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

Chrome logo Add to Chrome

Firefox logo Add to Firefox
Picture for Carlo Baldassi

Carlo Baldassi

Wide flat minima and optimal generalization in classifying high-dimensional Gaussian mixtures


Oct 27, 2020
Carlo Baldassi, Enrico M. Malatesta, Matteo Negri, Riccardo Zecchina

Figure 1 for Wide flat minima and optimal generalization in classifying high-dimensional Gaussian mixtures
Figure 2 for Wide flat minima and optimal generalization in classifying high-dimensional Gaussian mixtures
Figure 3 for Wide flat minima and optimal generalization in classifying high-dimensional Gaussian mixtures
Figure 4 for Wide flat minima and optimal generalization in classifying high-dimensional Gaussian mixtures
* 18 pages, 4 figures. arXiv admin note: text overlap with arXiv:2006.07897 
  Access Paper or Ask Questions

Ergodic Annealing


Aug 01, 2020
Carlo Baldassi, Fabio Maccheroni, Massimo Marinacci, Marco Pirazzini

Figure 1 for Ergodic Annealing
Figure 2 for Ergodic Annealing
Figure 3 for Ergodic Annealing

  Access Paper or Ask Questions

Entropic gradient descent algorithms and wide flat minima


Jun 14, 2020
Fabrizio Pittorino, Carlo Lucibello, Christoph Feinauer, Enrico M. Malatesta, Gabriele Perugini, Carlo Baldassi, Matteo Negri, Elizaveta Demyanenko, Riccardo Zecchina

Figure 1 for Entropic gradient descent algorithms and wide flat minima
Figure 2 for Entropic gradient descent algorithms and wide flat minima
Figure 3 for Entropic gradient descent algorithms and wide flat minima
Figure 4 for Entropic gradient descent algorithms and wide flat minima
* 24 pages (main text: 8 pages) 
  Access Paper or Ask Questions

Multialternative Neural Decision Processes


May 16, 2020
Carlo Baldassi, Simone Cerreia-Vioglio, Fabio Maccheroni, Massimo Marinacci, Marco Pirazzini


  Access Paper or Ask Questions

Clustering of solutions in the symmetric binary perceptron


Nov 18, 2019
Carlo Baldassi, Riccardo Della Vecchia, Carlo Lucibello, Riccardo Zecchina

Figure 1 for Clustering of solutions in the symmetric binary perceptron
Figure 2 for Clustering of solutions in the symmetric binary perceptron
Figure 3 for Clustering of solutions in the symmetric binary perceptron
Figure 4 for Clustering of solutions in the symmetric binary perceptron

  Access Paper or Ask Questions

Natural representation of composite data with replicated autoencoders


Sep 29, 2019
Matteo Negri, Davide Bergamini, Carlo Baldassi, Riccardo Zecchina, Christoph Feinauer

Figure 1 for Natural representation of composite data with replicated autoencoders
Figure 2 for Natural representation of composite data with replicated autoencoders
Figure 3 for Natural representation of composite data with replicated autoencoders
Figure 4 for Natural representation of composite data with replicated autoencoders
* 11 pages, 4 figures 
  Access Paper or Ask Questions

On the geometry of solutions and on the capacity of multi-layer neural networks with ReLU activations


Jul 17, 2019
Carlo Baldassi, Enrico M. Malatesta, Riccardo Zecchina

Figure 1 for On the geometry of solutions and on the capacity of multi-layer neural networks with ReLU activations
Figure 2 for On the geometry of solutions and on the capacity of multi-layer neural networks with ReLU activations
Figure 3 for On the geometry of solutions and on the capacity of multi-layer neural networks with ReLU activations
* 11 pages, 3 figures 
  Access Paper or Ask Questions

Shaping the learning landscape in neural networks around wide flat minima


May 21, 2019
Carlo Baldassi, Fabrizio Pittorino, Riccardo Zecchina

Figure 1 for Shaping the learning landscape in neural networks around wide flat minima
Figure 2 for Shaping the learning landscape in neural networks around wide flat minima
Figure 3 for Shaping the learning landscape in neural networks around wide flat minima
Figure 4 for Shaping the learning landscape in neural networks around wide flat minima
* 46 pages (20 main text), 7 figures (v2 fixes fig. 1 and a few typos in formulas) 
  Access Paper or Ask Questions

