Get our free extension to see links to code for papers anywhere online!

Add to Chrome

Add to Firefox

Picture for Diego Granziol

Diego Granziol

Applicability of Random Matrix Theory in Deep Learning

Feb 12, 2021
Nicholas P Baskerville, Diego Granziol, Jonathan P Keating

Figure 1 for Applicability of Random Matrix Theory in Deep Learning

Figure 2 for Applicability of Random Matrix Theory in Deep Learning

Figure 3 for Applicability of Random Matrix Theory in Deep Learning

Figure 4 for Applicability of Random Matrix Theory in Deep Learning

* 12 pages, 8 figures

Access Paper or Ask Questions

Curvature is Key: Sub-Sampled Loss Surfaces and the Implications for Large Batch Training

Jun 16, 2020
Diego Granziol

Figure 1 for Curvature is Key: Sub-Sampled Loss Surfaces and the Implications for Large Batch Training

Figure 2 for Curvature is Key: Sub-Sampled Loss Surfaces and the Implications for Large Batch Training

Figure 3 for Curvature is Key: Sub-Sampled Loss Surfaces and the Implications for Large Batch Training

Figure 4 for Curvature is Key: Sub-Sampled Loss Surfaces and the Implications for Large Batch Training

* 16 pages, 13 figures

Access Paper or Ask Questions

Flatness is a False Friend

Jun 16, 2020
Diego Granziol

Figure 1 for Flatness is a False Friend

Figure 2 for Flatness is a False Friend

Figure 3 for Flatness is a False Friend

Figure 4 for Flatness is a False Friend

* 9 pages, 10 figures

Access Paper or Ask Questions

Beyond Random Matrix Theory for Deep Networks

Jun 13, 2020
Diego Granziol

Figure 1 for Beyond Random Matrix Theory for Deep Networks

Figure 2 for Beyond Random Matrix Theory for Deep Networks

Figure 3 for Beyond Random Matrix Theory for Deep Networks

Figure 4 for Beyond Random Matrix Theory for Deep Networks

* 8 pages 5 Figures

Access Paper or Ask Questions

Iterate Averaging Helps: An Alternative Perspective in Deep Learning

Mar 02, 2020
Diego Granziol, Xingchen Wan, Stephen Roberts

Figure 1 for Iterate Averaging Helps: An Alternative Perspective in Deep Learning

Figure 2 for Iterate Averaging Helps: An Alternative Perspective in Deep Learning

Figure 3 for Iterate Averaging Helps: An Alternative Perspective in Deep Learning

Figure 4 for Iterate Averaging Helps: An Alternative Perspective in Deep Learning

* 9 pages, 8 figures, 21 pages including references and appendix

Access Paper or Ask Questions

MLRG Deep Curvature

Dec 20, 2019
Diego Granziol, Xingchen Wan, Timur Garipov, Dmitry Vetrov, Stephen Roberts

Figure 1 for MLRG Deep Curvature

Figure 2 for MLRG Deep Curvature

Figure 3 for MLRG Deep Curvature

Figure 4 for MLRG Deep Curvature

* 11 pages, 11 figures

Access Paper or Ask Questions

A Maximum Entropy approach to Massive Graph Spectra

Dec 19, 2019
Diego Granziol, Robin Ru, Stefan Zohren, Xiaowen Dong, Michael Osborne, Stephen Roberts

Figure 1 for A Maximum Entropy approach to Massive Graph Spectra

Figure 2 for A Maximum Entropy approach to Massive Graph Spectra

Figure 3 for A Maximum Entropy approach to Massive Graph Spectra

Figure 4 for A Maximum Entropy approach to Massive Graph Spectra

* 12 pages. 9 Figures

Access Paper or Ask Questions

MEMe: An Accurate Maximum Entropy Method for Efficient Approximations in Large-Scale Machine Learning

Jun 03, 2019
Diego Granziol, Binxin Ru, Stefan Zohren, Xiaowen Doing, Michael Osborne, Stephen Roberts

Figure 1 for MEMe: An Accurate Maximum Entropy Method for Efficient Approximations in Large-Scale Machine Learning

Figure 2 for MEMe: An Accurate Maximum Entropy Method for Efficient Approximations in Large-Scale Machine Learning

Figure 3 for MEMe: An Accurate Maximum Entropy Method for Efficient Approximations in Large-Scale Machine Learning

Figure 4 for MEMe: An Accurate Maximum Entropy Method for Efficient Approximations in Large-Scale Machine Learning

* MEMe: An Accurate Maximum Entropy Method for Efficient Approximations in Large-Scale Machine Learning. Entropy, 21(6), 551 (2019)
* 18 pages, 3 figures, Published at Entropy 2019: Special Issue Entropy Based Inference and Optimization in Machine Learning

Access Paper or Ask Questions