International Professional University of Technology in Nagoya Repository
Not a member yet
15131 research outputs found
Sort by
Topic segmentation in ASR transcripts using bidirectional rnns for change detection
No full textInternational audienceTopic segmentation methods are mostly based on the idea of lexical cohesion, in which lexical distributions are analysed across the document and segment boundaries are marked in areas of low cohesion. We propose a novel approach for topic segmentation in speech recognition transcripts by measuring lexical cohesion using bidirectional Recurrent Neural Networks (RNN). The bidirectional RNNs capture context in the past and the following set of words. The past and following contexts are compared to perform topic change detection. In contrast to existing works based on sequence and discrim-inative models for topic segmentation, our approach does not use a segmented corpus nor (pseudo) topic labels for training. Our model is trained using news articles obtained from the internet. Evaluation on ASR transcripts of French TV broadcast news programs demonstrates the effectiveness of our proposed approach
Foundational nonuniform (Co)datatypes for higher-order logic
No full textInternational audienceNonuniform (or " nested " or " heterogeneous ") data-types are recursively defined types in which the type arguments vary recursively. They arise in the implementation of finger trees and other efficient functional data structures. We show how to reduce a large class of nonuniform datatypes and codatatypes to uniform types in higher-order logic. We programmed this reduction in the Isabelle/HOL proof assistant, thereby enriching its specification language. Moreover, we derive (co)induction and (co)recursion principles based on a weak variant of parametricity
A Formal Proof of the Expressiveness of Deep Learning
No full textInternational audienceDeep learning has had a profound impact on computer science in recent years, with applications to image recognition, language processing, bioinformatics, and more. Recently, Cohen et al. provided theoretical evidence for the superiority of deep learning over shallow learning. We formalized their mathematical proof using Isabelle/HOL. The Isabelle development simplifies and generalizes the original proof, while working around the limitations of the HOL type system. To support the formalization, we developed reusable libraries of formalized mathematics, including results about the matrix rank, the Borel measure, and multivariate polynomials as well as a library for tensor analysis
A Minimax Optimal Algorithm for Crowdsourcing
No full textInternational audienceWe consider the problem of accurately estimating the reliability of workers based on noisy labels they provide, which is a fundamental question in crowdsourcing. We propose a novel lower bound on the minimax estimation error which applies to any estimation procedure. We further propose Triangular Estimation (TE), an algorithm for estimating the reliability of workers. TE has low complexity, may be implemented in a streaming setting when labels are provided by workers in real time, and does not rely on an iterative procedure. We prove that TE is minimax optimal and matches our lower bound. We conclude by assessing the performance of TE and other state-of-the-art algorithms on both synthetic and real-world data
Metamaterial for microwave absorption improvement of pyramidal absorbers
No full textInternational audienc
Imaging Comparison of GPR Based on Multi-and Bi-static Configurations
No full textInternational audienceIn this paper, a comparison of two ground penetrating radar (GPR) systems is presented, the classical bi-static one and the Multiple Input Multiple Output (MIMO) system both using N Vivaldi antennas. For both configurations a Linear Sampling Metthod (LSM) imaging algorithm and Kirchhoff migration technique are used for target detection and localization. The performance of both systems are compared through a parametric study based on synthetic data modeling a GPR scene
A feasibility study for cerebrovascular diseases monitoring via microwave imaging
No full textInternational audienceWith this paper, we will report about the feasibility study of a novel device which will be capable of performing the follow-up of a brain stroke event in the post-acute stage. The underlying technology, used in the device under design, is microwave imaging, which exploits the difference in dielectric properties of tissues. Numerical full-wave simulations, with realistic 3-D phantoms of the human head, will show the potential capabilities of the microwave imaging device
Combination of artificial materials with conventional pyramidal absorbers for microwave absorption improvement
No full textInternational audienc
Power pattern sensitivity analysis of reflectarray antennas to substrate uncertainties through the minkowski interval analysis
No full textInternational audienceIn this work, the effect of the fabrication uncertainties on the substrate thickness on the radiation performance of the reflectarray antenna is addressed. An interval analysis (7A)-based method is used to compute the upper and lower bounds of the power pattern as a function of the deviations in the substrate thickness. The Minkowski Sum (MS) is exploited to combine interval phasors in order to mitigate the over-estimation of the power pattern bounds affecting standard Cartesian IA (IA-CS) techniques. A representative numerical example shows that the proposed IAMS method provides narrower and more reliable bounds than the IA-CS
Synthesis of clustered linear arrays through a total variation compressive sensing approach
No full textInternational audienceIn this work the problem of synthesizing the excitations of a linear array, clustered into contiguous sub-arrays of irregular length, is addressed. By suitably exploiting the behavior of clustered array aperture distributions (i.e., step-wise discrete functions), the problem has been formulated as the minimization of the total variation (TV) of the excitations, satisfying a matching condition on a predefined reference pattern. In virtue of the sparse nature of the unknowns, the minimization problem has been solved by means of an efficient total variation compressive sensing (TV-CS) optimization approach. A simple example validating the proposed technique is finally reported