Showing 1–50 of 64 results for author: Chatelain, C

CTMRG study of the critical behavior of an interacting-dimer model

Authors: Christophe Chatelain

Abstract: The critical behavior of a dimer model with an interaction favoring parallel dimers in each plaquette of the square lattice is studied numerically by means of the Corner Transfer Matrix Renormalization Group algorithm. The critical exponents are known to depend on the chemical potential of vacancies, or monomers. At large average density of the latter, the phase transition becomes of first-order.… ▽ More The critical behavior of a dimer model with an interaction favoring parallel dimers in each plaquette of the square lattice is studied numerically by means of the Corner Transfer Matrix Renormalization Group algorithm. The critical exponents are known to depend on the chemical potential of vacancies, or monomers. At large average density of the latter, the phase transition becomes of first-order. We compute the scaling dimensions of both order parameter and temperature in the second order regime and compare them with the conjecture that the critical behavior is the same as the Ashkin-Teller model on its self-dual critical line. △ Less

Submitted 27 June, 2024; v1 submitted 26 January, 2024; originally announced January 2024.

Multiple Noises in Diffusion Model for Semi-Supervised Multi-Domain Translation

Authors: Tsiry Mayet, Simon Bernard, Clement Chatelain, Romain Herault

Abstract: Domain-to-domain translation involves generating a target domain sample given a condition in the source domain. Most existing methods focus on fixed input and output domains, i.e. they only work for specific configurations (i.e. for two domains, either $D_1\rightarrow{}D_2$ or $D_2\rightarrow{}D_1$). This paper proposes Multi-Domain Diffusion (MDD), a conditional diffusion framework for multi-doma… ▽ More Domain-to-domain translation involves generating a target domain sample given a condition in the source domain. Most existing methods focus on fixed input and output domains, i.e. they only work for specific configurations (i.e. for two domains, either $D_1\rightarrow{}D_2$ or $D_2\rightarrow{}D_1$). This paper proposes Multi-Domain Diffusion (MDD), a conditional diffusion framework for multi-domain translation in a semi-supervised context. Unlike previous methods, MDD does not require defining input and output domains, allowing translation between any partition of domains within a set (such as $(D_1, D_2)\rightarrow{}D_3$, $D_2\rightarrow{}(D_1, D_3)$, $D_3\rightarrow{}D_1$, etc. for 3 domains), without the need to train separate models for each domain configuration. The key idea behind MDD is to leverage the noise formulation of diffusion models by incorporating one noise level per domain, which allows missing domains to be modeled with noise in a natural way. This transforms the training task from a simple reconstruction task to a domain translation task, where the model relies on less noisy domains to reconstruct more noisy domains. We present results on a multi-domain (with more than two domains) synthetic image translation dataset with challenging semantic domain inversion. △ Less

Submitted 25 September, 2023; originally announced September 2023.

Dynamic Graph Representation Learning with Neural Networks: A Survey

Authors: Leshanshui Yang, Sébastien Adam, Clément Chatelain

Abstract: In recent years, Dynamic Graph (DG) representations have been increasingly used for modeling dynamic systems due to their ability to integrate both topological and temporal information in a compact representation. Dynamic graphs allow to efficiently handle applications such as social network prediction, recommender systems, traffic forecasting or electroencephalography analysis, that can not be ad… ▽ More In recent years, Dynamic Graph (DG) representations have been increasingly used for modeling dynamic systems due to their ability to integrate both topological and temporal information in a compact representation. Dynamic graphs allow to efficiently handle applications such as social network prediction, recommender systems, traffic forecasting or electroencephalography analysis, that can not be adressed using standard numeric representations. As a direct consequence of the emergence of dynamic graph representations, dynamic graph learning has emerged as a new machine learning problem, combining challenges from both sequential/temporal data processing and static graph learning. In this research area, Dynamic Graph Neural Network (DGNN) has became the state of the art approach and plethora of models have been proposed in the very recent years. This paper aims at providing a review of problems and models related to dynamic graph learning. The various dynamic graph supervised learning settings are analysed and discussed. We identify the similarities and differences between existing models with respect to the way time information is modeled. Finally, general guidelines for a DGNN designer when faced with a dynamic graph learning problem are provided. △ Less

Submitted 12 April, 2023; originally announced April 2023.

Faster DAN: Multi-target Queries with Document Positional Encoding for End-to-end Handwritten Document Recognition

Authors: Denis Coquenet, Clément Chatelain, Thierry Paquet

Abstract: Recent advances in handwritten text recognition enabled to recognize whole documents in an end-to-end way: the Document Attention Network (DAN) recognizes the characters one after the other through an attention-based prediction process until reaching the end of the document. However, this autoregressive process leads to inference that cannot benefit from any parallelization optimization. In this p… ▽ More Recent advances in handwritten text recognition enabled to recognize whole documents in an end-to-end way: the Document Attention Network (DAN) recognizes the characters one after the other through an attention-based prediction process until reaching the end of the document. However, this autoregressive process leads to inference that cannot benefit from any parallelization optimization. In this paper, we propose Faster DAN, a two-step strategy to speed up the recognition process at prediction time: the model predicts the first character of each text line in the document, and then completes all the text lines in parallel through multi-target queries and a specific document positional encoding scheme. Faster DAN reaches competitive results compared to standard DAN, while being at least 4 times faster on whole single-page and double-page images of the RIMES 2009, READ 2016 and MAURDOR datasets. Source code and trained model weights are available at https://github.com/FactoDeepLearning/FasterDAN. △ Less

Submitted 25 January, 2023; originally announced January 2023.

Journal ref: International Conference on Document Analysis and Recognition - ICDAR 2023

Numerical evidences of a universal critical behavior of 2D and 3D random quantum clock and Potts models

Authors: Valentin Anfray, Christophe Chatelain

Abstract: The random quantum $q$-state clock and Potts models are studied in 2 and 3 dimensions. The existence of Griffiths phases is tested in the 2D case with $q=6$ by sampling the integrated probability distribution of local susceptibilities of the equivalent McCoy-Wu 3D classical modelswith Monte Carlo simulations. No Griffiths phase is found for the clock model. In contrast, numerical evidences of the… ▽ More The random quantum $q$-state clock and Potts models are studied in 2 and 3 dimensions. The existence of Griffiths phases is tested in the 2D case with $q=6$ by sampling the integrated probability distribution of local susceptibilities of the equivalent McCoy-Wu 3D classical modelswith Monte Carlo simulations. No Griffiths phase is found for the clock model. In contrast, numerical evidences of the existence of Griffiths phases in the random Potts model are given and the Finite Size effects are analyzed. The critical point of the random quantum clock model is then studied by Strong-Disorder Renormalization Group. Despite a chaotic behavior of the Renormalization-Group flow at weak disorder, evidences are given that this critical behavior is governed by the same Infinite-Disorder Fixed Point as the Potts model, independently from the number of states $q$. △ Less

Submitted 30 December, 2022; originally announced December 2022.

Journal ref: Phys. Rev. E 108, 014124 (2023)

Domain Translation via Latent Space Map**

Authors: Tsiry Mayet, Simon Bernard, Clement Chatelain, Romain Herault

Abstract: In this paper, we investigate the problem of multi-domain translation: given an element $a$ of domain $A$, we would like to generate a corresponding $b$ sample in another domain $B$, and vice versa. Acquiring supervision in multiple domains can be a tedious task, also we propose to learn this translation from one domain to another when supervision is available as a pair $(a,b)\sim A\times B$ and l… ▽ More In this paper, we investigate the problem of multi-domain translation: given an element $a$ of domain $A$, we would like to generate a corresponding $b$ sample in another domain $B$, and vice versa. Acquiring supervision in multiple domains can be a tedious task, also we propose to learn this translation from one domain to another when supervision is available as a pair $(a,b)\sim A\times B$ and leveraging possible unpaired data when only $a\sim A$ or only $b\sim B$ is available. We introduce a new unified framework called Latent Space Map** (\model) that exploits the manifold assumption in order to learn, from each domain, a latent space. Unlike existing approaches, we propose to further regularize each latent space using available domains by learning each dependency between pairs of domains. We evaluate our approach in three tasks performing i) synthetic dataset with image translation, ii) real-world task of semantic segmentation for medical images, and iii) real-world task of facial landmark detection. △ Less

Submitted 6 December, 2022; originally announced December 2022.

