arXiv subjects

Yukito Iba

Publications and source records attributed to Yukito Iba.

24 records · Page 2Linked to original sources

arXiv1999

Multi-Self-Overlap Ensemble for protein folding: ground state search and thermodynamics

Long chains of the HP lattice protein model are studied by the Multi-Self-Overlap Ensemble(MSOE) Monte Carlo method, which was developed recently by the authors. MSOE successfully finds the lowest energy states reported before for sequences of the chain length $N=42\sim 100$ in two and three dimensions. Moreover, MSOE realizes the lowest energy state that ever found in a case of N=100. Finite-temperature properties of these sequences are also investigated by MSOE. Two successive transitions are observed between the native and random coil states. Thermodynamic analysis suggests that the ground state degeneracy is relevant to the order of the transitions in the HP model.

cond-mat.soft↗

arXiv1998

Simulation of Spin Glass with a Variable-system-size Ensemble

In this paper, we introduce a dynamical Monte Carlo algorithm for spin models in which the number of the spins fluctuates from zero to a given number by addition and deletion of spins with a probabilistic rule. Such simulations are realized with a variable-system-size ensemble, a mixture of canonical ensembles each of which corresponds to a system with different size. The weight of each component of the mixture is controlled by a penalty term and systematically tuned in a preliminary run in a way similar to the multicanonical algorithm. In a measurement run, the system grows and shrinks without violating the detailed balance condition and we can obtain the correct canonical averages if physical quantities is measured only when its size is equal to the prescribed maximum size. The mixing of Markov chain is facilitated by the fast relaxation at small system sizes. The algorithm is implemented for the SK model of spin glass and shows better performance than that of a conventional heat bath algorithm.

cond-mat.stat-mech↗

arXiv1998

The Nishimori line and Bayesian Statistics

``Nishimori line'' is a line or hypersurface in the parameter space of systems with quenched disorder, where simple expressions of the averages of physical quantities over the quenched random variables are obtained. It has been playing an important role in the theoretical studies of the random frustrated systems since its discovery around 1980. In this paper, a novel interpretation of the Nishimori line from the viewpoint of statistical information processing is presented. Our main aim is the reconstruction of the whole theory of the Nishimori line from the viewpoint of Bayesian statistics, or, almost equivalently, from the viewpoint of the theory of error-correcting codes. As a byproduct of our interpretation, counterparts of the Nishimori line in models without gauge invariance are given. We also discussed the issues on the ``finite temperature decoding'' of error-correcting codes in connection with our theme and clarify the role of gauge invariance in this topic.

cond-mat.dis-nn↗

arXiv1998

Mean Field Approximation in Bayesian Variable Selection

Variable selection for a multiple regression model (Noisy Linear Perceptron) is studied with a mean field approximation. In our Bayesian framework, variable selection is formulated as estimation of discrete parameters that indicate a subset of the explanatory variables. Then, a mean field approximation is introduced for the calculation of the posterior averages over the discrete parameters. An application to a real world example, Boston housing data, is shown.

cond-mat.dis-nn↗

arXiv1998

Design Equation: A Novel Approach to Heteropolymer Design

A novel approach to heteropolymer design is proposed. It is based on the criterion by Kurosky and Deutsch, with which the probability of a target conformation in a conformation space is maximized at low but finite temperature. The key feature of the proposed approach is the use of soft spins (fuzzy monomers) that leads to a design equation, which is an analog of the Boltzmann machine learning equation in the design problem. We implement an algorithm based on the design equation for the generalized HP model on the 3x3x3 cubic lattice and check its performance.

cond-mat.dis-nn↗

arXiv1998

Simulation of Lattice Polymers with Multi-Self-Overlap Ensemble

A novel family of dynamical Monte Carlo algorithms for lattice polymers is proposed. Our central idea is to simulate an extended ensemble in which the self-avoiding condition is systematically weakened. The degree of the self-overlap is controlled in a similar manner as the multicanonical ensemble. As a consequence, the ensemble --the multi-self-overlap ensemble-- contains adequate portions of self-overlapping conformations as well as higher energy ones. It is shown that the multi-self-overlap ensemble algorithm reproduce correctly the canonical averages at finite temperatures of the HP model of lattice proteins. Moreover, it outperforms massively a standard multicanonical algorithm for a difficult example of a polymer with 8-stickers. Alternative algorithm based on exchange Monte Carlo method is also discussed.

cond-mat↗