top of page
DSC_1057_edited_edited_edited_edited_edi

Research Fellow
Theoretical Physics, University of Oxford

UKRI Horizon Europe Guarantee Marie Skłodowska-Curie Fellowship Award
Project: Phenomenology of Deep Learning [ukri.org]

nayara.fonsecadesa@physics.ox.ac.uk | nayara.focs@gmail.com

News

  • 19 November: I will participate in the Physics for AI and AI for Physics: Landscaping Workshop at the Institute of Physics, London.
     

  • 1 November: I'm giving a talk at the AI Security Reading Group, Mathematical Institute, University of Oxford.
     

  • 28-29 October: I'm attending the 'Beyond the symbols vs signals debate' meeting at The Royal Society (London).
     

  • September: Our paper, An exactly solvable model for emergence and scaling laws, was accepted at NeurIPS 2024!
     

  • March-June: I'm participating in the AI Alignment Course run by BlueDot Impact. Please feel free to reach out if you'd like to exchange ideas or explore potential collaborations!

Benasque_May_2018_edited.jpg
RESEARCH
Hamburg_1.jpeg

ABOUT ME

I'm a Research Fellow at Oxford Theoretical Physics, where I'm exploring the interplay between Machine Learning and Physics. I'm hosted at Ard Louis group. Before Oxford, I worked as a Research Scientist at IBM Research UK. Previously, I was at ICTP in Trieste, DESY Theory Group in Hamburg, CTP at MIT, University of São Paulo, and University of Campinas.

 

For more details, please see my [CV].

I am a physicist seeking ways to understand generalization in deep learning. Toward this goal, I study AI as a scientific discipline, where both empirical validity and explainable methodologies are essential. For instance, see [here] for our work on emergence and scaling laws and [here] for our work on double descent and online/offline correspondence in similarity learning. 

 

Previously, I explored the construction of models to explain small- and large-scale phenomena in the universe. I worked on the interplay between particle physics and cosmology, learning from the early universe how to build and interpret our models of particle physics. See [here] for all publications. 

bottom of page