Finite-Size Scaling of the majority-voter model above the upper critical dimension

Authors: Christophe Chatelain

Abstract: The majority-voter model is studied by Monte Carlo simulations on hypercubic lattices of dimension $d=2$ to 7 with periodic boundary conditions. The critical exponents associated to the Finite-Size Scaling of the magnetic susceptibility are shown to be compatible with those of the Ising model. At dimension $d=4$, the numerical data are compatible with the presence of multiplicative logarithmic co… ▽ More The majority-voter model is studied by Monte Carlo simulations on hypercubic lattices of dimension $d=2$ to 7 with periodic boundary conditions. The critical exponents associated to the Finite-Size Scaling of the magnetic susceptibility are shown to be compatible with those of the Ising model. At dimension $d=4$, the numerical data are compatible with the presence of multiplicative logarithmic corrections. For $d\ge 5$,the estimates of the exponents are close to the prediction $d/2$ when taking into account the dangerous irrelevant variable at theGaussian fixed point. Moreover, the universal values of the Binder cumulant are also compatible with those of the Ising model. This indicates that the upper critical dimension of the majority-voter model is not $d_c=6$ as claimed in the literature, but $d_c=4$ like the equilibrium Ising model. △ Less

Submitted 6 January, 2023; v1 submitted 2 November, 2022; originally announced November 2022.

Comments: To be published in Condensed Matter Physics

DAN: a Segmentation-free Document Attention Network for Handwritten Document Recognition

Authors: Denis Coquenet, Clément Chatelain, Thierry Paquet

Abstract: Unconstrained handwritten text recognition is a challenging computer vision task. It is traditionally handled by a two-step approach, combining line segmentation followed by text line recognition. For the first time, we propose an end-to-end segmentation-free architecture for the task of handwritten document recognition: the Document Attention Network. In addition to text recognition, the model is… ▽ More Unconstrained handwritten text recognition is a challenging computer vision task. It is traditionally handled by a two-step approach, combining line segmentation followed by text line recognition. For the first time, we propose an end-to-end segmentation-free architecture for the task of handwritten document recognition: the Document Attention Network. In addition to text recognition, the model is trained to label text parts using begin and end tags in an XML-like fashion. This model is made up of an FCN encoder for feature extraction and a stack of transformer decoder layers for a recurrent token-by-token prediction process. It takes whole text documents as input and sequentially outputs characters, as well as logical layout tokens. Contrary to the existing segmentation-based approaches, the model is trained without using any segmentation label. We achieve competitive results on the READ 2016 dataset at page level, as well as double-page level with a CER of 3.43% and 3.70%, respectively. We also provide results for the RIMES 2009 dataset at page level, reaching 4.54% of CER. We provide all source code and pre-trained model weights at https://github.com/FactoDeepLearning/DAN. △ Less

Submitted 13 December, 2022; v1 submitted 23 March, 2022; originally announced March 2022.

Journal ref: IEEE Transactions on Pattern Analysis and Machine Intelligence 2023

Numerical evidence of a super-universality of the 2D and 3D random quantum Potts models

Authors: Valentin Anfray, Christophe Chatelain

Abstract: The random q-state quantum Potts model is studied on hypercubic lattices in dimensions 2 and 3 using the numerical implementation of the Strong Disorder Renormalization Group introduced by Kovacs and Igl{ó}i [Phys. Rev. B 82, 054437 (2010)]. Critical exponents $ν$, d f and $ψ$ at the Infinite Disorder Fixed Point are estimated by Finite-Size Scaling for several numbers of states q between 2 and 50… ▽ More The random q-state quantum Potts model is studied on hypercubic lattices in dimensions 2 and 3 using the numerical implementation of the Strong Disorder Renormalization Group introduced by Kovacs and Igl{ó}i [Phys. Rev. B 82, 054437 (2010)]. Critical exponents $ν$, d f and $ψ$ at the Infinite Disorder Fixed Point are estimated by Finite-Size Scaling for several numbers of states q between 2 and 50. When scaling corrections are not taken into account, the estimates of both d f and $ψ$ systematically increase with q. It is shown however that q-dependent scaling corrections are present and that the exponents are compatible within error bars, or close to each other, when these corrections are taking into account. This provides evidence of the existence of a super-universality of all 2D and 3D random Potts models. △ Less

Submitted 6 April, 2021; originally announced April 2021.

Journal ref: Phys. Rev. B 103, 174207 (2021)

SPAN: a Simple Predict & Align Network for Handwritten Paragraph Recognition

Authors: Denis Coquenet, Clément Chatelain, Thierry Paquet

Abstract: Unconstrained handwriting recognition is an essential task in document analysis. It is usually carried out in two steps. First, the document is segmented into text lines. Second, an Optical Character Recognition model is applied on these line images. We propose the Simple Predict & Align Network: an end-to-end recurrence-free Fully Convolutional Network performing OCR at paragraph level without an… ▽ More Unconstrained handwriting recognition is an essential task in document analysis. It is usually carried out in two steps. First, the document is segmented into text lines. Second, an Optical Character Recognition model is applied on these line images. We propose the Simple Predict & Align Network: an end-to-end recurrence-free Fully Convolutional Network performing OCR at paragraph level without any prior segmentation stage. The framework is as simple as the one used for the recognition of isolated lines and we achieve competitive results on three popular datasets: RIMES, IAM and READ 2016. The proposed model does not require any dataset adaptation, it can be trained from scratch, without segmentation labels, and it does not require line breaks in the transcription labels. Our code and trained model weights are available at https://github.com/FactoDeepLearning/SPAN. △ Less

Submitted 17 February, 2021; originally announced February 2021.

Journal ref: Document Analysis and Recognition - ICDAR 2021. ICDAR 2021. Lecture Notes in Computer Science, vol 12823

Recurrence-free unconstrained handwritten text recognition using gated fully convolutional network

Authors: Denis Coquenet, Clément Chatelain, Thierry Paquet

Abstract: Unconstrained handwritten text recognition is a major step in most document analysis tasks. This is generally processed by deep recurrent neural networks and more specifically with the use of Long Short-Term Memory cells. The main drawbacks of these components are the large number of parameters involved and their sequential execution during training and prediction. One alternative solution to usin… ▽ More Unconstrained handwritten text recognition is a major step in most document analysis tasks. This is generally processed by deep recurrent neural networks and more specifically with the use of Long Short-Term Memory cells. The main drawbacks of these components are the large number of parameters involved and their sequential execution during training and prediction. One alternative solution to using LSTM cells is to compensate the long time memory loss with an heavy use of convolutional layers whose operations can be executed in parallel and which imply fewer parameters. In this paper we present a Gated Fully Convolutional Network architecture that is a recurrence-free alternative to the well-known CNN+LSTM architectures. Our model is trained with the CTC loss and shows competitive results on both the RIMES and IAM datasets. We release all code to enable reproduction of our experiments: https://github.com/FactoDeepLearning/LinePytorchOCR. △ Less

Submitted 9 December, 2020; originally announced December 2020.

Journal ref: 2020 17th International Conference on Frontiers in Handwriting Recognition (ICFHR)

Have convolutions already made recurrence obsolete for unconstrained handwritten text recognition ?

Authors: Denis Coquenet, Yann Soullard, Clément Chatelain, Thierry Paquet

Abstract: Unconstrained handwritten text recognition remains an important challenge for deep neural networks. These last years, recurrent networks and more specifically Long Short-Term Memory networks have achieved state-of-the-art performance in this field. Nevertheless, they are made of a large number of trainable parameters and training recurrent neural networks does not support parallelism. This has a d… ▽ More Unconstrained handwritten text recognition remains an important challenge for deep neural networks. These last years, recurrent networks and more specifically Long Short-Term Memory networks have achieved state-of-the-art performance in this field. Nevertheless, they are made of a large number of trainable parameters and training recurrent neural networks does not support parallelism. This has a direct influence on the training time of such architectures, with also a direct consequence on the time required to explore various architectures. Recently, recurrence-free architectures such as Fully Convolutional Networks with gated mechanisms have been proposed as one possible alternative achieving competitive results. In this paper, we explore convolutional architectures and compare them to a CNN+BLSTM baseline. We propose an experimental study regarding different architectures on an offline handwriting recognition task using the RIMES dataset, and a modified version of it that consists of augmenting the images with notebook backgrounds that are printed grids. △ Less

Submitted 9 December, 2020; originally announced December 2020.

