
Andrew Gordon Wilson
Professor of Computer Science and Data Science · Computer Science — PhD Advisors · New York University (NYU)
At Prof Rod’s reading desk
Questions to explore
- When does a model-selection score disagree with performance on unseen data?
- How do prior assumptions change the model you select?
Research interests
Andrew Gordon Wilson studies generalization, foundation models and uncertainty.
- Generalization
- Foundation models
- Uncertainty
Explore the work
Related opportunities
- Explore a PhD in foundation-model theory and generalization
PhD opening · Standing invitation
- Explore postdoctoral research in foundation models and uncertainty
Postdoc opening · Standing invitation
Prof Rod’s reading notes
Start with a paper
ICML · 2022
Bayesian Model Selection, the Marginal Likelihood, and Generalization
Prof Rod’s reading note: Study why marginal likelihood can disagree with generalization. Separate the assumptions behind the score from the held-out test that supports a model choice.
Bring a question of your own
Turn reading into research
Choose two models that fit the same training observations. Write down one assumption that favors each model and a held-out test that could reverse your choice.
Keep exploring
More at New York University (NYU)
Explore the universityImprove this page
Suggest a correction, share a paper to start with, or point readers to a group page or opening.