Dr Kacper Sokol visits Università della Svizzera italiana to deliver new course on machine learning explainability
Author Kathy Nickels
Date 27 March 2023
Research Fellow Dr Kacper Sokol from the ARC Centre of Excellence for Automated Decision-Making and Society (ADM+S), RMIT University has recently visited Università della Svizzera italiana (USI) in Lugano, Switzerland to deliver training on machine learning explainability.
The training was developed to bridge the gap between the theoretical and practical aspects of explainability and interpretability of predictive models based on artificial intelligence and machine learning algorithms, and builds upon Dr Sokol’s research in this area.
Dr Sokol says that the course differs from others that commonly take an abstract approach.
“It takes an adversarial perspective and breaks these techniques up into core functional blocks, studies their role and configuration, and reassembles them to create bespoke explainers with well-understood properties, thus making them suitable for the problem at hand,” he said.
The course was offered along with other training opportunities available to postgraduate students from the informatics department at USI.
“Given its good reception and high modularity of the teaching materials, it will be adapted to support a variety of future training sessions,” said Dr Skolol.
The course resources are available online at Machine Learning Explainability: Exploring Automated Decision-Making Through Transparent Modelling and Peeking Inside Black Boxes.
This training is the most recent output stemming from Dr Sokol’s ongoing collaboration with Professor Marc Langheinrich and his Ubiquitous Computing Research Group at USI. Together they work on advancing explainability and interpretability of machine learning models. They recently presented BayCon: Model-agnostic Bayesian Counterfactual Generator at the 31st International Joint Conference on Artificial Intelligence 2022 (IJCAI-22) in Vienna, Austria.