Journal ref: 2019 International Conference on Document Analysis and Recognition Workshops (ICDARW)

End-to-end Handwritten Paragraph Text Recognition Using a Vertical Attention Network

Authors: Denis Coquenet, Clément Chatelain, Thierry Paquet

Abstract: Unconstrained handwritten text recognition remains challenging for computer vision systems. Paragraph text recognition is traditionally achieved by two models: the first one for line segmentation and the second one for text line recognition. We propose a unified end-to-end model using hybrid attention to tackle this task. This model is designed to iteratively process a paragraph image line by line… ▽ More Unconstrained handwritten text recognition remains challenging for computer vision systems. Paragraph text recognition is traditionally achieved by two models: the first one for line segmentation and the second one for text line recognition. We propose a unified end-to-end model using hybrid attention to tackle this task. This model is designed to iteratively process a paragraph image line by line. It can be split into three modules. An encoder generates feature maps from the whole paragraph image. Then, an attention module recurrently generates a vertical weighted mask enabling to focus on the current text line features. This way, it performs a kind of implicit line segmentation. For each text line features, a decoder module recognizes the character sequence associated, leading to the recognition of a whole paragraph. We achieve state-of-the-art character error rate at paragraph level on three popular datasets: 1.91% for RIMES, 4.45% for IAM and 3.59% for READ 2016. Our code and trained model weights are available at https://github.com/FactoDeepLearning/VerticalAttentionOCR. △ Less

Submitted 3 December, 2021; v1 submitted 7 December, 2020; originally announced December 2020.

Journal ref: IEEE Transactions on Pattern Analysis and Machine Intelligence 2022

Object Detection in the DCT Domain: is Luminance the Solution?

Authors: Benjamin Deguerre, Clement Chatelain, Gilles Gasso

Abstract: Object detection in images has reached unprecedented performances. The state-of-the-art methods rely on deep architectures that extract salient features and predict bounding boxes enclosing the objects of interest. These methods essentially run on RGB images. However, the RGB images are often compressed by the acquisition devices for storage purpose and transfer efficiency. Hence, their decompress… ▽ More Object detection in images has reached unprecedented performances. The state-of-the-art methods rely on deep architectures that extract salient features and predict bounding boxes enclosing the objects of interest. These methods essentially run on RGB images. However, the RGB images are often compressed by the acquisition devices for storage purpose and transfer efficiency. Hence, their decompression is required for object detectors. To gain in efficiency, this paper proposes to take advantage of the compressed representation of images to carry out object detection usable in constrained resources conditions. Specifically, we focus on JPEG images and propose a thorough analysis of detection architectures newly designed in regard of the peculiarities of the JPEG norm. This leads to a $\times 1.7$ speed up in comparison with a standard RGB-based architecture, while only reducing the detection performance by 5.5%. Additionally, our empirical findings demonstrate that only part of the compressed JPEG information, namely the luminance component, may be required to match detection accuracy of the full input methods. △ Less

Submitted 14 July, 2021; v1 submitted 10 June, 2020; originally announced June 2020.

Comments: ICPR 2020

Absence of logarithmic divergence of the entanglement entropies at the phase transitions of a 2D classical hard rod model

Authors: Christophe Chatelain, Andrej Gendiar

Abstract: Entanglement entropy is a powerful tool to detect continuous, discontinuous and even topological phase transitions in quantum as well as classical systems. In this work, von Neumann and Renyi entanglement entropies are studied numerically for classical lattice models in a square geometry. A cut is made from the center of the square to the midpoint of one of its edges, say the right edge. The entan… ▽ More Entanglement entropy is a powerful tool to detect continuous, discontinuous and even topological phase transitions in quantum as well as classical systems. In this work, von Neumann and Renyi entanglement entropies are studied numerically for classical lattice models in a square geometry. A cut is made from the center of the square to the midpoint of one of its edges, say the right edge. The entanglement entropies measure the entanglement between the left and right halves of the system. As in the strip geometry, von Neumann and Renyi entanglement entropies diverge logarithmically at the transition point while they display a jump for first-order phase transitions. The analysis is extended to a classical model of non-overlap** finite hard rods deposited on a square lattice for which Monte Carlo simulations have shown that, when the hard rods span over 7 or more lattice sites, a nematic phase appears in the phase diagram between two disordered phases. A new Corner Transfer Matrix Renormalization Group algorithm (CTMRG) is introduced to study this model. No logarithmic divergence of entanglement entropies is observed at the phase transitions in the CTMRG calculation discussed here. We therefore infer that the transitions neither can belong to the Ising universality class, as previously assumed in the literature, nor be discontinuous. △ Less

Submitted 28 May, 2020; v1 submitted 3 February, 2020; originally announced February 2020.

Journal ref: Eur. Phys. J. B 93, 134 (2020)

Fast object detection in compressed JPEG Images

Authors: Benjamin Deguerre, Clément Chatelain, Gilles Gasso

Abstract: Object detection in still images has drawn a lot of attention over past few years, and with the advent of Deep Learning impressive performances have been achieved with numerous industrial applications. Most of these deep learning models rely on RGB images to localize and identify objects in the image. However in some application scenarii, images are compressed either for storage savings or fast tr… ▽ More Object detection in still images has drawn a lot of attention over past few years, and with the advent of Deep Learning impressive performances have been achieved with numerous industrial applications. Most of these deep learning models rely on RGB images to localize and identify objects in the image. However in some application scenarii, images are compressed either for storage savings or fast transmission. Therefore a time consuming image decompression step is compulsory in order to apply the aforementioned deep models. To alleviate this drawback, we propose a fast deep architecture for object detection in JPEG images, one of the most widespread compression format. We train a neural network to detect objects based on the blockwise DCT (discrete cosine transform) coefficients {issued from} the JPEG compression algorithm. We modify the well-known Single Shot multibox Detector (SSD) by replacing its first layers with one convolutional layer dedicated to process the DCT inputs. Experimental evaluations on PASCAL VOC and industrial dataset comprising images of road traffic surveillance show that the model is about $2\times$ faster than regular SSD with promising detection performances. To the best of our knowledge, this paper is the first to address detection in compressed JPEG images. △ Less

Submitted 16 April, 2019; originally announced April 2019.

Improved Matrix Product Operator Renormalization Group: application to the N-color random Ashkin-Teller chain

Authors: Christophe Chatelain

Abstract: Strong-Disorder Renormalization Group (SDRG), despite being a relatively simple real-space renormalization procedure, provides in principle exact results on the critical properties at the infinite-randomness fixed point of random quantum spin chains. Numerically, SDRG can be efficiently implemented as a renormalization of Matrix Product Operators (MPO-RG). By considering larger blocks than SDRG, M… ▽ More Strong-Disorder Renormalization Group (SDRG), despite being a relatively simple real-space renormalization procedure, provides in principle exact results on the critical properties at the infinite-randomness fixed point of random quantum spin chains. Numerically, SDRG can be efficiently implemented as a renormalization of Matrix Product Operators (MPO-RG). By considering larger blocks than SDRG, MPO-RG was recently used to compute non-critical quantities of finite chains that are inaccessible to SDRG. In this work, the accuracy of this approach is studied and two simple and fast improvements are proposed. The accuracy on the ground state energy is improved by a factor at least equal to 4 for the random Ising chain in a transverse field. Finally, the proposed algorithms are shown to yield Binder cumulants of the 3-color random Ashkin-Teller chain that are compatible with a second-order phase transition while a first-order one is predicted by the original MPO-RG algorithm. △ Less

Submitted 24 June, 2019; v1 submitted 24 July, 2018; originally announced July 2018.

Neural Networks Regularization Through Class-wise Invariant Representation Learning

Authors: Soufiane Belharbi, Clément Chatelain, Romain Hérault, Sébastien Adam

Abstract: Training deep neural networks is known to require a large number of training samples. However, in many applications only few training samples are available. In this work, we tackle the issue of training neural networks for classification task when few training samples are available. We attempt to solve this issue by proposing a new regularization term that constrains the hidden layers of a network… ▽ More Training deep neural networks is known to require a large number of training samples. However, in many applications only few training samples are available. In this work, we tackle the issue of training neural networks for classification task when few training samples are available. We attempt to solve this issue by proposing a new regularization term that constrains the hidden layers of a network to learn class-wise invariant representations. In our regularization framework, learning invariant representations is generalized to the class membership where samples with the same class should have the same representation. Numerical experiments over MNIST and its variants showed that our proposal helps improving the generalization of neural network particularly when trained with few samples. We provide the source code of our framework https://github.com/sbelharbi/learning-class-invariant-features . △ Less

Submitted 22 December, 2017; v1 submitted 6 September, 2017; originally announced September 2017.

