2013 Workshop on
Statistical Physics of Disordered Systems
and Its Applications (SPDSA2013)
--- Prologue Series V of FSPIP2013 ---



Schedule: 20-21 March, 2013

Venue: Meeting Room Horai in Akiu Resort Hotel Sakan (Access)
    Yumoto Akiu, Taihaku-Ku, Sendai 982-0241, Japan

Scientific Program:

      Organized session only: We do not accept submissions from non-invited speakers.

    20 March, 2013

      13:20-13:30 Opening
        Kazuyuki Tanaka (Graduate School of Information Sciences, Tohoku University, Japan)

      13:30-15:00 Session 1
        Kazuyuki Tanaka (Graduate School of Information Sciences, Tohoku University, Japan)
          Title: Bayesian Image Modeling by Generalized Sparse Markov Random Fields and Loopy Belief Propagation
          Abstract:
            Bayesian image modeling is given for based on a generalized sparse prior probability. Our prior include a sparsity in each interaction term between every pair of neighbouring pixels in Markov random fields. A new scheme for hyperparameter estimations is given from the standpoint of the conditional maximization for entropy in our generalized sparse prior. Moreover, the criteria of the optimal value for sparseness in interactions is also given by using the maximization of marginal likelihood. Our practical algorithm is constructed by using the loopy belief propagation.
          References:
          1. M. Yasuda, S. Kataoka and K. Tanaka: Inverse problem in pairwise Markov random fields using loopy belief propagation, Journal of the Physical Society of Japan, Vol.81, No.4, Article No.044801, 2012.
          2. K. Tanaka, M. Yasuda and D. M. Titterington: Bayesian image modeling by means of generalized sparse prior and loopy belief propagation, Journal of the Physical Society of Japan, Vol.81, No.11, Article No.114802, 2012.
        Federico Ricci-Tersenghi (Dipartmento di Fisica, Universitá di Roma, "La Sapienza", Italy)
          Title: Making maximum entropy and linear response estimates consistent
          Abstract
            Mean field approximations are the most widespread method for predicting marginals in models of interacting variables (direct inference problem) and for estimating model parameters (e.g. couplings and fields) from the measured magnetizations and pairwise correlations (inverse inference problem). In mean field approximations, correlations can be obtained both from the maximum entropy principle (minimizing the approximated free energy) and by using the linear response theorem. Unfortunately these two estimates are usually different. We propose a new framework where these two estimates are force to be equal. The resulting method improves over existing mean field methods especially in models with frustrated interactions and short loops. Joint work with Jack Raymond.

