Seminar 1: BrIAS Fellow Prof. Andrea Roli
Title: Dynamical criticality: from cells to robots
Abstract: Systems that exhibit complex behaviours are often found in a particular dynamical condition, poised between order and disorder. This observation is at the core of the so-called criticality hypothesis, which states that systems in a dynamical regime between order and disorder attain the highest level of computational capabilities and achieve an optimal trade-off between robustness and flexibility. Recent results in cellular and evolutionary biology, neuroscience and computer science have revitalised the interest in the criticality hypothesis, emphasising its role as a viable candidate general law in adaptive complex systems. Dynamical criticality provides a powerful principle for designing adaptive artificial systems, as well as for addressing fundamental challenges related to human impact on nature. In this seminar I will review the main features of dynamical criticality and illustrate notable cases and recent results in biology, computer science and robotics.
Seminar 2: BrIAS Fellow Prof. Luca Magri
Title: Scientific machine learning for chaotic forecasting and real-time digital twinning
Abstract: The ability of modelling reality to predict the evolution of complex systems is enabled by principles and empirical approaches. Physical principles, for example conservation laws, are extrapolative (until the assumptions upon which they hinge break down): they provide predictions on phenomena that have not been observed. Human beings are excellent at extrapolating knowledge because we are excellent at finding physical principles. On the other hand, empirical modelling provides correlation functions within data, which are useful when principles are difficult to deduce. Artificial intelligence and machine learning are excellent at empirical modelling. In this talk, the complementary capabilities of both approaches will be merged (scientific machine learning). The approaches will achieve real-time modelling and optimization of nonlinear, unsteady and uncertain dynamical systems with chaotic and turbulent dynamics, which exhibit extreme events. The focus of the talk is on computational methodologies for modelling and optimization: (i) data assimilation with a Bayesian approach in which the model errors (unknown unknowns) are inferred in real-time, (ii) Bayesian optimization to optimize wind farms and (iii) auto-encoders to predict extreme events in turbulent flows. The flows under investigation are relevant to aerospace propulsion, with a focus on thermoacoustics, turbulence, and renewable energies. The talk will be concluded with some lessons learnt, and a discussion on future directions.
