This thesis explores the application of machine learning techniques to combinatorial optimization problems in computing systems, with a particular focus on those with timing constraints. Many modern computing systems, including real-time embedded systems, cloud infrastructures, and high-performance computing, require efficient decision-making to ensure optimal performance while meeting resource and scheduling requirements. Traditional optimization techniques often face challenges in scalability and real-time adaptability, making machine learning a promising alternative.

The research will investigate both reinforcement learning and unsupervised learning methods to tackle complex combinatorial optimization problems. Reinforcement learning is of interest due to its ability to learn optimal decision policies through interaction with the system, making it well-suited for dynamic and adaptive optimization tasks. Unsupervised learning, on the other hand, will be explored for its capability to discover hidden patterns and structures in system behavior, enabling efficient approximation and optimization strategies. 

This study aims to develop learning-based optimization algorithms that balance computational efficiency, scalability, and real-time feasibility. By integrating machine learning with combinatorial optimization, the thesis seeks to create adaptive, intelligent, and scalable solutions for performance-critical computing systems, with applications in real-time scheduling, resource allocation, and load balancing. The findings are expected to contribute significantly to both theoretical advancements and practical implementations in computing systems design optimization. 

Some basic knowledge in:

- solid programming background

- operations research

- combinatorial optimization

- graph theory

- reinforcement learning

- unsupervised learning

- graph neural networks

Please send your application email with cv and transcripts (in English) to binqi.sun@tum.de. 

(Students from CIT are welcome to apply for this topic as an IDP or Master's thesis)