International Professional University of Technology in Nagoya Repository
Not a member yet
15131 research outputs found
Sort by
Optimized Binary64 and Binary128 Arithmetic with GNU MPFR
No full textInternational audienceWe describe algorithms used to optimize the GNU MPFR library when the operands fit into one or two words. On modern processors, a correctly rounded addition of two quadruple precision numbers is now performed in 22 cycles, a subtraction in 24 cycles, a multiplication in 32 cycles, a division in 64 cycles, and a square root in 69 cycles. We also introduce a new faithful rounding mode, which enables even faster computations. Those optimizations will be available in version 4 of MPFR
A Discrete MRF Framework for Integrated Multi-Atlas Registration and Segmentation
No full textInternational audienceMulti-atlas segmentation has emerged in recent years as a simple yet powerful approach in medical image segmentation. It commonly comprises two steps: i) a series of pairwise registrations that establish correspondences between a query image and a number of atlases, and ii) the fusion of the available seg-mentation hypotheses towards labeling objects of interest. In this paper, we introduce a novel approach that solves simultaneously for the underlying segmentation labels and the multi-atlas registration. The proposed approach is formulated as a pairwise Markov Random Field, where registration and segmentation nodes are coupled towards simultaneously recovering all atlas deformations and labeling the query image. The coupling is achieved by promoting the consistency between selected deformed atlas segmentations and the estimated query segmentation. Additional membership fields are estimated, determining the participation of each atlas in labeling each voxel. Inference is performed by using a sequential relaxation scheme. The proposed approach is validated on the IBSR dataset and is compared against standard post-registration label fusion strategies. Promising results demonstrate the potential of our method
Fusion of Heterogeneous Data in Convolutional Networks for Urban Semantic Labeling (Invited Paper)
No full textInternational audienceIn this work, we present a novel module to perform fusion of heterogeneous data using fully convolutional networks for semantic labeling. We introduce residual correction as a way to learn how to fuse predictions coming out of a dual stream architecture. Especially, we perform fusion of DSM and IRRG optical data on the ISPRS Vaihingen dataset over a urban area and obtain new state-of-the-art results
Microscale temperature and SAR measurements in cell monolayer models exposed to millimeter waves
No full textInternational audienceDue to shallow penetration of millimeter waves (MMW) and convection in liquid medium surrounding cells, the problem of accurate assessment of local MMW heating in in vitro experiments remains unsolved. Conventional dosimetric MMW techniques, such as infrared imaging or fiber optic (FO) sensors, face several inherent limits. Here we propose a methodology for accurate local temperature measurement and subsequent specific absorption rate (SAR) retrieval using microscale thermocouples (TC). SAR was retrieved by fitting the measured initial temperature rise to the numerical solution of an equivalent thermal model. It was found that the accuracy of temperature measurement depends on thermosensor size, that is, the smaller TC, the more accurate the temperature measurement. SAR determined using TC with lead diameters of 25 and 75m demonstrated 98.5% and 80.4% match with computed SAR, respectively. However, both TC provided the same temperature rises in long run (>10min). FO probe failed to measure adequately local heating both for short and long exposures due to the relatively large size of the probe sensor (400m) and time constant (0.6s). Calculated SAR in the cell monolayer was almost two times lower than that in the surrounding liquid. It was shown that the impact of the cell monolayer on heating due to its small thickness (5 to 10m) can be considered as negligible. Moreover, we demonstrated the possibility of accurate measurement of MMW-induced thermal pulses (up to 10 degrees C) using 25m TC. Bioelectromagnetics. 38:11-21, 2017. (c) 2016 Wiley Periodicals, Inc
An empirical classification-based framework for the safety criticality assessment of energy production systems, in presence of inconsistent data
No full textInternational audienceThe technical problem addressed in the present paper is the assessment of the safety criticality of energy production systems. An empirical classification model is developed, based on the Majority Rule Sorting method, to evaluate the class of criticality of the plant/system of interest, with respect to safety. The model is built on the basis of a (limited-size) set of data representing the characteristics of a number of plants and their corresponding criticality classes, as assigned by experts.The construction of the classification model may raise two issues. First, the classification examples provided by the experts may contain contradictions: a validation of the consistency of the considered dataset is, thus, required. Second, uncertainty affects the process: a quantitative assessment of the performance of the classification model is, thus, in order, in terms of accuracy and confidence in the class assignments. In this paper, two approaches are proposed to tackle the first issue: the inconsistencies in the data examples are “resolved” by deleting or relaxing, respectively, some constraints in the model construction process. Three methods are proposed to address the second issue: (i) a model retrieval-based approach, (ii) the Bootstrap method and (iii) the cross-validation technique. Numerical analyses are presented with reference to an artificial case study regarding the classification of Nuclear Power Plants
Two Evidential Data Based Models for Influence Maximization in Twitter
No full textInternational audienceInfluence maximization is the problem of selecting a set of influential users in the social network. Those users could adopt the product and trigger a large cascade of adoptions through the " word of mouth " effect. In this paper, we propose two evidential influence maximization models for Twitter social network. The proposed approach uses the theory of belief functions to estimate users influence. Furthermore, the proposed influence estimation measure fuses many influence aspects in Twitter, like the importance of the user in the network structure and the popularity of user's tweets (messages). In our experiments, we compare the proposed solutions to existing ones and we show the performance of our models
Stability Analysis of Polynomially Dependent Systems by Eigenvalue Perturbation
No full textaccepté le 23 décembre 2016International audienc
Approximating the standard condition number for cognitive radio spectrum sensing with finite number of sensors
No full textInternational audienc
Data fusion for paroxysmal events' classification from EEG.
No full textInternational audienceSpatiotemporal analysis of electroencephalography is commonly used for classification of events since it allows capturing dependencies across channels. The significant increase of feature vector dimensionality however introduce noise and thus it does not allow the classification models to be trained using a limited number of samples usually available in clinical studies.Thus, we investigate the classification of epileptic and non-epileptic events based on temporal and spectral analysis through the application of three different fusion schemes for the combination of information across channels. We compare the commonly used early-integration (EI) scheme - in which features are fused from all channels prior to classification - with two late-integration (LI) schemes performing per channel classification when: (i) the temporal context varies significantly across channels, thus local spatial training models are required, and (ii) the spatial variations are negligible in comparison to the inter-subject variation, thus only the temporal variation is modeled using a single global spatial training model. Furthermore, we perform dimensionality reduction either by feature selection or by principal component analysis.The framework is applied on events that manifest across most channels, as generalized epileptic seizures, psychogenic non-epileptic seizures and vasovagal syncope. The three classification architectures were evaluated on EEG epochs from 11 subjects.Although direct comparison with other studies is difficult due to the different characteristics of each dataset, the achieved recognition accuracy of the LI fusion schemes outperforms the performance reported in the literature.The best scheme was the LI with global model which achieved 97% accuracy
