Regression approaches for large-scale highdimensional data (C04)

• Subject Area: Data Management, Data-Intensive Systems, Computer Science Methods in Business Informatics ; Mathematics ; Theoretical Computer Science
• Term: from 2011 to 2022
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 124020371

The scalability of modern regression approaches is often stretched to its limits when applying them to big data or in embedded systems. The goal of this project is therefore the development of highly efficient regression methods. We pursue the design of algorithms to reduce the number of observations for generalized linear and Bayesian regression models using, e.g. random linear projections and sampling. Furthermore we develop methods to solve nonparametric regression models under resource constraints imposed on their description complexity and structural constraints, e.g. monotonicity.

• DFG Programme: Collaborative Research Centres
• Subproject of: SFB 876: Providing Information by Resource-Constrained Data Analysis
• Applicant Institution: Technische Universität Dortmund
• Project Heads: Professorin Dr. Katja Ickstadt; Dr. Alexander Munteanu, since 10/2019; Professor Dr. Christian Sohler, until 9/2019