We use computers to learn from data and make better decisions — developing new optimization and learning algorithms, integrating learning with decision-making, and deploying AI systems for real-world impact.
This is our GitHub landing page. For our official page visit iol.zib.de
Our research spans optimization theory, algorithm design, and machine-learning applications.
Conditional gradient algorithms, non-smooth and distributed optimization, accelerated methods, and online learning.
Mixed-integer (nonlinear) programming, GPU-accelerated techniques, cutting planes, branch-and-bound, and the SCIP Optimization Suite.
First-order methods, convex and min-max optimization, and optimization for deep learning.
Model compression, training efficiency, decentralized training, robustness, XAI, AI4Science, and Agentic AI.
Computer-assisted proofs, flag algebras in combinatorics, interactive theorem provers, and learning-based heuristics.
Tensor decompositions, operator theory, numerical PDE methods, kernel-based techniques, and Bell nonlocality.
Human-AI interaction, AI for learning and education, decision-making with AI, and AI in AR/VR.
Autonomous systems, AI agents, cyber-physical systems, smart manufacturing, mobility, and digital twins.
We maintain research software used across academia and industry.
Conditional gradient algorithms in Julia
Branch-and-bound with Frank–Wolfe
Local polytope membership & Bell inequalities
We co-develop the SCIP Optimization Suite — one of the fastest academically developed MIP solvers.
From One Trivial Observation at a Time — mathematics, optimization, machine learning, and AI.
Loading posts…
A diverse group of mathematicians, computer scientists, and engineers working together in Berlin.