Recombinator-k-means: Enhancing k-means++ by seeding from pools of previous runs


May 01, 2019
Carlo Baldassi

Figure 1 for Recombinator-k-means: Enhancing k-means++ by seeding from pools of previous runs
Figure 2 for Recombinator-k-means: Enhancing k-means++ by seeding from pools of previous runs
Figure 3 for Recombinator-k-means: Enhancing k-means++ by seeding from pools of previous runs
Figure 4 for Recombinator-k-means: Enhancing k-means++ by seeding from pools of previous runs
* 11 pages, 3 figures, 3 tables 
  Access Paper or Ask Questions

On the role of synaptic stochasticity in training low-precision neural networks


Mar 20, 2018
Carlo Baldassi, Federica Gerace, Hilbert J. Kappen, Carlo Lucibello, Luca Saglietti, Enzo Tartaglione, Riccardo Zecchina

Figure 1 for On the role of synaptic stochasticity in training low-precision neural networks
Figure 2 for On the role of synaptic stochasticity in training low-precision neural networks
Figure 3 for On the role of synaptic stochasticity in training low-precision neural networks
Figure 4 for On the role of synaptic stochasticity in training low-precision neural networks
* Phys. Rev. Lett. 120, 268103 (2018) 
 * 7 pages + 14 pages of supplementary material 
  Access Paper or Ask Questions

Efficiency of quantum versus classical annealing in non-convex learning problems


Oct 16, 2017
Carlo Baldassi, Riccardo Zecchina

Figure 1 for Efficiency of quantum versus classical annealing in non-convex learning problems
Figure 2 for Efficiency of quantum versus classical annealing in non-convex learning problems
Figure 3 for Efficiency of quantum versus classical annealing in non-convex learning problems
* Proceedings of the National Academy of Sciences Jan 2018, 201711456 
 * 31 pages, 10 figures 
  Access Paper or Ask Questions

Parle: parallelizing stochastic gradient descent


Sep 10, 2017
Pratik Chaudhari, Carlo Baldassi, Riccardo Zecchina, Stefano Soatto, Ameet Talwalkar, Adam Oberman

Figure 1 for Parle: parallelizing stochastic gradient descent
Figure 2 for Parle: parallelizing stochastic gradient descent
Figure 3 for Parle: parallelizing stochastic gradient descent
Figure 4 for Parle: parallelizing stochastic gradient descent

  Access Paper or Ask Questions

Entropy-SGD: Biasing Gradient Descent Into Wide Valleys


Apr 21, 2017
Pratik Chaudhari, Anna Choromanska, Stefano Soatto, Yann LeCun, Carlo Baldassi, Christian Borgs, Jennifer Chayes, Levent Sagun, Riccardo Zecchina

Figure 1 for Entropy-SGD: Biasing Gradient Descent Into Wide Valleys
Figure 2 for Entropy-SGD: Biasing Gradient Descent Into Wide Valleys
Figure 3 for Entropy-SGD: Biasing Gradient Descent Into Wide Valleys
Figure 4 for Entropy-SGD: Biasing Gradient Descent Into Wide Valleys
* ICLR '17 
  Access Paper or Ask Questions

Unreasonable Effectiveness of Learning Neural Networks: From Accessible States and Robust Ensembles to Basic Algorithmic Schemes


Oct 06, 2016
Carlo Baldassi, Christian Borgs, Jennifer Chayes, Alessandro Ingrosso, Carlo Lucibello, Luca Saglietti, Riccardo Zecchina

Figure 1 for Unreasonable Effectiveness of Learning Neural Networks: From Accessible States and Robust Ensembles to Basic Algorithmic Schemes
Figure 2 for Unreasonable Effectiveness of Learning Neural Networks: From Accessible States and Robust Ensembles to Basic Algorithmic Schemes
Figure 3 for Unreasonable Effectiveness of Learning Neural Networks: From Accessible States and Robust Ensembles to Basic Algorithmic Schemes
Figure 4 for Unreasonable Effectiveness of Learning Neural Networks: From Accessible States and Robust Ensembles to Basic Algorithmic Schemes
* Proc. Natl. Acad. Sci. U.S.A. 113(48):E7655-E7662, 2016 
 * 31 pages (14 main text, 18 appendix), 12 figures (6 main text, 6 appendix) 
  Access Paper or Ask Questions

