Modelling human preferences for ranking and collaborative filtering: a probabilistic ordered partition approach

Learning preference models from human generated data is an important task in modern information processing systems. Its popular setting consists of simple input ratings, assigned with numerical values to indicate their relevancy with respect to a specific query. Since ratings are often specified within a small range, several objects may have the same ratings, thus creating ties among objects for a given query. Dealing with this phenomena presents a general problem of modelling preferences in the presence of ties and being query-specific. To this end, we present in this paper a novel approach by constructing probabilistic models directly on the collection of objects exploiting the combinatorial structure induced by the ties among them. The proposed probabilistic setting allows exploration of a super-exponential combinatorial state-space with unknown numbers of partitions and unknown order among them. Learning and inference in such a large state-space are challenging, and yet we present in this paper efficient algorithms to perform these tasks. Our approach exploits discrete choice theory, imposing generative process such that the finite set of objects is partitioned into subsets in a stagewise procedure, and thus reducing the state-space at each stage significantly. Efficient Markov chain Monte Carlo algorithms are then presented for the proposed models. We demonstrate that the model can potentially be trained in a large-scale setting of hundreds of thousands objects using an ordinary computer. In fact, in some special cases with appropriate model specification, our models can be learned in linear time. We evaluate the models on two application areas: (i) document ranking with the data from the Yahoo! challenge and (ii) collaborative filtering with movie data. We demonstrate that the models are competitive against state-of-the-arts.

Effect of elastic modulus and poisson's ratio on guided wave dispersion using transversely isotropic material modelling

Timber poles are commonly used for telecommunication and power distribution networks, wharves or jetties, piling or as a substructure of short span bridges. Most of the available techniques currently used for non-destructive testing (NDT) of timber structures are based on one-dimensional wave theory. If it is essential to detect small sized damage, it becomes necessary to consider guided wave (GW) propagation as the behaviour of different propagating modes cannot be represented by one-dimensional approximations. However, due to the orthotropic material properties of timber, the modelling of guided waves can be complex. No analytical solution can be found for plotting dispersion curves for orthotropic thick cylindrical waveguides even though very few literatures can be found on the theory of GW for anisotropic cylindrical waveguide. In addition, purely numerical approaches are available for solving these curves. In this paper, dispersion curves for orthotropic cylinders are computed using the scaled boundary finite element method (SBFEM) and compared with an isotropic material model to indicate the importance of considering timber as an anisotropic material. Moreover, some simplification is made on orthotropic behaviour of timber to make it transversely isotropic due to the fact that, analytical approaches for transversely isotropic cylinder are widely available in the literature. Also, the applicability of considering timber as a transversely isotropic material is discussed. As an orthotropic material, most material testing results of timber found in the literature include 9 elastic constants (three elastic moduli and six Poisson's ratios), hence it is essential to select the appropriate material properties for transversely isotropic material which includes only 5 elastic constants. Therefore, comparison between orthotropic and transversely isotropic material model is also presented in this article to reveal the effect of elastic moduli and Poisson's ratios on dispersion curves. Based on this study, some suggestions are proposed on selecting the parameters from an orthotropic model to transversely isotropic condition.

Strain gradient plasticity modelling of high-pressure torsion

Gradient plasticity modelling combining a micro-structure-related constitutive description of the local material behaviour with a particular gradient plasticity frame is presented. The constitutive formulation is based on a phase mixture model in which the dislocation cell walls and the cell interiors are considered as separate 'phases', the respective dislocation densities entering as internal variables. Two distinct physical mechanisms, which give rise to gradient plasticity, are considered. The first one is associated with the occurrence of geometrically necessary dislocations leading to first-order strain gradients; the second one is associated with the reaction stresses due to plastic strain incompatibilities between neighbouring grains, which lead to second-order strain gradients. These two separate variants of gradient plasticity were applied to the case of high-pressure torsion: a process known to result in a fairly uniform, ultrafine grained structure of metals. It is shown that the two complementary variants of gradient plasticity can both account for the experimental results, thus resolving a controversial issue of the occurrence of a uniform micro-structure as a result of an inherently non-uniform process. © 2007 Elsevier Ltd. All rights reserved.

Prediction of mode I delamination resistance of z-pinned laminates using the embedded finite element technique

A new finite modelling approach is presented to analyse the mode I delamination fracture toughness of z-pinned laminates using the computationally efficient embedded element technique. In the FE model,each z-pin is represented by a single one-dimensional truss element that is embedded within the laminate. Each truss is given the material, geometric and spatial properties associated with the global crackbridging traction response of a z-pin in the laminate; this simplification provides a computationally efficient and flexible model where pin elements are independent of the underlying structural mesh for thelaminate. The accuracy of the FE modelling approach is assessed using mode I interlaminar fracture toughness data for a carbon-epoxy laminate reinforced with z-pins made of copper, titanium or stainless steel. The model is able to predict with good accuracy the crack growth resistance curves and fracture toughness properties for the different types of z-pinned laminate.