Data Driven Modelling

Data-driven modeling uses machine learning to create models from data, aiming to understand and predict complex systems where traditional analytical approaches are insufficient. Current research emphasizes developing robust models for partially observed systems and improving the accuracy and interpretability of predictions, employing architectures like neural ordinary differential equations, neural networks combined with symbolic regression, and diffusion models. These advancements are impacting diverse fields, from weather forecasting and power grid management to brain activity analysis and battery temperature prediction, enabling more accurate simulations and improved decision-making. A key challenge remains ensuring data quality and avoiding trivial solutions, particularly in data-scarce scenarios.