Mirko Goldmann receives international award for research carried out at IFISC

• The award recognizes research carried out at IFISC showing that recurrent neural networks can remain adaptive beyond the training phase.
• The awarded study introduces a conceptor-based adaptive control loop that enhances network robustness and performance under perturbations.

Researcher Mirko Goldmann has been awarded the Edward N. Lorenz Early Career Award, granted annually by the journal Chaos to outstanding early-career scientists in the field of nonlinear science.

The distinction recognizes his work “Adaptive control of recurrent neural networks using conceptors”, study developed during Goldmann’s research period at the Institute for Cross-Disciplinary Physics and Complex Systems (IFISC), in collaboration with Guillaume Pourcel (University of Groningen) and Ingo Fischer (IFISC, UIB-CSIC) and Miguel C. Soriano (IFISC, UIB-CSIC). The work, carried out in the context of the POST-DIGITAL project, focuses on recurrent neural networks and shows that these systems can remain adaptive even after the training phase, challenging the conventional view that their internal parameters must remain fixed once learning is complete.

Using the conceptor framework, the researchers introduced an adaptive control loop that allows neural networks to continuously adjust their internal representations. This approach improves performance in tasks such as interpolating temporal patterns, maintaining functionality despite partial network damage, and resisting distorted inputs.

According to Chaos, the study enhances both the robustness and interpretability of neural networks, opening new possibilities for adaptive machine learning systems and hardware implementations. The award highlights the international impact of research conducted at IFISC and its contribution to advancing interdisciplinary approaches in nonlinear science.

• Celebrating the 2024 Edward N. Lorenz Early Career Award Winners