ترغب بنشر مسار تعليمي؟ اضغط هنا

Inference of compressed Potts graphical models

194   0   0.0 ( 0 )
 نشر من قبل Remi Monasson
 تاريخ النشر 2019
  مجال البحث فيزياء
والبحث باللغة English




اسأل ChatGPT حول البحث

We consider the problem of inferring a graphical Potts model on a population of variables, with a non-uniform number of Potts colors (symbols) across variables. This inverse Potts problem generally involves the inference of a large number of parameters, often larger than the number of available data, and, hence, requires the introduction of regularization. We study here a double regularization scheme, in which the number of colors available to each variable is reduced, and interaction networks are made sparse. To achieve this color compression scheme, only Potts states with large empirical frequency (exceeding some threshold) are explicitly modeled on each site, while the others are grouped into a single state. We benchmark the performances of this mixed regularization approach, with two inference algorithms, the Adaptive Cluster Expansion (ACE) and the PseudoLikelihood Maximization (PLM) on synthetic data obtained by sampling disordered Potts models on an Erdos-Renyi random graphs. We show in particular that color compression does not affect the quality of reconstruction of the parameters corresponding to high-frequency symbols, while drastically reducing the number of the other parameters and thus the computational time. Our procedure is also applied to multi-sequence alignments of protein families, with similar results.



قيم البحث

اقرأ أيضاً

We study the stochastic dynamics of infinitely many globally interacting $q$-state units on a ring that is externally driven. While repulsive interactions always lead to uniform occupations, attractive interactions give rise to much richer phenomena: We analytically characterize a Hopf bifurcation which separates a high-temperature regime of uniform occupations from a low-temperature one where all units coalesce into a single state. For odd $q$ below the critical temperature starts a synchronization regime which ends via a second phase transition at lower temperatures, while for even $q$ this intermediate phase disappears. We find that interactions have no effects except below critical temperature for attractive interactions. A thermodynamic analysis reveals that the dissipated work is reduced in this regime, whose temperature range is shown to decrease as $q$ increases. The $q$-dependence of the power-efficiency trade-off is also analyzed.
Phase transition of the two- and three-state quantum Potts models on the Sierpinski pyramid are studied by means of a tensor network framework, the higher-order tensor renormalization group method. Critical values of the transverse magnetic field and the magnetic exponent $beta$ are evaluated. Despite the fact that the Hausdorff dimension of the Sierpinski pyramid is exactly two $( = log_2^{~} 4)$, the obtained critical properties show that the effective dimension is lower than two.
We investigate the two-dimensional $q=3$ and 4 Potts models with a variable interaction range by means of Monte Carlo simulations. We locate the phase transitions for several interaction ranges as expressed by the number $z$ of equivalent neighbors. For not too large $z$, the transitions fit well in the universality classes of the short-range Potts models. However, at longer ranges the transitions become discontinuous. For $q=3$ we locate a tricritical point separating the continuous and discontinuous transitions near $z=80$, and a critical fixed point between $z=8$ and 12. For $q=4$ the transition becomes discontinuous for $z > 16$. The scaling behavior of the $q=4$ model with $z=16$ approximates that of the $q=4$ merged critical-tricritical fixed point predicted by the renormalization scenario.
The phase transitions of random-field q-state Potts models in d=3 dimensions are studied by renormalization-group theory by exact solution of a hierarchical lattice and, equivalently, approximate Migdal-Kadanoff solutions of a cubic lattice. The recu rsion, under rescaling, of coupled random-field and random-bond (induced under rescaling by random fields) coupled probability distributions is followed to obtain phase diagrams. Unlike the Ising model (q=2), several types of random fields can be defined for q >= 3 Potts models, including random-axis favored, random-axis disfavored, random-axis randomly favored or disfavored cases, all of which are studied. Quantitatively very similar phase diagrams are obtained, for a given q for the three types of field randomness, with the low-temperature ordered phase persisting, increasingly as temperature is lowered, up to random-field threshold in d=3, which is calculated for all temperatures below the zero-field critical temperature. Phase diagrams thus obtained are compared as a function of $q$. The ordered phase in the low-q models reaches higher temperatures, while in the high-q models it reaches higher random fields. This renormalization-group calculation result is physically explained.
106 - Michael Chertkov 2008
It is well known that an arbitrary graphical model of statistical inference defined on a tree, i.e. on a graph without loops, is solved exactly and efficiently by an iterative Belief Propagation (BP) algorithm convergent to unique minimum of the so-c alled Bethe free energy functional. For a general graphical model on a loopy graph the functional may show multiple minima, the iterative BP algorithm may converge to one of the minima or may not converge at all, and the global minimum of the Bethe free energy functional is not guaranteed to correspond to the optimal Maximum-Likelihood (ML) solution in the zero-temperature limit. However, there are exceptions to this general rule, discussed in cite{05KW} and cite{08BSS} in two different contexts, where zero-temperature version of the BP algorithm finds ML solution for special models on graphs with loops. These two models share a key feature: their ML solutions can be found by an efficient Linear Programming (LP) algorithm with a Totally-Uni-Modular (TUM) matrix of constraints. Generalizing the two models we consider a class of graphical models reducible in the zero temperature limit to LP with TUM constraints. Assuming that a gedanken algorithm, g-BP, funding the global minimum of the Bethe free energy is available we show that in the limit of zero temperature g-BP outputs the ML solution. Our consideration is based on equivalence established between gapless Linear Programming (LP) relaxation of the graphical model in the $Tto 0$ limit and respective LP version of the Bethe-Free energy minimization.
التعليقات
جاري جلب التعليقات جاري جلب التعليقات
سجل دخول لتتمكن من متابعة معايير البحث التي قمت باختيارها
mircosoft-partner

هل ترغب بارسال اشعارات عن اخر التحديثات في شمرا-اكاديميا