Skip to main content

Resultados relacionados

Resultados totales (Incluyendo duplicados): 4
Encontrada(s) 1 página(s)

RiuNet. Repositorio Institucional de la Universitat Politécnica de Valéncia
Publication Artículo científico (JournalArticle)

Efficient training of energy-based models via spin-glass control

  • Pozas-Kerstjens, Alejandro
  • Muñoz-Gil, Gorka
  • Piñol, Eloy
  • Acín, Antonio
  • Lewenstein, Maciej
  • Grzybowski, Przemyslaw R.
  • Garcia March, Miguel Angel|||0000-0001-7092-838X
[EN] We introduce a new family of energy-based probabilistic graphical models for efficient unsupervised learning. Its definition is motivated by the control of the spin-glass properties of the Ising model described by the weights of Boltzmann machines. We use it to learn the Bars and Stripes dataset of various sizes and the MNIST dataset, and show how they quickly achieve the performance offered by standard methods for unsupervised learning. Our results indicate that the standard initialization of Boltzmann machines with random weights equivalent to spin-glass models is an unnecessary bottleneck in the process of training. Furthermore, this new family allows for very easy access to low-energy configurations, which points to new, efficient training algorithms. The simplest variant of such algorithms approximates the negative phase of the log-likelihood gradient with no Markov chain Monte Carlo sampling costs at all, and with an accuracy sufficient to achieve good learning and generalization., ML and AA groups acknowledge the Spanish Ministry MINECO and State Research Agency AEI (FIDEUA PID2019-106901GBI00/10.13039/501100011033, Severo Ochoa Grant Nos. SEV-2015-0522 and CEX2019-000910-S, FPI), the European Social Fund, Fundacio Cellex, Fundacio Mir-Puig, Generalitat de Catalunya (AGAUR Grant Nos. 2017 SGR 1341 and SGR 1381, CERCA program, QuantumCAT U16-011424, co-funded by ERDF Operational Program of Catalonia 2014-2020), ERC AdG NOQIA and CERQUTE, EU FEDER, MINECO-EU QUANTERA MAQS (funded by the State Research Agency AEI PCI2019-111828-2/10.13039/501100011033), the National Science Centre, Poland-Symfonia Grant No. 2016/20/W/ST4/00314 and the AXA Chair in Quantum Information Science. A P-K acknowledges funding from Fundacio Obra Socialla Caixa' (LCF/BQ/ES15/10360001) and the European Union's Horizon 2020 research and innovation programme-Grant Agreement No. 648913. G M-G acknowledges funding from Fundacio Obra Social 'la Caixa' (LCF-ICFO grant). M A G-M acknowledges funding from the Spanish Ministry of Education and Vocational Training (MEFP) through the Beatriz Galindo program 2018 (BEAGAL18/00203).
Dipòsit Digital de la UB
Publication Artículo científico (JournalArticle)

Probing Thermalization through Spectral Analysis with Matrix Product Operators

  • Yang, Yilun
  • Iblisdir, Sofyan
  • Cirac, Juan Ignacio
  • Bañuls, Mari Carmen
We combine matrix product operator techniques with Chebyshev polynomial expansions and present a method that is able to explore spectral properties of quantum many-body Hamiltonians. In particular, we show how this method can be used to probe thermalization of large spin chains without explicitly simulating their time evolution, as well as to compute full and local densities of states. The performance is illustrated with the examples of the Ising and PXP spin chains. For the nonintegrable Ising chain, our findings corroborate the presence of thermalization for several initial states, well beyond what direct timedependent simulations have been able to achieve so far.
DIGITAL.CSIC. Repositorio Institucional del CSIC
Publication Artículo científico (JournalArticle)

Projected Entangled Pair States: Fundamental Analytical and Numerical Limitations

  • Scarpato, G.
  • Molnár, A.
  • Ge, Y.
  • García-Ripoll, Juan José
  • Schuch, N.
  • Pérez-García, D.
  • Iblisdir, S.
6 pags., 2 figs., Matrix product states and projected entangled pair states (PEPS) are powerful analytical and numerical tools to assess quantum many-body systems in one and higher dimensions, respectively. While matrix product states are comprehensively understood, in PEPS fundamental questions, relevant analytically as well as numerically, remain open, such as how to encode symmetries in full generality, or how to stabilize numerical methods using canonical forms. Here, we show that these key problems, as well as a number of related questions, are algorithmically undecidable, that is, they cannot be fully resolved in a systematic way. Our work thereby exposes fundamental limitations to a full and unbiased understanding of quantum many-body systems using PEPS., This project has received funding from the European Research Council (ERC) under the European Union’s Horizon 2020 research and innovation programme (Grants No. 648913 and No. 636201). S. I., D. P. G., and G. S. acknowledge financial support from MINECO (Grants No. MTM2014-54240-P and No. MTM2017- 88385-P), CAM research consortium QUITEMAD-CM (Grant No. P2018/TCS-4342), and Severo Ochoa Project No. SEV-2015-556. S. I. also acknowledges funding by the Ministerio de Ciencias, Innovacion y Universidades, PGC2018-095862-B-C22 (Spain) and by Quantum CAT (001-P-001644), Generalitat de Catalunya, and María de Maeztu Project No. CEX2019-000918-M (Spain). J. J. G. R. acknowledges financial support from MINECO Project No. FIS2015-70856-P, CAM research consortion QUITEMAD-CM Grant No. P2018/TCS-4342, MINECO research network FIS2016-81891-REDT. A. M., Y. G., and N. S. acknowledge support from the DFG (German Research Foundation) under Germany’s Excellence Strategy (EXC2111-390814868).
Recercat. Dipósit de la Recerca de Catalunya
Publication Artículo científico (JournalArticle)

Probing Thermalization through Spectral Analysis with Matrix Product Operators

  • Yang, Yilun
  • Iblisdir, Sofyan
  • Cirac, Juan Ignacio
  • Bañuls, Mari Carmen
We combine matrix product operator techniques with Chebyshev polynomial expansions and present a method that is able to explore spectral properties of quantum many-body Hamiltonians. In particular, we show how this method can be used to probe thermalization of large spin chains without explicitly simulating their time evolution, as well as to compute full and local densities of states. The performance is illustrated with the examples of the Ising and PXP spin chains. For the nonintegrable Ising chain, our findings corroborate the presence of thermalization for several initial states, well beyond what direct timedependent simulations have been able to achieve so far.