Program Overview

Automated Machine Learning Course

The Automated Machine Learning course is a flipped-classroom lecture and exercise program that delves into the complexities of applying machine learning (ML) and deep learning (DL) in practice. The course aims to equip participants with the knowledge and skills necessary to utilize automated machine learning (AutoML) systems, develop their own systems, and comprehend the underlying ideas behind state-of-the-art AutoML approaches.

Course Description

Applying machine learning and deep learning in practice is a challenging task that requires a significant amount of expertise. The success of ML/DL applications depends on various design decisions, including data preprocessing, choosing a suitable machine learning algorithm, and tuning its hyperparameters. Even experts can spend days, weeks, or months finding well-performing pipelines and may still make mistakes when optimizing their pipelines. The course will discuss meta-algorithmic approaches to automatically search for and obtain well-performing machine learning systems through AutoML.

Requirements

Participants are strongly recommended to have a solid foundation in machine learning (ML) and deep learning (DL). Additionally, hands-on experience with:

• Python (3.6+)
• Machine learning
• Deep learning is highly recommended. It is also expected that participants have attended at least one other course on ML and DL in the past.

Topics

The lectures are partitioned into several parts, including:

• Hyperparameter Optimization
• Bayesian Optimization for Hyperparameter Optimization
• Neural Architecture Search
• Dynamic Configuration Analysis and Interpretability of AutoML
• Algorithm Selection/Meta-Learning

Organization

The course will be taught in a flipped-classroom style, with weekly combined Q/A sessions and exercises. A new exercise sheet will be provided roughly every week, with most exercises being practical and involving programming in Python and teamwork.

Grading and Assessment

The grading of exercises will be done via a designated platform. The final assessment will involve a student conference, where groups of three students will write a short report on one of three potential topics, peer-review the reports, and present their work in a virtual poster session.

Exam Details

• Report: 4 pages + #students pages appendix
• Contributions of each student need to be described for the report
• A paper template will be provided
• Peer review: Each student has to write a review for one other paper
• Poster Conference: Jointly with Hannover Online in a virtual environment
• Oral presentation of the poster during a predefined time-slot

Important Dates

• 20.07.2022: Project handout
• 31.08.2022: Report Deadline
• 09.09.2022: Review Deadline
• 13.09.2022 (16:00 CEST): Poster Submission Deadline
• 15 - 16.09.2022: Poster Session
• 23.09.2022: Final Deadline (incorporate feedback from the poster session)