Comments: Submitted to ELSEVIER, 13 pages, 5 figures

LV-ROVER: Lexicon Verified Recognizer Output Voting Error Reduction

Authors: Bruno Stuner, Clément Chatelain, Thierry Paquet

Abstract: Offline handwritten text line recognition is a hard task that requires both an efficient optical character recognizer and language model. Handwriting recognition state of the art methods are based on Long Short Term Memory (LSTM) recurrent neural networks (RNN) coupled with the use of linguistic knowledge. Most of the proposed approaches in the literature focus on improving one of the two componen… ▽ More Offline handwritten text line recognition is a hard task that requires both an efficient optical character recognizer and language model. Handwriting recognition state of the art methods are based on Long Short Term Memory (LSTM) recurrent neural networks (RNN) coupled with the use of linguistic knowledge. Most of the proposed approaches in the literature focus on improving one of the two components and use constraint, dedicated to a database lexicon. However, state of the art performance is achieved by combining multiple optical models, and possibly multiple language models with the Recognizer Output Voting Error Reduction (ROVER) framework. Though handwritten line recognition with ROVER has been implemented by combining only few recognizers because training multiple complete recognizers is hard. In this paper we propose a Lexicon Verified ROVER: LV-ROVER, that has a reduce complexity compare to the original one and that can combine hundreds of recognizers without language models. We achieve state of the art for handwritten line text on the RIMES dataset. △ Less

Submitted 24 July, 2017; originally announced July 2017.

Comments: Submitted to Pattern Recognition Letters

Diverging conductance at the contact between random and pure quantum XX spin chains

Authors: Christophe Chatelain

Abstract: A model consisting in two quantum XX spin chains, one homogeneous and the second with random couplings drawn from a binary distribution, is considered. The two chains are coupled to two different non-local thermal baths and their dynamics is governed by a Lindblad equation. In the steady state, a current J is induced between the two chains by coupling them together by their edges and imposing diff… ▽ More A model consisting in two quantum XX spin chains, one homogeneous and the second with random couplings drawn from a binary distribution, is considered. The two chains are coupled to two different non-local thermal baths and their dynamics is governed by a Lindblad equation. In the steady state, a current J is induced between the two chains by coupling them together by their edges and imposing different chemical potentials $μ$ to the two baths. While a regime of linear characteristics J versus $Δ$$μ$ is observed in the absence of randomness, a gap opens as the disorder strength is increased. In the infinite-randomness limit, this behavior is related to the density of states of the localized states contributing to the current. The conductance is shown to diverge in this limit. △ Less

Submitted 7 September, 2017; v1 submitted 11 July, 2017; originally announced July 2017.

Comments: 15 pages, 18 figures

Handwriting recognition using Cohort of LSTM and lexicon verification with extremely large lexicon

Authors: Bruno Stuner, Clément Chatelain, Thierry Paquet

Abstract: State-of-the-art methods for handwriting recognition are based on Long Short Term Memory (LSTM) recurrent neural networks (RNN), which now provides very impressive character recognition performance. The character recognition is generally coupled with a lexicon driven decoding process which integrates dictionaries. Unfortunately these dictionaries are limited to hundred of thousands words for the b… ▽ More State-of-the-art methods for handwriting recognition are based on Long Short Term Memory (LSTM) recurrent neural networks (RNN), which now provides very impressive character recognition performance. The character recognition is generally coupled with a lexicon driven decoding process which integrates dictionaries. Unfortunately these dictionaries are limited to hundred of thousands words for the best systems, which prevent from having a good language coverage, and therefore limit the global recognition performance. In this article, we propose an alternative to the lexicon driven decoding process based on a lexicon verification process, coupled with an original cascade architecture. The cascade is made of a large number of complementary networks extracted from a single training (called cohort), making the learning process very light. The proposed method achieves new state-of-the art word recognition performance on the Rimes and IAM databases. Dealing with gigantic lexicon of 3 millions words, the methods also demonstrates interesting performance with a fast decision stage. △ Less

Submitted 25 September, 2017; v1 submitted 22 December, 2016; originally announced December 2016.

Comments: 31 pages, paper submitted to Pattern Recognition

Infinite disorder and correlation fixed point in the Potts model with correlated disorder

Authors: Christophe Chatelain

Abstract: Recent Monte Carlo simulations of the q-state Potts model with a disorder displaying slowly-decaying correlations reported a violation of hyperscaling relation caused by large disorder fluctuations and the existence of a Griffiths phase, as in random systems governed by an infinite-disorder fixed point. New simulations, directly made in the limit of an infinite disorder strength, are presented. Th… ▽ More Recent Monte Carlo simulations of the q-state Potts model with a disorder displaying slowly-decaying correlations reported a violation of hyperscaling relation caused by large disorder fluctuations and the existence of a Griffiths phase, as in random systems governed by an infinite-disorder fixed point. New simulations, directly made in the limit of an infinite disorder strength, are presented. The magnetic scaling dimension is shown to correspond to the correlated percola-tion fixed point. The latter is shown to be unstable at finite disorder strength but with a large cross-over length which is not accessible to Monte Carlo simulations. △ Less

Submitted 18 October, 2016; originally announced October 2016.

Numerical study of the critical phases of the frustrated Z(5) model

Authors: Christophe Chatelain

Abstract: The phase diagram of the $\mathbb{Z}(5)$ spin model is studied numerically on the square lattice by means of the {\sl Density Matrix Renormalization Group}. In the regime where the two nearest-neighbor couplings have opposite signs, a critical phase, not observed in earlier Monte Carlo simulations, is identified. The new phase diagram is in agreement with predictions made by M. den Nijs [{\sl Phys… ▽ More The phase diagram of the $\mathbb{Z}(5)$ spin model is studied numerically on the square lattice by means of the {\sl Density Matrix Renormalization Group}. In the regime where the two nearest-neighbor couplings have opposite signs, a critical phase, not observed in earlier Monte Carlo simulations, is identified. The new phase diagram is in agreement with predictions made by M. den Nijs [{\sl Phys. Rev. B} {\bf 31}, 266 (1985)] but for the $\mathbb{Z}(7)$ model rather than the $\mathbb{Z}(5)$ one. All critical phases are shown to be compatible with a central charge $c=1$. The magnetization scaling dimension displays however a different behavior in the different critical phases. △ Less

Submitted 3 June, 2016; v1 submitted 25 February, 2016; originally announced February 2016.

Comments: 20 pages

Numerical evidence of the double-Griffiths phase of the random quantum Ashkin-Teller chain

Authors: Christophe Chatelain, Dimitrios Voliotis

Abstract: The random quantum Ashkin-Teller chain is studied numerically by means of time-dependent Density-Matrix Renormalization Group. The critical lines are estimated as the location of the peaks of the integrated autocorrelation times, computed from spin-spin and polarization-polarization autocorrelation functions. Disorder fluctuations of magnetization and polarization are observed to be maximum on the… ▽ More The random quantum Ashkin-Teller chain is studied numerically by means of time-dependent Density-Matrix Renormalization Group. The critical lines are estimated as the location of the peaks of the integrated autocorrelation times, computed from spin-spin and polarization-polarization autocorrelation functions. Disorder fluctuations of magnetization and polarization are observed to be maximum on these critical lines. Entanglement entropy leads to the same phase diagram, though with larger Finite-Size effects. The decay of spin-spin and polarization-polarization autocorrelation functions provides numerical evidence of the existence of a double Griffiths phase when taking into account finite-size effects. The two associated dynamical exponents z increase rapidly as the critical lines are approached, in agreement with the recent conjecture of a divergence at the two transitions in the thermodynamic limit. △ Less

Submitted 26 January, 2016; v1 submitted 1 June, 2015; originally announced June 2015.

Journal ref: The European Physical Journal B, Springer, 2016, 89, pp.18

Deep Neural Networks Regularization for Structured Output Prediction

Authors: Soufiane Belharbi, Romain Hérault, Clément Chatelain, Sébastien Adam

Abstract: A deep neural network model is a powerful framework for learning representations. Usually, it is used to learn the relation $x \to y$ by exploiting the regularities in the input $x$. In structured output prediction problems, $y$ is multi-dimensional and structural relations often exist between the dimensions. The motivation of this work is to learn the output dependencies that may lie in the outpu… ▽ More A deep neural network model is a powerful framework for learning representations. Usually, it is used to learn the relation $x \to y$ by exploiting the regularities in the input $x$. In structured output prediction problems, $y$ is multi-dimensional and structural relations often exist between the dimensions. The motivation of this work is to learn the output dependencies that may lie in the output data in order to improve the prediction accuracy. Unfortunately, feedforward networks are unable to exploit the relations between the outputs. In order to overcome this issue, we propose in this paper a regularization scheme for training neural networks for these particular tasks using a multi-task framework. Our scheme aims at incorporating the learning of the output representation $y$ in the training process in an unsupervised fashion while learning the supervised map** function $x \to y$. We evaluate our framework on a facial landmark detection problem which is a typical structured output task. We show over two public challenging datasets (LFPW and HELEN) that our regularization scheme improves the generalization of deep neural networks and accelerates their training. The use of unlabeled data and label-only data is also explored, showing an additional improvement of the results. We provide an opensource implementation (https://github.com/sbelharbi/structured-output-ae) of our framework. △ Less

Submitted 30 October, 2017; v1 submitted 28 April, 2015; originally announced April 2015.

