Follow
Diego Granziol
Title
Cited by
Cited by
Year
Fast information-theoretic Bayesian optimisation
B Ru, MA Osborne, M McLeod, D Granziol
International Conference on Machine Learning, 4384-4392, 2018
542018
Learning rates as a function of batch size: A random matrix theory approach to neural network training
D Granziol, S Zohren, S Roberts
Journal of Machine Learning Research 23 (173), 1-65, 2022
30*2022
Entropic trace estimates for log determinants
J Fitzsimons, D Granziol, K Cutajar, M Osborne, M Filippone, S Roberts
Machine Learning and Knowledge Discovery in Databases: European Conference …, 2017
262017
MEMe: An accurate maximum entropy method for efficient approximations in large-scale machine learning
D Granziol, B Ru, S Zohren, X Dong, M Osborne, S Roberts
Entropy 21 (6), 551, 2019
192019
Towards understanding the true loss surface of deep neural networks using random matrix theory and iterative spectral methods
D Granziol, T Garipov, D Vetrov, S Zohren, S Roberts, AG Wilson
172019
Beyond random matrix theory for deep networks
D Granziol
arXiv preprint arXiv:2006.07721, 2020
162020
Appearance of random matrix theory in deep learning
NP Baskerville, D Granziol, JP Keating
Physica A: Statistical Mechanics and its Applications 590, 126742, 2022
132022
MLRG deep curvature
D Granziol, X Wan, T Garipov, D Vetrov, S Roberts
arXiv preprint arXiv:1912.09656, 2019
122019
Flatness is a false friend
D Granziol
arXiv preprint arXiv:2006.09091, 2020
102020
Iterative averaging in the quest for best test error
D Granziol, X Wan, S Albanie, S Roberts
arXiv preprint arXiv:2003.01247, 2020
62020
Iterate averaging helps: An alternative perspective in deep learning
D Granziol, X Wan, S Roberts
stat 1050, 2, 2020
52020
Deep curvature suite
D Granziol, X Wan, T Garipov
arXiv preprint arXiv:1912.09656, 2019
52019
Universal characteristics of deep neural network loss surfaces from random matrix theory
NP Baskerville, JP Keating, F Mezzadri, J Najnudel, D Granziol
Journal of Physics A: Mathematical and Theoretical 55 (49), 494002, 2022
32022
A random matrix theory approach to damping in deep learning
D Granziol, N Baskerville
Journal of Physics: Complexity 3 (2), 024001, 2022
32022
Ranker-agnostic contextual position bias estimation
OB Mayor, V Bellini, A Buchholz, G Di Benedetto, DM Granziol, M Ruffini, ...
arXiv preprint arXiv:2107.13327, 2021
32021
Applicability of Random Matrix Theory in Deep Learning
NP Baskerville, D Granziol, JP Keating
arXiv e-prints, arXiv: 2102.06740, 2021
32021
Explaining the Adapative Generalisation Gap
D Granziol, S Albanie, X Wan, S Roberts
stat 1050, 15, 2020
32020
Gadam: Combining adaptivity with iterate averaging gives greater generalisation
D Granziol, X Wan, S Roberts
stat 1050, 10, 2020
32020
VBALD-Variational Bayesian approximation of log determinants
D Granziol, E Wagstaff, BX Ru, M Osborne, S Roberts
arXiv preprint arXiv:1802.08054, 2018
32018
The Deep Learning Limit: are negative neural network eigenvalues just noise?
D Granziol, T Garipov, S Zohren, D Vetrov, S Roberts, AG Wilson
ICML 2019 workshop on theoretical physics for deep learning, 2019
22019
The system can't perform the operation now. Try again later.
Articles 1–20