This doctoral research will be at the intersection of sparsity and artificial intelligence. The research will investigate the potential of sparse-to-sparse training of deep neural networks within reinforcement learning frameworks. This innovative approach holds promise for creating highly efficient and scalable AI systems capable of learning with limited data and computational resources, pertinent in areas such as autonomous systems, online resource allocation, and complex decision-making processes.

Main Responsibilities:

• Conduct original research on sparse-to-sparse training techniques, exploring new frontiers in algorithmic development for DRL.
• Investigate the mathematical underpinnings of sparsity in deep reinforcement learning and its effects on learning dynamics, and generalization.
• Design and evaluate experiments to validate the effectiveness of sparse-to-sparse training in various scenarios and benchmarks.
• Publish and present research findings in top-tier conferences (e.g., Machine Learning, JMLR) and journals (e.g., NeurIPS, ICLR, ICML, IJCAI, AAMAS, ECMLPKDD).
• Collaborate with a international team of researchers and industry partners.

The successful candidate will be embedded in the DMB research group, and the supervision will be ensured by Dr. Elena Mocanu and Prof. dr. Maurice van Keulen. This PhD position is part of the Modular Integrated Sustainable Datacenter (MISD) project and will have ample collaboration opportunities. As part of the MISD project effort led by Elena Mocanu, we are opening multiple positions (two Ph.D. candidates and one PostDoc) to join us and work at the intersection of dynamic sparse training in neural networks on various tasks.

Useful links:

• Elena Mocanu webpage
• DMB research group
• MISD project
• Sample of our work on sparsity

The candidate is expected to have

• A master degree (or will shortly, acquire) in Artificial Intelligence, Computer Science, Mathematics, Engineering, or a related discipline.
• Excellent skills in machine learning and deep learning (experience with deep reinforcement learning is a plus).
• Excellent programming skills (e.g. Python, PyTorch).
• Experience with sparsity in computational models is a plus.
• Good communication skills, with proficiency in English (written and oral).

• As a PhD candidate at UT, you will be appointed to a full-time position for four years, with a qualifier in the first year, within a very stimulating and exciting scientific environment;
• The University offers a dynamic ecosystem with enthusiastic colleagues;
• Your salary and associated conditions are in accordance with the collective labour agreement for Dutch universities (CAO-NU);
• You will receive a gross monthly salary ranging from € 2.770,- (first year) to € 3.539,- (fourth year);
• There are excellent benefits including a holiday allowance of 8% of the gross annual salary, an end-of-year bonus of 8.3%, and a solid pension scheme;
• The flexibility to work (partially) from home;
• A minimum of 232 leave hours in case of full-time employment based on a formal workweek of 38 hours. A full-time employment in practice means 40 hours a week, therefore resulting in 96 extra leave hours on an annual basis.
• Free access to sports facilities on campus
• A family-friendly institution that offers parental leave (both paid and unpaid);
• You will have a training programme as part of the Twente Graduate School where you and your supervisors will determine a plan for a suitable education and supervision;
• We encourage a high degree of responsibility and independence, while collaborating with close colleagues, researchers and other staff.

Our DMB collective stands by its diversity, inclusivity, and interdisciplinary composition. We are doing research at the forefront of advancements in machine learning, deep learning, and computer vision to advance scientific knowledge and societal welfare in a large spectrum of data science applications. We disseminate our research findings through publications in leading conferences (such as NeurIPS, ICLR, ICML, AAMAS, and CVPR) and prestigious journals (e.g. Nature Communications, Machine Learning, etc.).