Comments: Submitted to Neurocomputing, 8 figures

Critical behavior in lattice models with two symmetric absorbing state

Authors: Áttila L. Rodrigues, Christophe Chatelain, Tânia Tomé, Mário J. De Oliveira

Abstract: We analyze nonequilibrium lattice models with up-down symmetry and two absorbing states by mean-field approximations and numerical simulations in two and three dimensions. The phase diagram displays three phases: paramagnetic, ferromagnetic and absorbing. The transition line between the first two phases belongs to the Ising universality class and between the last two, to the direct percolation uni… ▽ More We analyze nonequilibrium lattice models with up-down symmetry and two absorbing states by mean-field approximations and numerical simulations in two and three dimensions. The phase diagram displays three phases: paramagnetic, ferromagnetic and absorbing. The transition line between the first two phases belongs to the Ising universality class and between the last two, to the direct percolation universality class. The two lines meet at the point describing the voter model and the size $\ell$ of the ferromagnetic phase vanishes with the distance $\varepsilon$ to the voter point as $\ell\sim\varepsilon$, with possible logarithm corrections in two dimensions. △ Less

Submitted 6 January, 2015; originally announced January 2015.

DMRG study of the Berezinskii-Kosterlitz-Thouless transitions of the 2D five-state clock model

Authors: Christophe Chatelain

Abstract: The two Berezinskii-Kosterlitz-Thouless phase transitions of the two-dimensional 5-state clock model are studied on infinite strips using the DMRG algorithm. Because of the open boundary conditions, the helicity modulus $Υ_2$ is computed by imposing twisted magnetic fields at the two boundaries. Its scaling behavior is in good agreement with the existence of essential singularities with $σ=1/2$ at… ▽ More The two Berezinskii-Kosterlitz-Thouless phase transitions of the two-dimensional 5-state clock model are studied on infinite strips using the DMRG algorithm. Because of the open boundary conditions, the helicity modulus $Υ_2$ is computed by imposing twisted magnetic fields at the two boundaries. Its scaling behavior is in good agreement with the existence of essential singularities with $σ=1/2$ at the two transitions. The predicted universal values of $Υ_2$ are shown to be reached in the thermodynamic limit. The fourth-order helicity modulus is observed to display a dip at the high-temperature BKT transition, like the XY model, and shown to take a new universal value at the low-temperature one. Finally, the scaling behavior of magnetization at the low-temperature transition is compatible with $η=1/4$. △ Less

Submitted 12 October, 2014; v1 submitted 22 July, 2014; originally announced July 2014.

Comments: 12 pages, 13 figures

Stability of the Griffiths phase in the 2D Potts model with correlated disorder

Authors: Christophe Chatelain

Abstract: A Griffiths phase has recently been observed by Monte Carlo simulations in the 2D $q$-state Potts model with strongly correlated quenched random couplings. In particular, the magnetic susceptibility was shown to diverge algebraically with the lattice size in a broad range of temperatures. However, only relatively small lattice sizes could be considered so one can wonder whether this Griffiths phas… ▽ More A Griffiths phase has recently been observed by Monte Carlo simulations in the 2D $q$-state Potts model with strongly correlated quenched random couplings. In particular, the magnetic susceptibility was shown to diverge algebraically with the lattice size in a broad range of temperatures. However, only relatively small lattice sizes could be considered so one can wonder whether this Griffiths phase will not shrink and collapse into a single point, the critical point, as the lattice size is increased to much larger values. In this paper, the 2D eight-state Potts model is numerically studied for four different disorder correlations. It is shown that the Griffiths phase cannot be explained as a simple spreading of local transition temperatures caused by disorder fluctuations. As a consequence, the vanishing of the latter in the thermodynamic limit does not necessarily imply the collapse of the Griffiths phase into a single point. In contrast, the width of the Griffiths phase is controlled by the disorder strength. However, for disorder correlations decaying slower than $1/r$, no cross-over to a more usual critical behavior could be observed as this strength is tuned to weaker values. △ Less

Submitted 25 April, 2014; originally announced April 2014.

Comments: 10 pages, 16 figures

Journal ref: Condensed Matter Physics 17, 3 (2014) 33601

Griffiths phase and critical behavior of the 2D Potts models with long-range correlated disorder

Authors: Christophe Chatelain

Abstract: The $q$-state Potts model with a long-range correlated disorder is studied by means of large-scale Monte Carlo simulations for $q=2,4,8$ and 16. Evidence is given of the existence of a Griffiths phase, where the thermodynamic quantities display an algebraic Finite-Size Scaling, in a finite range of temperatures around the self-dual point. The critical exponents are shown to depend on both the temp… ▽ More The $q$-state Potts model with a long-range correlated disorder is studied by means of large-scale Monte Carlo simulations for $q=2,4,8$ and 16. Evidence is given of the existence of a Griffiths phase, where the thermodynamic quantities display an algebraic Finite-Size Scaling, in a finite range of temperatures around the self-dual point. The critical exponents are shown to depend on both the temperature and the exponent of the algebraic decay of disorder correlations, but not on the number of states of the Potts model. The mechanism leading to the violation of hyperscaling relations is observed in the entire Griffiths phase. △ Less

Submitted 23 February, 2014; v1 submitted 3 August, 2013; originally announced August 2013.

Comments: 18 pages, 16 figures, 12 tables

Hyperscaling violation in the 2D 8-state Potts model with long-range correlated disorder

Authors: Christophe Chatelain

Abstract: The first-order phase transition of the two-dimensional eight-state Potts model is shown to be rounded when long-range correlated disorder is coupled to energy density. Critical exponents are estimated by means of large-scale Monte Carlo simulations. In contrast to uncorrelated disorder, a violation of the hyperscaling relation $γ/ν=d-2x_σ$ is observed. Even though the system is not frustrated, di… ▽ More The first-order phase transition of the two-dimensional eight-state Potts model is shown to be rounded when long-range correlated disorder is coupled to energy density. Critical exponents are estimated by means of large-scale Monte Carlo simulations. In contrast to uncorrelated disorder, a violation of the hyperscaling relation $γ/ν=d-2x_σ$ is observed. Even though the system is not frustrated, disorder fluctuations are strong enough to cause this violation in the very same way as in the 3D random-field Ising model. In the thermal sector too, evidence is given for such violation in the two hyperscaling relations $α/ν=d-2x_\varepsilon$ and $1/ν=d-x_\varepsilon$. In contrast to the random field Ising model, at least two hyperscaling violation exponents are needed. The scaling dimension of energy is conjectured to be $x_\varepsilon=a/2$, where $a$ is the exponent of the algebraic decay of disorder correlations. △ Less

Submitted 12 July, 2013; v1 submitted 8 March, 2013; originally announced March 2013.

Comments: 6 pages, 6 figures. Published version

Journal ref: EPL 102 (2013) 66007

Exact correlation functions in particle-reaction models with immobile particles

Authors: Christophe Chatelain, Malte Henkel, Mário J. De Oliveira, Tânia Tomé

Abstract: Exact results on particle-densities as well as correlators in two models of immobile particles, containing either a single species or else two distinct species, are derived. The models evolve following a descent dynamics through pair-annihilation where each particle interacts at most once throughout its entire history. The resulting large number of stationary states leads to a non-vanishing config… ▽ More Exact results on particle-densities as well as correlators in two models of immobile particles, containing either a single species or else two distinct species, are derived. The models evolve following a descent dynamics through pair-annihilation where each particle interacts at most once throughout its entire history. The resulting large number of stationary states leads to a non-vanishing configurational entropy. Our results are established for arbitrary initial conditions and are derived via a generating-function method. The single-species model is the dual of the 1D zero-temperature kinetic Ising model with Kimball-Deker-Haake dynamics. In this way, both infinite and semi-infinite chains and also the Bethe lattice can be analysed. The relationship with the random sequential adsorption of dimers and weakly tapped granular materials is discussed. △ Less

Submitted 21 September, 2012; v1 submitted 10 July, 2012; originally announced July 2012.