Learning may need only a few bits of synaptic precision


May 27, 2016
Carlo Baldassi, Federica Gerace, Carlo Lucibello, Luca Saglietti, Riccardo Zecchina

Figure 1 for Learning may need only a few bits of synaptic precision
Figure 2 for Learning may need only a few bits of synaptic precision
Figure 3 for Learning may need only a few bits of synaptic precision
Figure 4 for Learning may need only a few bits of synaptic precision
* Phys. Rev. E 93, 052313 (2016) 
 * 38 pages (main text: 16 pages), 5 figures; http://link.aps.org/doi/10.1103/PhysRevE.93.052313 
  Access Paper or Ask Questions

Local entropy as a measure for sampling solutions in Constraint Satisfaction Problems


Feb 25, 2016
Carlo Baldassi, Alessandro Ingrosso, Carlo Lucibello, Luca Saglietti, Riccardo Zecchina

Figure 1 for Local entropy as a measure for sampling solutions in Constraint Satisfaction Problems
Figure 2 for Local entropy as a measure for sampling solutions in Constraint Satisfaction Problems
Figure 3 for Local entropy as a measure for sampling solutions in Constraint Satisfaction Problems
Figure 4 for Local entropy as a measure for sampling solutions in Constraint Satisfaction Problems
* J. Stat. Mech. 2016 (2) 023301 
 * 46 pages (main text: 22), 7 figures. This is an author-created, un-copyedited version of an article published in Journal of Statistical Mechanics: Theory and Experiment. IOP Publishing Ltd is not responsible for any errors or omissions in this version of the manuscript or any version derived from it. The Version of Record is available online at http://dx.doi.org/10.1088/1742-5468/2016/02/023301 
  Access Paper or Ask Questions

Subdominant Dense Clusters Allow for Simple Learning and High Computational Performance in Neural Networks with Discrete Synapses


Sep 18, 2015
Carlo Baldassi, Alessandro Ingrosso, Carlo Lucibello, Luca Saglietti, Riccardo Zecchina

Figure 1 for Subdominant Dense Clusters Allow for Simple Learning and High Computational Performance in Neural Networks with Discrete Synapses
Figure 2 for Subdominant Dense Clusters Allow for Simple Learning and High Computational Performance in Neural Networks with Discrete Synapses
Figure 3 for Subdominant Dense Clusters Allow for Simple Learning and High Computational Performance in Neural Networks with Discrete Synapses
* Physical Review Letters, 15, 128101 (2015) url=http://journals.aps.org/prl/abstract/10.1103/PhysRevLett.115.128101 
 * 11 pages, 4 figures (main text: 5 pages, 3 figures; Supplemental Material: 6 pages, 1 figure) 
  Access Paper or Ask Questions

A Max-Sum algorithm for training discrete neural networks


Aug 13, 2015
Carlo Baldassi, Alfredo Braunstein

Figure 1 for A Max-Sum algorithm for training discrete neural networks
Figure 2 for A Max-Sum algorithm for training discrete neural networks
Figure 3 for A Max-Sum algorithm for training discrete neural networks
Figure 4 for A Max-Sum algorithm for training discrete neural networks
* Journal of Statistical Mechanics: Theory and Experiment 2015, no. 8, P08008 
  Access Paper or Ask Questions

Efficient supervised learning in networks with binary synapses


Jul 09, 2007
Carlo Baldassi, Alfredo Braunstein, Nicolas Brunel, Riccardo Zecchina

Figure 1 for Efficient supervised learning in networks with binary synapses
Figure 2 for Efficient supervised learning in networks with binary synapses
Figure 3 for Efficient supervised learning in networks with binary synapses
Figure 4 for Efficient supervised learning in networks with binary synapses
* PNAS 104, 11079-11084 (2007) 
 * 10 pages, 4 figures 
  Access Paper or Ask Questions