PhD Position: Physical Learning in Dynamical Systems

This PhD position at AMOLF’s Learning Machines Group offers a unique opportunity to investigate how learning emerges in physical systems—from biological networks to adaptive materials—operating far from equilibrium. As part of an ERC Starting Grant project, the successful candidate will develop a theoretical framework for physical learning, combining analytical theory with numerical simulations. The work spans condensed matter physics, non-equilibrium statistical mechanics, and machine learning, aiming to uncover fundamental principles like scaling laws and phase boundaries. The PhD student will contribute to cutting-edge research at the intersection of physics and AI, with support from a collaborative, inclusive environment at Amsterdam Science Park.

• Develop local learning rules for dynamical physical systems without centralized optimization or backpropagation.
• Compare physical learning approaches to idealized gradient-based machine learning methods.
• Investigate constraints on learning imposed by locality, causality, non-reciprocity, and dissipation.
• Study diverse adaptive dynamical networks (flow, mechanical, neuronal systems) via analytical theory and large-scale simulations.
• Identify scaling laws, phase boundaries, and universal features of learning in matter.
• Develop efficient simulation tools for adaptive dynamical networks.
• Contribute to establishing a new physics of adaptive matter.