Comments: 43 pages, 18 figures

Journal ref: J. Stat. Mech. (2012) P11006

Relaxation at finite temperature in Fully-Frustrated Ising Models

Authors: Jean-Charles Walter, Christophe Chatelain

Abstract: We consider by means of Monte Carlo simulations the relaxation in the paramagnetic phase of the anti-ferromagnetic Ising model on a triangular lattice and of a fully-frustrated Ising model on a square lattice. In contradistinction to previous studies of the second model, we show that spin-spin correlation functions do not decay with a stretched-exponential law at low temperature but that both mode… ▽ More We consider by means of Monte Carlo simulations the relaxation in the paramagnetic phase of the anti-ferromagnetic Ising model on a triangular lattice and of a fully-frustrated Ising model on a square lattice. In contradistinction to previous studies of the second model, we show that spin-spin correlation functions do not decay with a stretched-exponential law at low temperature but that both models display an exponential decay with logarithmic corrections that are interpreted as the signature of topological defects. △ Less

Submitted 20 December, 2011; originally announced December 2011.

Comments: 12 pages, 9 figures

Journal ref: Journal of Statistical Mechanics: Theory and Experiment (2012) P02010

Aging and stationary properties of non-equilibrium symmetrical three-state models

Authors: Christophe Chatelain, Tânia Tomé, Mario J. De Oliveira

Abstract: We consider a non-equilibrium three-state model whose dynamics is Markovian and displays the same symmetry as the three-state Potts model, i.e., the transition rates are invariant under the permutation of the states. Unlike the Potts model, detailed balance is in general not satisfied. The aging and the stationary properties of the model defined on a square lattice are obtained by means of large-s… ▽ More We consider a non-equilibrium three-state model whose dynamics is Markovian and displays the same symmetry as the three-state Potts model, i.e., the transition rates are invariant under the permutation of the states. Unlike the Potts model, detailed balance is in general not satisfied. The aging and the stationary properties of the model defined on a square lattice are obtained by means of large-scale Monte Carlo simulations. We show that the phase diagram presents a critical line, belonging to the three-state Potts universality class, that ends at a point whose universality class is that of the voter model. Aging is considered on the critical line, at the voter point and in the ferromagnetic phase. △ Less

Submitted 15 December, 2010; originally announced December 2010.

Comments: 25 pages

Numerical study of Schramm-Loewner Evolution in the random 3-state Potts model

Authors: Christophe Chatelain

Abstract: We have numerically studied the properties of the interface induced in the ferromagnetic random-bond three-state Potts model by symmetry-breaking boundary conditions. The fractal dimension $d_f$ of the interface was determined. The corresponding SLE parameter $κ$ was estimated to be $κ\simeq 3.18(6)$, compatible with previous estimate. On the other hand, we estimated $κ$ independently from the pro… ▽ More We have numerically studied the properties of the interface induced in the ferromagnetic random-bond three-state Potts model by symmetry-breaking boundary conditions. The fractal dimension $d_f$ of the interface was determined. The corresponding SLE parameter $κ$ was estimated to be $κ\simeq 3.18(6)$, compatible with previous estimate. On the other hand, we estimated $κ$ independently from the probability of passage of the interface at the left of a given point. The numerical data are well reproduced by the Schramm theoretical prediction and the fit leads to $κ\simeq 3.245(10)$, in agreement with the first estimate. This provides evidences that the geometric properties of spin interfaces in the random 3-state Potts model may be described by chordal ${\rm SLE}_κ$. △ Less

Submitted 18 June, 2010; originally announced June 2010.

Comments: 11 pages

Journal ref: Journal of Statistical Mechanics: Theory and Experiment (2010) P08004

Mesoscopic rings with Spin-Orbit interactions

Authors: Bertrand Berche, Christophe Chatelain, Ernesto Medina

Abstract: A didactic description of charge and spin equilibrium currents on mesoscopic rings in the presence of Spin-Orbit interaction is presented. Emphasis is made on the non trivial construction of the correct Hamiltonian in polar coordinates, the calculation of eigenvalues and eigenfunctions and the symmetries of the ground state properties. Spin currents are derived following an intuitive definition an… ▽ More A didactic description of charge and spin equilibrium currents on mesoscopic rings in the presence of Spin-Orbit interaction is presented. Emphasis is made on the non trivial construction of the correct Hamiltonian in polar coordinates, the calculation of eigenvalues and eigenfunctions and the symmetries of the ground state properties. Spin currents are derived following an intuitive definition and then a more thorough derivation is built upon the canonical Lagrangian formulation that emphasizes the SU(2) gauge structure of the transport problem of spin 1/2 fermions in spin-orbit active media. The quantization conditions that follow from the constraint of single-valued Pauli spinors are also discussed. The targeted students are those of a graduate Condensed Matter Physics course. △ Less

Submitted 13 July, 2010; v1 submitted 27 May, 2010; originally announced May 2010.

Journal ref: Eur. J. Phys.31:1267-1286, 2010

Symmetry relation for multifractal spectra at random critical points

Authors: Cecile Monthus, Bertrand Berche, Christophe Chatelain

Abstract: Random critical points are generically characterized by multifractal properties. In the field of Anderson localization, Mirlin, Fyodorov, Mildenberger and Evers [Phys. Rev. Lett 97, 046803 (2006)] have proposed that the singularity spectrum $f(α)$ of eigenfunctions satisfies the exact symmetry $f(2d-α)=f(α)+d-α$ at any Anderson transition. In the present paper, we analyse the physical origin of… ▽ More Random critical points are generically characterized by multifractal properties. In the field of Anderson localization, Mirlin, Fyodorov, Mildenberger and Evers [Phys. Rev. Lett 97, 046803 (2006)] have proposed that the singularity spectrum $f(α)$ of eigenfunctions satisfies the exact symmetry $f(2d-α)=f(α)+d-α$ at any Anderson transition. In the present paper, we analyse the physical origin of this symmetry in relation with the Gallavotti-Cohen fluctuation relations of large deviation functions that are well-known in the field of non-equilibrium dynamics: the multifractal spectrum of the disordered model corresponds to the large deviation function of the rescaling exponent $γ=(α-d)$ along a renormalization trajectory in the effective time $t=\ln L$. We conclude that the symmetry discovered on the specific example of Anderson transitions should actually be satisfied at many other random critical points after an appropriate translation. For many-body random phase transitions, where the critical properties are usually analyzed in terms of the multifractal spectrum $H(a)$ and of the moments exponents X(N) of two-point correlation function [A. Ludwig, Nucl. Phys. B330, 639 (1990)], the symmetry becomes $H(2X(1) -a)= H(a) + a-X(1)$, or equivalently $Δ(N)=Δ(1-N)$ for the anomalous parts $Δ(N) \equiv X(N)-NX(1)$. We present numerical tests in favor of this symmetry for the 2D random $Q-$state Potts model with various $Q$. △ Less

Submitted 6 November, 2009; v1 submitted 3 September, 2009; originally announced September 2009.

Comments: 15 pages, 3 figures, v2=final version

Journal ref: J. Stat. Mech. (2009) P12002

Numerical investigation of the aging of the Fully-Frustrated XY model

Authors: Jean-Charles Walter, Christophe Chatelain

Abstract: We study the out-of-equilibrium dynamics of the fully-frustrated XY model. At equilibrium, this model undergoes two phase transitions at two very close temperatures: a Kosterlitz-Thouless topological transition and a second-order phase transition between a paramagnetic phase and a low-temperature phase where the chiralities of the lattice plaquettes are anti-ferromagnetically ordered. We compute… ▽ More We study the out-of-equilibrium dynamics of the fully-frustrated XY model. At equilibrium, this model undergoes two phase transitions at two very close temperatures: a Kosterlitz-Thouless topological transition and a second-order phase transition between a paramagnetic phase and a low-temperature phase where the chiralities of the lattice plaquettes are anti-ferromagnetically ordered. We compute by Monte Carlo simulations two-time spin-spin and chirality-chirality autocorrelation and response functions. From the dynamics of the spin waves in the low temperature phase, we extract the temperature-dependent exponent $η$. We provide evidences for logarithmic corrections above the Kosterlitz-Thouless temperature and interpret them as a manifestation of free topological defects. Our estimates of the autocorrelation exponent and the fluctuation-dissipation ratio differ from the XY values, while $η(T_{\rm KT})$ lies at the boundary of the error bar. Indications for logarithmic corrections at the second-order critical temperature are presented. However, the coupling between angles and chiralities is still strong and explains why autocorrelation exponent and fluctuation-dissipation ratio are far from the Ising values and seems stable. △ Less

Submitted 19 October, 2009; v1 submitted 9 July, 2009; originally announced July 2009.

