Interpretable surrogates for efficient analog timeseries forecasting

Systems with complex dynamics are omnipresent in nature and in everyday life. Examples include systems such as the human heart, traffic models, weather systems and the global climate. For all these examples, there is a desire to be able to predict the temporal development of dynamic variables. For systems that exhibit complex or chaotic dynamics, this is a challenging task. The fact that often not all measurement data is available makes prediction even more complicated. In addition to the ever-growing need for algorithms that can make accurate predictions, there is also a need for these algorithms to be energy-efficient.

The aim of Dr. Jaurigue's research project is to develop computationally efficient algorithms. These can be implemented in physical hardware rather than as software on a conventional computer. The aim is to develop a compact and task-adaptable algorithm that can be used in energy-efficient hardware. The algorithm will be used as edge devices for time series prediction in medical and industrial applications.