learn, Keras, and TensorFlow, Géro, O’Reilly, ISBN 9781098125974 |