Comments: 16 pages, 11 figures

Journal ref: J. Stat. Mech. (2009) P10017

Extended Scaling in High Dimensions

Authors: Bertrand Berche, Christophe Chatelain, Chania Dhall, Ralph Kenna, Robert Low, Jean-Charles Walter

Abstract: We apply and test the recently proposed "extended scaling" scheme in an analysis of the magnetic susceptibility of Ising systems above the upper critical dimension. The data are obtained by Monte Carlo simulations using both the conventional Wolff cluster algorithm and the Prokof'ev-Svistunov worm algorithm. As already observed for other models, extended scaling is shown to extend the high-tempe… ▽ More We apply and test the recently proposed "extended scaling" scheme in an analysis of the magnetic susceptibility of Ising systems above the upper critical dimension. The data are obtained by Monte Carlo simulations using both the conventional Wolff cluster algorithm and the Prokof'ev-Svistunov worm algorithm. As already observed for other models, extended scaling is shown to extend the high-temperature critical scaling regime over a range of temperatures much wider than that achieved conventionally. It allows for an accurate determination of leading and sub-leading scaling indices, critical temperatures and amplitudes of the confluent corrections. △ Less

Submitted 24 October, 2008; v1 submitted 16 July, 2008; originally announced July 2008.

Comments: 16 pages, 8 figures. Improved version to appear in JSTAT

Journal ref: J.Stat. Mech. (2008) P11010

Probability distributions for polymer translocation

Authors: Clément Chatelain, Yacov Kantor, Mehran Kardar

Abstract: We study the passage (translocation) of a self-avoiding polymer through a membrane pore in two dimensions. In particular, we numerically measure the probability distribution Q(T) of the translocation time T, and the distribution P(s,t) of the translocation coordinate s at various times t. When scaled with the mean translocation time <T>, Q(T) becomes independent of polymer length, and decays exp… ▽ More We study the passage (translocation) of a self-avoiding polymer through a membrane pore in two dimensions. In particular, we numerically measure the probability distribution Q(T) of the translocation time T, and the distribution P(s,t) of the translocation coordinate s at various times t. When scaled with the mean translocation time <T>, Q(T) becomes independent of polymer length, and decays exponentially for large T. The probability P(s,t) is well described by a Gaussian at short times, with a variance that grows sub-diffusively as t^α with α~0.8. For times exceeding <T>, P(s,t) of the polymers that have not yet finished their translocation has a non-trivial stable shape. △ Less

Submitted 27 May, 2008; originally announced May 2008.

Comments: 5 pages, 4 figures

Journal ref: Phys. Rev. E 78, 021129 (2008)

Logarithmic corrections in the aging of the fully-frustrated Ising model

Authors: Jean-Charles Walter, Christophe Chatelain

Abstract: We study the dynamics of the critical two-dimensional fully-frustrated Ising model by means of Monte Carlo simulations. The dynamical exponent is estimated at equilibrium and is shown to be compatible with the value $z_c=2$. In a second step, the system is prepared in the paramagnetic phase and then quenched at its critical temperature $T_c=0$. Numerical evidences for the existence of logarithmi… ▽ More We study the dynamics of the critical two-dimensional fully-frustrated Ising model by means of Monte Carlo simulations. The dynamical exponent is estimated at equilibrium and is shown to be compatible with the value $z_c=2$. In a second step, the system is prepared in the paramagnetic phase and then quenched at its critical temperature $T_c=0$. Numerical evidences for the existence of logarithmic corrections in the aging regime are presented. These corrections may be related to the topological defects observed in other fully-frustrated models. The autocorrelation exponent is estimated to be $λ=d$ as for the Ising chain quenched at $T_c=0$. △ Less

Submitted 28 May, 2008; v1 submitted 13 July, 2007; originally announced July 2007.

Comments: 12 pages, 9 figures

Journal ref: J. Stat. Mech. (2008) P07005

Temperature-extended Jarzynski relation: Application to the numerical calculation of the surface tension

Authors: Christophe Chatelain

Abstract: We consider a generalization of the Jarzynski relation to the case where the system interacts with a bath for which the temperature is not kept constant but can vary during the transformation. We suggest to use this relation as a replacement to the thermodynamic perturbation method or the Bennett method for the estimation of the order-order surface tension by Monte Carlo simulations. To demonstr… ▽ More We consider a generalization of the Jarzynski relation to the case where the system interacts with a bath for which the temperature is not kept constant but can vary during the transformation. We suggest to use this relation as a replacement to the thermodynamic perturbation method or the Bennett method for the estimation of the order-order surface tension by Monte Carlo simulations. To demonstrate the feasibility of the method, we present some numerical data for the 3D Ising model. △ Less

Submitted 5 April, 2007; v1 submitted 2 February, 2007; originally announced February 2007.

Journal ref: J.Stat.Mech.0704:P04011,2007

Historical and interpretative aspects of quantum mechanics: a physicists' naive approach

Authors: B. Berche, C. Chatelain, C. Dufour, T. Gourieux, D. Karevski

Abstract: Many theoretical predictions derived from quantum mechanics have been confirmed experimentally during the last 80 years. However, interpretative aspects have long been subject to debate. Among them, the question of the existence of hidden variables is still open. We review these questions, paying special attention to historical aspects, and argue that one may definitively exclude local realism o… ▽ More Many theoretical predictions derived from quantum mechanics have been confirmed experimentally during the last 80 years. However, interpretative aspects have long been subject to debate. Among them, the question of the existence of hidden variables is still open. We review these questions, paying special attention to historical aspects, and argue that one may definitively exclude local realism on the basis of present experimental outcomes. Other interpretations of Quantum Mechanics are nevertheless not excluded. △ Less

Submitted 24 May, 2006; originally announced May 2006.

Comments: 30 pages

Journal ref: Condensed Matter Physics 9 (2006) 319

Probability distributions of the work in the 2D-Ising model

Authors: Christophe Chatelain, Dragi Karevski

Abstract: Probability distributions of the magnetic work are computed for the 2D Ising model by means of Monte Carlo simulations. The system is first prepared at equilibrium for three temperatures below, at and above the critical point. A magnetic field is then applied and grown linearly at different rates. Probability distributions of the work are stored and free energy differences computed using the Jar… ▽ More Probability distributions of the magnetic work are computed for the 2D Ising model by means of Monte Carlo simulations. The system is first prepared at equilibrium for three temperatures below, at and above the critical point. A magnetic field is then applied and grown linearly at different rates. Probability distributions of the work are stored and free energy differences computed using the Jarzynski equality. Consistency is checked and the dynamics of the system is analyzed. Free energies and dissipated works are reproduced with simple models. The critical exponent $δ$ is estimated in an usual manner. △ Less

Submitted 24 February, 2006; originally announced February 2006.

Comments: 12 pages, 6 figures. Comments are welcome

Journal ref: Journal of Statistical Mechanics: Theory and Experiment 18 (2006) P06005

Monte Carlo Study of Phase Transitions in the Bond-Diluted 3D 4-State Potts Model

Authors: Christophe Chatelain, Bertrand Berche, Wolfhard Janke, Pierre-Emmanuel Berche

Abstract: Large-scale Monte Carlo simulations of the bond-diluted three-dimensional 4-state Potts model are performed. The phase diagram and the physical properties at the phase transitions are studied using finite-size scaling techniques. Evidences are given for the existence of a tricritical point dividing the phase diagram into a regime where the transitions remain of first order and a second regime wh… ▽ More Large-scale Monte Carlo simulations of the bond-diluted three-dimensional 4-state Potts model are performed. The phase diagram and the physical properties at the phase transitions are studied using finite-size scaling techniques. Evidences are given for the existence of a tricritical point dividing the phase diagram into a regime where the transitions remain of first order and a second regime where the transitions are softened to continuous ones by the influence of disorder. In the former regime, the nature of the transition is essentially clarified through an analysis of the energy probability distribution. In the latter regime critical exponents are estimated. Rare and typical events are identified and their role is qualitatively discussed in both regimes. △ Less

Submitted 14 April, 2005; v1 submitted 6 January, 2005; originally announced January 2005.

