Albert Catalan-Tatjer

Profile Picture

I am a first year PhD student at the safety- & efficiency- aligned learning (🦭) group advised by Jonas Geiping funded by IMPRS-IS and Amazon Science Hub. I am excited about most things deep learning, especially optimization and inference efficiency. Fundamental to my research is the implementation of mathematical ideas on computer technologies. On my free time I enjoy science and sports.

I am currently working on inference for multi-agent systems, fine-tunability/continual learning, and tokenization.

News

Publications

Decoding class dynamics in learning with noisy labels.

Albert Catalan-Tatjer, Bhalaji Nagarajan, Ricardo Marques, Petia Radeva. Pattern Recognition Letters 2024

CCLM: Class-conditional label noise modelling.

Albert Catalan-Tatjer, Bhalaji Nagarajan, Ricardo Marques, Petia Radeva. IbPRIA 2023 Best Paper