Prof. Paolo Missier – University of Birmingham, UK
Part II of the PhD teaching “Scalable Data Science – Explanable AI”, described here, includes 6 hours and follows on from part I, as follows:
- Part I introduced methods for providing explanations to ML models (XAI). We focused specifically on Influence Functions. In this part we consider the problem of extending those explanations to the data processing that occurs when raw data, possibly from multiple sources, is engineered into a training set. We will introduce key concepts of data and process provenance, and explain how they apply to the problem, leading to “data explanations”, using a specific implementation we have developed over the years. We then connect the tool to the Influence Function framework, resulting in “End-to-End explanations” (XEE).
- We introduce well-known concepts of “algorithmic fairness” and provide definitions that apply specifically to classification models. We then introduce methods to ensure a model is “fair”, or at least “fairness-aware”, and finally we present our recent work on ensuring continued fairness in the presence of data drifts
The programme is structured as follows.
- About 3 hours of seminar-style lectures, using key papers including from our own group
- About 3 hours of in-depth study, guided by relevant literature. Students are required to select recent work from a portfolio of recommended papers, and to present their insights to the class
Class schedule (all classes take place in the Math building M018 of the FIM Department, via Campi 213/b – Modena)
Mon 11-11 h.15:00-17:00 – Meeting Room 2° floor – MO18 Building
Thu 13-11 h. 9:00-11:00 – Class Room M1.6
Thu 13-11 h. 11:30-13:30 – Class Room M2.4
Prof. Paolo Missier. is Professor of Data Management in the School of Computer Science, and the Director of the Institute for Data and AI at University of Birmingham.
Bio, research interests, and link to research portal can be found here:
https://www.birmingham.ac.uk/staff/profiles/computer-science/academic-staff/missier-paolo
Host Prof. Federica Mandreoli