Description
The 73rd Seminar
Date and Time: September 11th 2:00pm – 3:00 pm (JST)
Venue: Online and Open Space at the RIKEN AIP Nihonbashi office
Language: English
Speaker: Jörg Wicker (University of Auckland)
Title: Reliable Machine Learning – Methods and Applications in Environmental Sciences
Short Abstract:
I will discuss reliability of Machine Learning (ML) algorithms, from the perspective of adversarial learning and its implications in terms of model quality and robustness given new data. The talk will highlight how adversarial learning can be used as a valuable tool to measure and improve reliability, particularly in the domain of environmental sciences.
I will explore adversarial defenses against evasion and poisoning attacks, as well as attacks on time series data. Defenses are crucial for maintaining the integrity of ML systems, but also give us a way to better understand model behavior and propose improvements.
Highlighting the central role of reliable ML in environmental sciences, I will showcase our work in atmospheric chemistry, recognized with an Ig Nobel Prize. In this work, we identified links between breath and smell and emotional responses of an audience in a cinema.
Furthermore, I will cheminformatics, I will illustrate how ML contributes to predicting metabolic pathways for environmental pollutants.
Concluding with an outlook, I will discuss our innovative approach of integrating mechanistic models with AI for freshwater modeling, encompassing adversarial techniques. This holistic approach holds great promise for addressing urgent ecological challenges.
Bio:
Joerg Simon Wicker is a Senior Lecturer at the School of Computer Science at the University of Auckland, founding CEO and the current CTO of enviPath, and leads the Machine Learning Group at University of Auckland.Before his current role, Joerg had a Postdoctoral position at the University of Mainz in Germany, and earned his PhD from the Technical University of Munich.
Joerg’s research covers both the development of machine learning algorithms and their practical application across various domains, including bioinformatics, cheminformatics, and environmental sciences. His current focus is on reliability of machine learning algorithms, exploring adversarial machine learning, and addressing bias within the field. These works have exciting applications in domains such as chemistry, epidemiology, and environmental research.
With a rich publication history, Joerg has contributed significantly to the scientific community as an author of over 45 peer-review papers in machine learning, cheminformatics, environmental sciences, and other disciplines, as well as a member of the program committee of top-ranked conferences. In recognition of his contributions to the field, he was awarded the Ig Nobel Prize in Chemistry in 2021. For a comprehensive overview of his ongoing research projects and recent developments, please visit his lab’s webpage (https://wickerlab.org).
