Explore a comprehensive lecture on the Automatic Statistician, delivered by Professor Zoubin Ghahramani from the University of Cambridge and the Alan Turing Institute. Delve into Bayesian model selection strategies for automated model selection and human-readable report generation. Learn about the ingredients of an automatic statistician, including the language of regression models, Gaussian processes, and composition rules. Examine real-world applications through examples like the Mauna Loa Keeling Curve analysis and discover how this approach achieves good predictive performance. Gain insights into model checking, criticism, and the rational allocation of computational resources in this cutting-edge field of machine learning and statistics.
Overview
Syllabus
Intro
AUTOMATING MACHINE LEARNING
THE AUTOMATIC STATISTICIAN
INGREDIENTS OF AN AUTOMATIC STATISTICIAN
DEFINING A LANGUAGE OF REGRESSION MODELS
WE CAN BUILD REGRESSION MODELS WITH GAUSSIAN PROCESSES
A PICTURE: GPS, LINEAR AND LOGISTIC REGRESSION, AND SVMS
THE COMPOSITION RULES OF OUR LANGUAGE
MODEL SEARCH: MAUNA LOA KEELING CURVE
EXAMPLE: AN ENTIRELY AUTOMATIC ANALYSIS
EXAMPLE REPORTS
GOOD PREDICTIVE PERFORMANCE AS WELL
MODEL CHECKING AND CRITICISM
RATIONAL ALLOCATION OF COMPUTATIONAL RESOURCES
CONCLUSIONS
Taught by
Alan Turing Institute
