Lecture
LCN2 seminar January 2025
- Date
- Friday 31 January 2025
- Time
- Address
- Gorlaeus
- Room
- BM. 2.26 (second floor)
71th LCN2 seminar
Speaker: Lucia Cavallaro (Radboud University)
Title: Network Science meets Machine Learning: Insights and Industrial Challenges
Abstract: Machine learning models, particularly artificial neural networks, are rapidly evolving to tackle real-world challenges. Yet, as models grow in complexity, new questions arise about their efficiency, interpretability, and optimization. This talk aims of bridging the gap between network science and machine learning, presenting how these domains complement each other to address industrial challenges. We begin by exploring pruning techniques for artificial neural networks, demonstrating how edge-level sparsity can be achieved without sacrificing accuracy. From there, we overview the powerful toolkit of network science exploring directions such as similarity metrics, centrality measures, and more to gain insights on graph neural networks and deep learning models at large. By leveraging network properties, we uncover ways to optimize model setups and training strategies for specific applications. Finally, we briefly discuss a practical use case from industrial power grids, showcasing how network science can drive innovation and efficiency in addressing challenges in this critical sector.
Snacks and drinks will be available after the seminar at the Fusiebar.