Two-day format

Following the success of previous years, the event will span over two days. The first day, May 25th, will feature a dedicated session for students, including a series of workshops and talks. The main Complexitat Day is scheduled for May 26th, and will follow its regular format.

Day 1 Program (Warm-up)

Day 2 Program (Jornada)

Confirmed Invited Speakers

We are very happy to announce that this year's invited speakers are:

• Pablo Almaraz
  Instituto de Ciencias Marinas de Andalucía (ICMAN)-CSIC.


• Jordi Garcia-Ojalvo
  Universitat Pompeu Fabra, Barcelona.


• Rachel Lowe
  Barcelona Supercomputing Center, Barcelona.

Program for Warm-Up day
(Day 1, 25th May)

10:00 - 10:30	Registration and Welcome
10:30 - 12:00	Tutorial 1: Joan Garriga Reasoning about complex high-dimensional systems See more
12:00 - 12:30	Coffee Break
12:30 - 14:00	Tutorial 2: Camille Simon Chane Training famous deep-learning models on real-world data See more
14:00 - 15:30	Group Lunch
15:30 - 18:00	Special Activity Data collection in nearby ecological settings

Tutorial 1: Joan Garriga
Reasoning about complex high-dimensional systems

High-dimensional datasets exhibit complex and often non-linear interdependencies between features that cannot be directly apprehended through low-dimensional intuition. Yet these interrelations define the intrinsic structure of the data — its neighborhoods, clusters, decision boundaries, and correlations — which play a central role in analysis and modeling.

Understanding the structure and distribution of data in high-dimensional spaces is therefore a key challenge in complex systems. Many machine learning methods can be viewed as mechanisms for capturing and exploiting this structure, whether through neighborhood preservation, margin maximization, or the learning of latent representations.

In unsupervised settings, where prior knowledge is limited and assumptions are necessarily weak, exploratory visualization becomes a natural first step. Visualization and dimensionality reduction techniques such as PCA, t-SNE, and UMAP provide practical means to probe data organization, revealing structural patterns that would otherwise remain inaccessible.

However, visualization entails unavoidable trade-offs, particularly between preserving local versus global structure. This talk discusses these trade-offs and illustrates how data visualization can serve as a principled tool for understanding and reasoning about complex high-dimensional systems.

Tutorial 2: Camille Simon Chane
Training famous deep-learning models on real-world data

The past ten years has seen a boost in powerful deep-learning models that can tackle many computer vision tasks such as detection and segmentation. However, scientists in many domains are still performing such tasks manually or using non adapted tools. This practical talk will show how to apply a few famous architectures such as SAM − Segment Anything Model and Yolo − You Only Look Once, to niche applications through examples in heritage conservation science and insect recognition.

Program for Complexitat Day
(Day 2, 26th May)

We will keep you informed with updates and additional details as they become available.