Скачать или смотреть LARS KOTHOFF: Optimizing Predictive Performance through Automated Machine Learning

Recording of a lecture by Lars Kothoff on Automated Machine Learning. The seminar took place on June 16, 2023 at IDEAS NCBR in Warsaw.

Abstract:
Machine Learning is ubiquitous, but Machine Learning expertise is not. Choosing the right ML approach, tuning its hyperparameters, and getting good performance is a challenging task, especially for ML novices. In this talk, I will introduce the field of Automated Machine Learning, which aims to make this task easier for novices and experts alike. At the heart of AutoML are meta-algorithmic techniques that automatically learn how to use existing approaches better, rather than inventing new ones. I will give a brief overview of techniques and approaches to AutoML and pointers on how you can use it to get better ML performance.

Bio:
Lars Kotthoff is an Associate Professor of Computer Science at the University of Wyoming and has previously held positions at the University of British Columbia, University College Cork, and the University of St Andrews. He is known for contributions to the field of meta-algorithmics, which includes automated parameter tuning, and helped to birth the new field of automated machine learning. Lars has applied machine learning in different areas of AI and other disciplines, for example materials science, and is one of the core contributors to a popular machine learning framework in R. His research has been widely cited and supported by the National Science Foundation, NASA, and Microsoft, among others.