1.5k citations
- University of BirminghamGB29 papers
- Centre National de la Recherche ScientifiqueFR22 papers
- Novosibirsk State UniversityRU14 papers
- Tokyo Institute of TechnologyJP13 papers
- Hong Kong University of Science and TechnologyHK12 papers
- Institute of Automation and ElectrometryRU9 papers
- University of Electronic Science and Technology of ChinaCN9 papers
- Weizmann Institute of ScienceIL9 papers
- Education University of Hong KongHK8 papers
- Laboratoire Interdisciplinaire Carnot de BourgogneFR8 papers
- Université de BourgogneFR8 papers
- B. Verkin Institute for Low Temperature Physics and Engineering of the National Academy of Sciences of UkraineUA7 papers
54 papers · 1 filter
Large Deviation Analysis of Function Sensitivity in Random Deep Neural Networks
Bo Li, David Saad
Mean field theory has been successfully used to analyze deep neural networks (DNN) in the infinite size limit. Given the finite size of realistic DNN, we utilize the large deviatio…
Exploring the Function Space of Deep-Learning Machines
Bo Li, David Saad
The function space of deep-learning machines is investigated by studying growth in the entropy of functions of a given error with respect to a reference function, realized by a dee…
Slow Spin Dynamics and Self-Sustained Clusters in Sparsely Connected Systems
Jacopo Rocchi, David Saad, Chi Ho Yeung
To identify emerging microscopic structures in low temperature spin glasses, we study self-sustained clusters (SSC) in spin models defined on sparse random graphs. A message-passin…
High storage capacity in the Hopfield model with auto-interactions - stability analysis
Jacopo Rocchi, David Saad, Daniele Tantari
Recent studies point to the potential storage of a large number of patterns in the celebrated Hopfield associative memory model, well beyond the limits obtained previously. We inve…
Self-Sustained Clusters as Drivers of Computational Hardness in -spin Models
Jacopo Rocchi, David Saad, Chi Ho Yeung
While macroscopic properties of spin glasses have been thoroughly investigated, their manifestation in the corresponding microscopic configurations is much less understood. Cases w…
Shortest node-disjoint paths on random graphs
Caterina De Bacco, Silvio Franz, David Saad +1
A localized method to distribute paths on random graphs is devised, aimed at finding the shortest paths between given source/destination pairs while avoiding path overlaps at nodes…