Hands-On:
AutoML Decathlon
Abstract
While the field of machine learning has witnessed tremendous advances in domains such as computer vision and natural language processing, a wide range of areas beyond these traditional ML domains also seek to take advantage of data-driven tools. As a consequence, there is a growing need for ML systems that can adapt to diverse tasks in an efficient and automated fashion. We recently launched the NeurIPS22 AutoML Decathlon competition to help catalyze research in this area and establish a benchmark for the current state of AutoML on diverse tasks. Building on these efforts, the AutoML Fall School Hackathon provides a self-contained conceptual and hands-on introduction to the problem of AutoML for diverse tasks.
We will begin with a short overview of the underlying problem followed by technical details about the competition, which together should allow you to quickly begin crafting and testing methods. To facilitate your hands-on efforts, we will provide a starter kit of code; create scaled-down versions of the competition datasets that can be processed using free tools such as Google Colab; and be available during the event to answer questions and provide support. The aims of the hackathon are for you to (1) learn about the conceptual problem of AutoML for diverse tasks; (2) gain hands-on experience in running and evaluating AutoML methods on a diverse set of tasks; and optionally (3) make an initial submission to the competition itself. Afterwards, we hope your experience will encourage you to continue participating in AutoML Decathlon through its November 10 deadline (and win prizes)!
Bio
Nicholas Roberts is a Ph.D. student in the Computer Sciences Department at the University of Wisconsin-Madison advised by Frederic Sala. His research focus is on principled approaches and emerging applications of automated machine learning (AutoML). In 2021, he obtained his M.S. in Machine Learning at Carnegie Mellon University, and in 2019, he obtained his B.S. in Computer Science at the University of California, San Diego.
Samuel Guo is an M.S. in Machine Learning student at Carnegie Mellon University. In 2021, he obtained his B.S. in Computer Science and B.S. in Statistics at the University of Illinois at Urbana-Champaign. With direction from Ameet Talwalkar, he does research focused on automated machine learning (AutoML) and neural architecture search.
Ameet Talwalkar is an associate professor in the Machine Learning Department at CMU. His work is motivated by the goal of democratizing machine learning, focusing on core challenges related to automation, efficiency, and human-in-the-loop learning. He co-founded Determined AI (acquired by HPE), helped create the MLlib project in Apache Spark, co-authored the textbook 'Foundations of Machine Learning,' and created an award-winning edX MOOC on distributed machine learning. He also helped to start the MLSys conference and is currently President of the MLSys Board.