New and innovative ideas were discussed at IDA 2024. Photo: Åse Karlén.

The Department of Computer and Systems Sciences (DSV) hosted the international Symposium on Intelligent Data Analysis (IDA) on April 24–26. Being “intentionally small-scale and single-track”, the conference gathered nearly 100 researchers from countries like Germany, the Netherlands, Portugal, United Kingdom, Greece, Italy, Slovenia and France.

Panagiotis Papapetrou, DSV. Photo: Åse Karlén.

“IDA has a unique focus on innovative ideas that can have a potentially high impact”, says DSV professor Panagiotis Papapetrou who was the general chair of the conference.

DSV’s Ioanna Miliou, program committee co-chair, agrees:

“The uniqueness of this conference is in its motto: Ideas over performance. So, if someone has an amazing idea, an innovative concept or methodology, this is prioritized over merely achieving high metrics or the best performance”, she says.

Addressing climate change and pandemics

The three-day-conference covered AI and machine learning themes, including paper presentations on foundational theories, representation learning, and real-world applications.

Key areas like natural language processes, temporal data mining, statistical learning, and optimization were discussed, focusing on improving AI’s understanding and efficiency. Explainable AI was emphasized for its role in making AI decisions transparent and trustworthy.

Ioanna Miliou, DSV. Photo: Carina Bergholm.

“Some of the emerging trends that we discussed were the importance of ethical AI, and interdisciplinary applications. We also addressed global challenges such as climate change and pandemics, raising critical questions about fairness, transparency, and societal impact”, says Miliou.

Robots find human tasks difficult

IDA 2024 featured three keynote speakers. First up was professor Danica Kragic from the KTH Royal Institute of Technology. She shared insights from her research on robots that can assist humans in day-to-day tasks.

Danica Kragic, KTH. Photo: Åse Karlén.

This is a challenging and exciting field, according to Kragic. How do you teach a robot to dress a person, pick grapes or inflate a balloon? These seemingly “easy” tasks are very difficult for a robot, she explained. The person could be running around, grapes are easily crushed, and to blow air into a balloon you need to coordinate many small movements at the same time.

“You’d expect that we’d have robots with hands to help us do the dishes by now. But except for lawn mowers and vacuum cleaners, we haven’t seen many robots come out on consumer markets in the last three decades”, Danica Kragic said in her keynote.

Another aspect of her research has to do with interactions between humans and human-like robots. To be frank, humanoids are quite scary.

“I’m interested in what happens in the human brain when we collaborate with another human being, versus when we collaborate with a machine”, Kragic said.

Dino Pedreschi. Photo: Åse Karlén.

Explainability and fairness

The two other invited keynote speakers were Dino Pedreschi, University of Pisa, and Dimitrios Gunopulos, National and Kapodistrian University of Athens.

“The keynotes were very diverse but all three raised important issues”, says Ioanna Miliou.

“For me, a takeaway from Dimitrios’s speech was the need for explainability and fairness in machine learning so that the produced models and predictions are trustworthy. From Dino’s speech, an important takeaway was the impact of AI on socio-technical systems. The next-generation AIs should team up with humans to help overcome societal problems, rather than exacerbate them”, says Panagiotis Papapetrou.

Dimitrios Gunopulos. Photo: Åse Karlén.

New models to predict engine failure

The conference also included a PhD forum and an Industrial Challenge. DSV’s Tony Lindgren was in charge of the Industrial Challenge where participants came up with ideas on how to predict if – and when – a specific engine component in a truck is at risk of imminent failure.

“We have had a long collaboration with Scania, spanning several years and multiple research projects. We organized a similar industrial challenge together with them in 2016, so it wasn’t a far-fetched idea to repeat this. Of course, the data set was new this year, and we used another type of modeling challenge”, says Tony Lindgren.

He was very impressed with the amount of work that the contestants put in.

Tony Lindgren, DSV. Photo: Åse Karlén.

“It was both interesting and inspiring to see the different approaches they tried out for modeling the problem”.

Lindgren explains that the Industrial Challenge gives a realistic use case for creating a model for predictive maintenance in vehicles.

“The results from the challenge are not applicable directly, as the data have been altered for proprietary reasons. However, the methodologies for creating the models are potentially useable in a real-life scenario. This is what is valuable and interesting for the research community and industry in the field of predictive maintenance”, says Tony Lindgren.

On the final day of the conference, three teams that entered the challenge were awarded for their efforts. Third prize went to Louis Carpentier, Arne De Temmerman and Mathias Verbeke. Second prize went to Maurizio Parton, Andrea Fois, Michelangelo Vegliò, Carlo Metta and Marco Gregnanin. And finally, the challenge winners were Jie Zhong and Zhenkan Wang.

Industrial Challenge accepted – and solved. From left: Tony Lindgren, Louis Carpentier, Arne De Temmerman, Carlo Metta, Zhenkan Wang, Jie Zhong and Jesper Brauer from Scania. Photo: Åse Karlén.

Re-connecting and growing the network

IDA is a European flagship conference in data analysis. As such, it is of high interest to the large group of researchers and PhD students at DSV who are involved in the research area called AI and Data Science.

“We were very happy to be able to organise IDA 2024 at our campus. DSV and Stockholm University focus on social aspects of machine learning and AI – and this is highly related to the objectives and goals of this conference. IDA 2024 provided good opportunities for our PhD students and researchers to network and initiate collaborations with leading experts in the area”, says Panagiotis Papapetrou.

“Yes, we got to re-connect with some of our international colleagues who came to attend the conference and to help with the PhD forum, by mentoring PhD students. We also made new friends and established connections that could lead to future collaborations. Besides that, the food was great – everybody commented on us being Greeks as a guarantee of good food! The social activities included a visit to the Stockholm City Hall as well as a typical Swedish dinner at a restaurant in Djurgården”, says Ioanna Miliou.

The prestigious IDA 2024 Frontier Prize went to Anton Björklund, Lauri Seppäläinen and Kai Puolamäki from University of Helsinki. Their paper is titled “SLIPMAP: Fast and Robust Manifold Visualisation for Explainable AI”.

Proud winners of the IDA 2024 Frontier Prize. From left: Lauri Seppäläinen, Panagiotis Papapetrou, Kai Puolamäki and Anton Björklund. Photo: Åse Karlén.

Text: Åse Karlén