      15:00-15:15 Break

      15:15-17:30 Session 2

        Tomoyuki Obuchi (Cybermedia Center, Osaka University, Japan)
          Title: Partition-Function Zeros of Spin Glasses and Singularities in Quantum Quench Dynamics
          Abstract:
            Although Yang-Lee theory of partition-function zeros compactly demonstrated that phase transitions are expressed by singularities of free energy, physical significances of partition-function zeros have been less derived. In this talk, I will provide several patterns of partition-function zeros of some spin-glass models derived by the replica and cavity methods, and discuss the physical consequence for the replica symmetry breaking and the chaos effect of spin glasses. As a more recent progress, dynamical singularities of spin glasses in a quantum quench are analyzed based on the theory of partition-function zeros. The details will also be presented.
          References:
          1. T. Obuchi and K. Takahashi: Dynamical singularities of glassy systems in a quantum quench, Physical Review E, Vol.86, No.5, Article No.051125, 2012.
          2. T. Obuchi and K. Takahashi: Partition-function zeros of spherical spin glasses and their relevance to chaos, Journal of Physics A: Mathematical and Theoretical, Vol.45, No.12, Article No.125003, 2012
          3. Y. Matsuda, M. Müller, H. Nishimori, T. Obuchi and A. Scardicchio: Distribution of partition function zeros for the ±J model on the Bethe lattice, Journal of Physics A: Mathematical and Theoretical, Vol.43, No.28, Article No.285002, 2010.
        Tommaso Rizzo (Dipartmento di Fisica, Universitá di Roma, "La Sapienza", Italy)
          Title: Replica Cluster Variational Method
          Abstract:
            Replica Cluster Variational Method is a general formalism to implement the Replica-Symmetric (RS) and Replica-Symmetry-Breaking ansatz in the context of Kikuchi's Cluster Variational Method (CVM). I will present and discuss the derivation of the main equations and their physical interpretation. Recently the method has been applied to the 2D random bond Ising Model yielding predictions on the spin-glass transition temperatures and the multicritical point largely improved over the Bethe predictions. The solution of the RS equation deep in the glassy phase has also been obtained by means of a generalization of population dynamics to negative functions. I will discuss these recent developments and their connection with single-instance message-passing algorithms.
          References:
          1. T. Rizzo, A. Lage-Castellanos, R. Mulet and F. Ricci-Tersenghi: Replica Cluster Variational Method, Journal of Statistical Physics, Vol.139, No.3, p.375, 2010.
          2. E. Dominguez, A. Lage, R.Mulet, F. Ricci-Tersenghi and T. Rizzo: Characterizing and improving generalized belief propagation algorithms on the 2D Edwards-Anderson model, Journal of Statistical Mechanics: Theory and Experiment, Article No.P12007, 2011.
          3. A. Lage-Castellanos, R. Mulet, F. Ricci-Tersenghi and T. Rizzo: Inference algorithm for finite-dimensional spin glasses: Belief propagation on the dual lattice, Physical Review E, Vol.84, No.4, Article No.046706, 2011.
          4. A. Lage-Castellanos, R. Mulet, F. Ricci-Tersenghi and T. Rizzo: Replica Cluster Variational Method: the Replica Symmetric solution for the 2D random bond Ising model, Journal of Physics A: Mathematical and Theoretical, Vol.46, No.13, Article No.135001, 2013.
        Shun Kataoka (Graduate School of Information Sciences, Tohoku University, Japan)
          Title: Gaussian graphical modeling for traffic density estimation
          Reference:
            S. Kataoka, M. Yasuda and K. Tanaka: Statistical Analysis of Gaussian Image Inpainting Problems, Journal of the Physical Society of Japan, Vol.81, No.2, Article No.025001, 2012.

    21 March, 2013
      10:00-12:15 Session 3
        Takehisa Hasegawa (Graduate School of Information Sciences, Tohoku University, Japan)
          Title: Classification of Phase Transitions in Complex Networks
        Ryuhei Mori (Graduate School of Informatics, Kyoto University, Japan)
          Title: New Generalization of the Bethe Approximation using Graph Covers
          Reference:
            R. Mori and T. Tanaka: New generalizations of the Bethe approximation via asymptotic expansion, 2012 IEICE SITA, Oita, Japan, 11-14 December, 2012 (arXiv:1210.2592).
        Masayuki Ohzeki (Graduate School of Informatics, Kyoto University, Japan)
          Title: Variational Principle in Statistical Mechanics
      12:15-12:30 Closing
        Masayuki Ohzeki (Graduate School of Informatics, Kyoto University, Japan)

Organized by

    Masayuki Ohzeki (Graduate School of Informatics, Kyoto University, Japan)
    Kazuyuki Tanaka (Graduate School of Information Sciences, Tohoku University, Japan)

Sponsered by

    Center of Education and Research for Information Electronics Systems, ECEI, Tohoku University, Japan

Supported by

    Graduate School of Information Sciences (GSIS), Tohoku University, Japan

Related Webpages

    Global COE (Center of Excellence) Program "Center of Education and Research for Information Electronics Systems"
    Frontier of Statistical Physics and Information Processing --- Perspectives from Nonequilibrium Behaviors --- (FSPIP2013) (July, 2013, Kyoto, Japan)
    Group of Electrical Engineering, Communication Engineering, Electronic Engineering, and Information Engineering (ECEI), Tohoku University, Japan

Contact to SPDSA2013 office