Flexible and Scalable Functional Regression using Neural Networks

• Applicant: Professor Dr. David Rügamer
• Subject Area: Statistics and Econometrics ; Image and Language Processing, Computer Graphics and Visualisation, Human Computer Interaction, Ubiquitous and Wearable Computing
• Term: since 2024
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 548823575

Today's challenges of statistical modeling involve millions of data points with heterogeneously structured data sources, data-driven hypotheses, and an unprecedented complexity of data generating processes that statistical models need to be able to express. Especially in the era of digitization, research on neural networks becomes increasingly more important, and integrating statistical models with approaches from deep learning offers a promising and under-researched perspective to tackle diverse and often highly complex modeling tasks. This project aims to bridge the gap between conventional statistical regression and neural networks by developing innovative statistical modeling techniques that are embedded in and optimized through neural networks. Motivated by biomechanics research applications, this proposal aims to: 1) introduce a neural implementation of functional regression models; 2) develop a concept for achieving sparsity in neural additive models through optimization transfer; and 3) integrate these approaches to create a scalable functional regression model that ensures sparsity and is suitable for large-scale applications involving numerous observations and functional predictors.

• DFG Programme: Research Grants