Phoneme recognition can be viewed as classifying multivariate observations. Multi-layer perceptrons (MLP) and probabilistic neural networks (PNN) approach the decision problem using two complementary models. The MLP models the discriminant surfaces between different phoneme categories, essentially by piece-wise planar approximations, while the PNN approximates class conditional probability dens...

Time-wise velocity signals obtained from large-eddy simulation (LES) within the near-wall, logarithmic region of the zeropressure gradient, flat-plate turbulent boundary layer are used as input to a calibrated, empirical wall model to calculate the statistics of the fluctuating, wall shear stress τw. The results, together with those from numerical simulation and experiment indicate a log-variat...

While neural machine translation (NMT) has made remarkable progress in recent years, it is hard to interpret its internal workings due to the continuous representations and non-linearity of neural networks. In this work, we propose to use layer-wise relevance propagation (LRP) to compute the contribution of each contextual word to arbitrary hidden states in the attention-based encoderdecoder fr...

To infer a multilayer representation of high-dimensional count vectors, we propose the Poisson gamma belief network (PGBN) that factorizes each of its layers into the product of a connection weight matrix and the nonnegative real hidden units of the next layer. The PGBN’s hidden layers are jointly trained with an upward-downward Gibbs sampler, each iteration of which upward samples Dirichlet di...

In the present manuscript, a Layer-Wise (LW) generalized model is proposed for linear static analysis of doublycurved shells constrained with general boundary conditions under influence concentrated and surface loads. The unknown field variable modelled employing polynomials various orders, each them defined within layer structure. As particular case LW model, an Equivalent Single Layer (ESL) f...

Despite the tremendous achievements of deep convolutional neural networks (CNNs) in most of computer vision tasks, understanding how they actually work remains a significant challenge. In this paper, we propose a novel two-step visualization method that aims to shed light on how deep CNNs recognize images and the objects therein. We start out with a layer-wise relevance propagation (LRP) step w...

This paper deals with the development of a new three-dimensional element with two-dimensional kinematic constraints capable of analysing the mechanical behaviour. of the laminated anisotropic shell-type structures. This element, originally developed for the linear 'analysis of plates, is extended for the linear analysis of laminated composite shells. The element can represent arbitrarily curved...

and quantum settings Model reduction is a highly desirable process for deep neural networks. While large networks are theoretically capable of learning arbitrarily complex models, overfitting and model redundancy negatively affects the prediction accuracy and model variance. Net-Trim is a layer-wise convex framework to prune (sparsify) deep neural networks. The method is applicable to neural ne...