Følg
Stefano SARAO MANNELLI
Stefano SARAO MANNELLI
Assistant Professor, DSAI CSE, Chalmers University of Technology and Gothenburg University
Verifisert e-postadresse på chalmers.se - Startside
Tittel
Sitert av
Sitert av
År
Marvels and pitfalls of the langevin algorithm in noisy high-dimensional inference
S Sarao Mannelli, G Biroli, C Cammarota, F Krzakala, P Urbani, ...
Physical Review X 10 (1), 011057, 2020
642020
Optimization and Generalization of Shallow Neural Networks with Quadratic Activation Functions
S Sarao Mannelli, E Vanden-Eijnden, L Zdeborová
Advances in Neural Information Processing Systems 33, 13445--13455, 2020
63*2020
Passed & Spurious: Descent Algorithms and Local Minima in Spiked Matrix-Tensor Models
S Sarao Mannelli, F Krzakala, P Urbani, L Zdeborova
International Conference on Machine Learning, 4333-4342, 2019
632019
Who is Afraid of Big Bad Minima? Analysis of gradient-flow in spiked matrix-tensor models
S Sarao Mannelli, G Biroli, C Cammarota, F Krzakala, L Zdeborová
Advances in Neural Information Processing Systems, 8676-8686, 2019
472019
Epidemic mitigation by statistical inference from contact tracing data
A Baker, I Biazzo, A Braunstein, G Catania, L Dall’Asta, A Ingrosso, ...
Proceedings of the National Academy of Sciences 118 (32), 2021
402021
Probing transfer learning with a model of synthetic correlated datasets
F Gerace, L Saglietti, S Sarao Mannelli, A Saxe, L Zdeborová
Machine Learning: Science and Technology 3 (1), 2022
342022
Complex Dynamics in Simple Neural Networks: Understanding Gradient Flow in Phase Retrieval
S Sarao Mannelli, G Biroli, C Cammarota, F Krzakala, P Urbani, ...
Advances in Neural Information Processing Systems 33, 3265--3274, 2020
342020
An Analytical Theory of Curriculum Learning in Teacher-Student Networks
L Saglietti, S Sarao Mannelli, A Saxe
Advances in Neural Information Processing Systems 35 (2022), 21113-21127, 2021
262021
Analytical Study of Momentum-Based Acceleration Methods in Paradigmatic High-Dimensional Non-Convex Problems
S Sarao Mannelli, P Urbani
Advances in Neural Information Processing Systems, 2021, 2021
17*2021
Maslow's Hammer for Catastrophic Forgetting: Node Re-Use vs Node Activation
S Lee, S Sarao Mannelli, C Clopath, S Goldt, A Saxe
ICML 2022, 12455-12477, 2022
142022
The RL Perceptron: Generalisation Dynamics of Policy Learning in High Dimensions
N Patel, S Lee, SS Mannelli, S Goldt, A Saxe
arXiv preprint arXiv:2306.10404, 2023
10*2023
Bias-inducing geometries: exactly solvable data model with fairness implications
S Sarao Mannelli, F Gerace, N Rostamzadeh, L Saglietti
arXiv preprint arXiv:2205.15935, 2022
8*2022
Thresholds of descending algorithms in inference problems
S Sarao Mannelli, L Zdeborova
Journal of Statistical Mechanics: Theory and Experiment 2020 (3), 034004, 2020
82020
Why Do Animals Need Shaping? A Theory of Task Composition and Curriculum Learning
JH Lee, S Sarao Mannelli, A Saxe
Forty-first International Conference on Machine Learning, ICML 2024, 2024
72024
Bias in Motion: Theoretical Insights into the Dynamics of Bias in SGD Training
A Jain, R Nobahari, A Baratin, S Sarao Mannelli
Advances in Neural Information Processing Systems, 2025, 2024
42024
How to choose the right transfer learning protocol? A qualitative analysis in a controlled set-up
F Gerace, D Doimo, SS Mannelli, L Saglietti, A Laio
Transactions on Machine Learning Research, 2024
3*2024
Optimal Protocols for Continual Learning via Statistical Physics and Control Theory
F Mori, SS Mannelli, F Mignacco
arXiv preprint arXiv:2409.18061, 2024
22024
Tilting the Odds at the Lottery: the Interplay of Overparameterisation and Curricula in Neural Networks
SS Mannelli, Y Ivashynka, A Saxe, L Saglietti
Forty-first International Conference on Machine Learning, ICML 2024, 2024
2024
A meta-learning framework for rationalizing cognitive fatigue in neural systems
Y Li, R Carrasco-Davis, Y Strittmatter, S Sarao Mannelli, S Musslick
Proceedings of the Annual Meeting of the Cognitive Science Society 46, 2024
2024
A Theory of Initialisation's Impact on Specialisation
D Jarvis, S Lee, CCJ Dominé, AM Saxe, SS Mannelli
NeurIPS 2024 Workshop on Mathematics of Modern Machine Learning, 0
Systemet kan ikke utføre handlingen. Prøv på nytt senere.
Artikler 1–20