Vittorio Erba
Scholar

Vittorio Erba

Google Scholar ID: mUgeZ3YAAAAJ
EPFL, Lausanne
Statistical mechanics of combinatorial optimization problems and machine learning
Homepage↗Google Scholar↗
Citations & Impact
All-time
Citations
193
 
H-index
7
 
i10-index
6
 
Publications
20
 
Co-authors
22
list available
Contact
Emailvittorio.erba@epfl.chCVOpen ↗GitHubOpen ↗
Publications
7 items
A solvable high-dimensional model where nonlinear autoencoders learn structure invisible to PCA while test loss misaligns with generalization
2026
Cited
0
Inductive Bias and Spectral Properties of Single-Head Attention in High Dimensions
2025
Cited
0
Scaling Laws and Spectra of Shallow Neural Networks in the Feature Learning Regime
2025
Cited
0
The maximum-average subtensor problem: equilibrium and out-of-equilibrium properties
2025
Cited
0
Bayes optimal learning of attention-indexed models
2025
Cited
0
The Nuclear Route: Sharp Asymptotics of ERM in Overparameterized Quadratic Networks
2025
Cited
0
Bilinear Sequence Regression: A Model for Learning from Long Sequences of High-dimensional Tokens
arXiv.org · 2024
Cited
0
Resume (English only)
Research Experience
  • Work Experience: Physicist at EPFL; Position: Teacher and student supervisor.
Background
  • Research Interests: Statistical mechanics, spin glass theory and its application to understanding computational phenomena; Brief Introduction: Works as a physicist at EPFL in Lausanne, also teaches and supervises students.
Miscellany
  • Personal Interests: Books, crosswords, hikes, websites, chess.
Co-authors
22 total
Co-author 1
Co-author 1
Lenka Zdeborová
Lenka Zdeborová
EPFL, Switzerland
Florent Krzakala
Florent Krzakala
École polytechnique fédérale de Lausanne
Emanuele Troiani
Emanuele Troiani
École polytechnique fédérale de Lausanne
Co-author 5
Co-author 5
Co-author 6
Co-author 6
Co-author 7
Co-author 7
Antoine Maillard
Antoine Maillard
Inria Paris