Comments: Revised version, 40 pages, 30 figures

Journal ref: Nuclear Physics B 719/3 (2005) 275

Random Ising model in three dimensions: theory, experiment and simulation - a difficult coexistence

Authors: Bertrand Berche, Pierre-Emmanuel Berche, Christophe Chatelain, Wolfhard Janke

Abstract: We discuss different approaches for studying the influence of disorder in the three-dimensional Ising model. From the theoretical point of view, renormalisation group calculations provide quite accurate results. Experiments carried out on crystalline mixtures of compounds lead to measurements as accurate as three digits on the values of critical exponents. Numerically, extensive Monte Carlo simu… ▽ More We discuss different approaches for studying the influence of disorder in the three-dimensional Ising model. From the theoretical point of view, renormalisation group calculations provide quite accurate results. Experiments carried out on crystalline mixtures of compounds lead to measurements as accurate as three digits on the values of critical exponents. Numerically, extensive Monte Carlo simulations then pretend to be of comparable accuracy. Life becomes complicated when details are compared between the three approaches. △ Less

Submitted 3 December, 2004; v1 submitted 10 November, 2004; originally announced November 2004.

Comments: Latex file, 4 figures

Journal ref: Condensed Matter Physics 8 (2005) pp 47-58

On universality in aging ferromagnets

Authors: Christophe Chatelain

Abstract: This work is a contribution to the study of universality in out-of-equilibrium lattice models undergoing a second-order phase transition at equilibrium. The experimental protocol that we have chosen is the following: the system is prepared in its high-temperature phase and then quenched at the critical temperature $T_c$. We investigated by mean of Monte Carlo simulations two quantities that are… ▽ More This work is a contribution to the study of universality in out-of-equilibrium lattice models undergoing a second-order phase transition at equilibrium. The experimental protocol that we have chosen is the following: the system is prepared in its high-temperature phase and then quenched at the critical temperature $T_c$. We investigated by mean of Monte Carlo simulations two quantities that are believed to take universal values: the exponent $λ/z$ obtained from the decay of autocorrelation functions and the asymptotic value $X_\infty$ of the fluctuation-dissipation ratio $X(t,s)$. This protocol was applied to the Ising model, the 3-state clock model and the 4-state Potts model on square, triangular and honeycomb lattices and to the Ashkin-Teller model at the point belonging at equilibrium to the 3-state Potts model universality class and to a multispin Ising model and the Baxter-Wu model both belonging to the 4-state Potts model universality class at equilibrium. △ Less

Submitted 17 June, 2004; v1 submitted 1 April, 2004; originally announced April 2004.

Comments: 17 pages

Journal ref: J. Stat. Mech.: Theor. Exp. - (2004) P06006

Bond dilution in the 3D Ising model: a Monte Carlo study

Authors: Pierre-Emmanuel Berche, Christophe Chatelain, Bertrand Berche, Wolfhard Janke

Abstract: We study by Monte Carlo simulations the influence of bond dilution on the three-dimensional Ising model. This paradigmatic model in its pure version displays a second-order phase transition with a positive specific heat critical exponent $α$. According to the Harris criterion disorder should hence lead to a new fixed point characterized by new critical exponents. We have determined the phase dia… ▽ More We study by Monte Carlo simulations the influence of bond dilution on the three-dimensional Ising model. This paradigmatic model in its pure version displays a second-order phase transition with a positive specific heat critical exponent $α$. According to the Harris criterion disorder should hence lead to a new fixed point characterized by new critical exponents. We have determined the phase diagram of the diluted model, between the pure model limit and the percolation threshold. For the estimation of critical exponents, we have first performed a finite-size scaling study, where we concentrated on three different dilutions. We emphasize in this work the great influence of the cross-over phenomena between the pure, disorder and percolation fixed points which lead to effective critical exponents dependent on the concentration. In a second set of simulations, the temperature behaviour of physical quantities has been studied in order to characterize the disorder fixed point more accurately. In particular this allowed us to estimate ratios of some critical amplitudes. In accord with previous observations for other models this provides stronger evidence for the existence of the disorder fixed point since the amplitude ratios are more sensitive to the universality class than the critical exponents. Moreover, the question of non-self-averaging at the disorder fixed point is investigated and compared with recent results for the bond-diluted $q=4$ Potts model. Overall our numerical results provide evidence that, as expected on theoretical grounds, the critical behaviour of the bond-diluted model is governed by the same universality class as the site-diluted model. △ Less

Submitted 24 February, 2004; originally announced February 2004.

Journal ref: European Physical Journal B 38 (2004) 463-474

Quenched Disorder Distributions in Three-Dimensional Diluted Ferromagnets

Authors: W. Janke, P. -E. Berche, C. Chatelain, B. Berche

Abstract: We report on large-scale Monte Carlo simulations of the three-dimensional 4-state Potts ferromagnet subject to quenched, random bond dilution. For small dilutions the rather strong first-order phase transition of the pure system is found to persist, whereas for larger dilutions the theoretically expected softening to a continuous transition is confirmed. The properties of the underlying disorder… ▽ More We report on large-scale Monte Carlo simulations of the three-dimensional 4-state Potts ferromagnet subject to quenched, random bond dilution. For small dilutions the rather strong first-order phase transition of the pure system is found to persist, whereas for larger dilutions the theoretically expected softening to a continuous transition is confirmed. The properties of the underlying disorder distributions of thermal observables are discussed and illustrated with a few selected examples. △ Less

Submitted 28 April, 2003; originally announced April 2003.

Comments: 5 pages, LaTeX, 4 Postscript figures, Author Information under http://www.physik.uni-leipzig.de/CQT

Report number: Leipzig-LU-ITP 2003/008

Journal ref: Computer Simulation Studies in Condensed-Matter Physics XVI (eds. D.P. Landau, S.P. Lewis, H.-B. Schüttler) Berlin 2004 , p.89-94

A far-from-equilibrium fluctuation-dissipation relation for an Ising-Glauber-like model

Authors: Christophe Chatelain

Abstract: We derive an exact expression of the response function to an infinitesimal magnetic field for an Ising-Glauber-like model with arbitrary exchange couplings. The result is expressed in terms of thermodynamic averages and does not depend on the initial conditions or on the dimension of the space. The response function is related to time-derivatives of a complicated correlation function and so the… ▽ More We derive an exact expression of the response function to an infinitesimal magnetic field for an Ising-Glauber-like model with arbitrary exchange couplings. The result is expressed in terms of thermodynamic averages and does not depend on the initial conditions or on the dimension of the space. The response function is related to time-derivatives of a complicated correlation function and so the expression is a generalisation of the equilibrium fluctuation-dissipation theorem in the special case of this model. Correspondence with the Ising-Glauber model is discussed. A discrete-time version of the relation is implemented in Monte Carlo simulations and then used to study the aging regime of the ferromagnetic two-dimensional Ising-Glauber model quenched from the paramagnetic phase to the ferromagnetic one. Our approach has the originality to give direct access to the response function and the fluctuation-dissipation ratio. △ Less

Submitted 21 July, 2003; v1 submitted 26 March, 2003; originally announced March 2003.

Comments: 16 pages, 4 figures, major improvement

Journal ref: J. Phys. A 36 (2003) 10739

Monte Carlo Studies of Three-Dimensional Bond-Diluted Ferromagnets

Authors: P. -E. Berche, C. Chatelain, B. Berche, W. Janke

Abstract: In this report we give an overview on recent results obtained from extensive Monte Carlo (MC) computer simulations of the 3D 2-state (Ising) and 4-state Potts models with bond-dilution. The motivation to study the 4-state Potts model derives from the fact that, in the pure case, this model is known to exhibit a fairly strong first-order transition, such that a disorder-induced softening to a sec… ▽ More In this report we give an overview on recent results obtained from extensive Monte Carlo (MC) computer simulations of the 3D 2-state (Ising) and 4-state Potts models with bond-dilution. The motivation to study the 4-state Potts model derives from the fact that, in the pure case, this model is known to exhibit a fairly strong first-order transition, such that a disorder-induced softening to a second-order transition would give clear support of the theoretical expectations. Modeling the disorder by bond-dilution enables in the Ising case a test of the expected universality with respect to the type of disorder. Furthermore, for both models this choice facilitates a quantitative comparison with recent high-temperature series expansions for general random-bond q-state Potts models. △ Less

Submitted 19 December, 2002; originally announced December 2002.

Comments: 12 pages, LaTeX, 11 Postscript figures

Report number: Leipzig-LU-ITP 2002/026

Journal ref: High Performance Computing in Science and Engineering, eds. S. Wagner, W. Hanke, A. Bode, F. Durst (2003) Springer (Berlin), p. 227-238