Our group is open to any research collaborations. Students interested in a Master's thesis in our group please send an email to aurelien.lucchi@unibas.ch with short research interests, transcript of records, and CV.

Research projects

Deep Learning

• Stochastic Gradient Langevin Dynamics: SDE Foundations and Sign-Based Improvements

Optimization

• Memory-Efficient Optimizers for Training LLMs

• Benchmarking Muon-type Algorithms for LLM Training

• Optimization Under Generalized Smoothness 

• Scheduling precision for quantized training 

Quantum

• Computational Complexity of Approximated Parameterized Quantum Models

• Spectral Bias in Classical and Quantum Neural Networks

• Spectral de-biasing in hybrid classical-quantum neural networks  (INDUSTRY-PARTNERED)

Continual Learning

• Comparing LoRA with Replay Methods Under the Same Memory Constraint