Linear regression under fixed-rank constraints: A Riemannian approach

In this paper, we tackle the problem of learning a linear regression model whose parameter is a fixed-rank matrix. We study the Riemannian manifold geometry of the set of fixed-rank matrices and develop efficient line-search algorithms. The proposed algorithms have many applications, scale to high-dimensional problems, enjoy local convergence properties and confer a geometric basis to recent contributions on learning fixed-rank matrices. Numerical experiments on benchmarks suggest that the proposed algorithms compete with the state-of-the-art, and that manifold optimization offers a versatile framework for the design of rank-constrained machine learning algorithms. Copyright 2011 by the author(s)/owner(s).

Non linear constitutive models for lattice materials

We use a computational homogenisation approach to derive a non linear constitutive model for lattice materials. A representative volume element (RVE) of the lattice is modelled by means of discrete structural elements, and macroscopic stress-strain relationships are numerically evaluated after applying appropriate periodic boundary conditions to the RVE. The influence of the choice of the RVE on the predictions of the model is discussed. The model has been used for the analysis of the hexagonal and the triangulated lattices subjected to large strains. The fidelity of the model has been demonstrated by analysing a plate with a central hole under prescribed in plane compressive and tensile loads, and then comparing the results from the discrete and the homogenised models. © 2013 Elsevier Ltd.

Viscoelastic microbuckling of fiber composites

A theoretical study is given of viscoelastic microbuckling of fiber composites. The analysis is formulated in terms of general linear viscoelastic behavior within the kink band. Material outside the kink band is assumed to behave elastically. Two specific forms of linear viscoelastic behavior are considered: a standard linear viscoelastic model and a logarithmically creeping model. Results are provided as deformation versus time histories and failure life versus applied stress. Failure is due to either the attainment of a critical failure strain in the kink band or to the intervention of a different failure mechanism such as plastic microbuckling.

Fault-tolerant aspects of MPC

The architecture of model predictive control (MPC), with its explicit internal model and constrained optimization is presented. Since MPC relies on an explicit internal model, one can imagine dealing with failures by updating the internal model, and letting the on-line optimizer work out how to control the system in its new condition. This aspects rely on assumptions such that the nature of the fault can be located, and the model can be updated automatically. A standard form of MPC, with linear inequality constraints on inputs and outputs, linear internal model, and quadriatic cost function.

Slowness: an objective for spike-timing-dependent plasticity?

Our nervous system can efficiently recognize objects in spite of changes in contextual variables such as perspective or lighting conditions. Several lines of research have proposed that this ability for invariant recognition is learned by exploiting the fact that object identities typically vary more slowly in time than contextual variables or noise. Here, we study the question of how this "temporal stability" or "slowness" approach can be implemented within the limits of biologically realistic spike-based learning rules. We first show that slow feature analysis, an algorithm that is based on slowness, can be implemented in linear continuous model neurons by means of a modified Hebbian learning rule. This approach provides a link to the trace rule, which is another implementation of slowness learning. Then, we show analytically that for linear Poisson neurons, slowness learning can be implemented by spike-timing-dependent plasticity (STDP) with a specific learning window. By studying the learning dynamics of STDP, we show that for functional interpretations of STDP, it is not the learning window alone that is relevant but rather the convolution of the learning window with the postsynaptic potential. We then derive STDP learning windows that implement slow feature analysis and the "trace rule." The resulting learning windows are compatible with physiological data both in shape and timescale. Moreover, our analysis shows that the learning window can be split into two functionally different components that are sensitive to reversible and irreversible aspects of the input statistics, respectively. The theory indicates that irreversible input statistics are not in favor of stable weight distributions but may generate oscillatory weight dynamics. Our analysis offers a novel interpretation for the functional role of STDP in physiological neurons.

Fertile-free fuels in pressurized water reactors: Design challenges and solutions

This paper investigates the basic feasibility of using reactor-grade Pu in fertile-free fuel (FFF) matrix in pressurized water reactors (PWRs). Several important issues were investigated in this work: the Pu loading required to achieve a specific interrefueling interval, the impact of inert matrix composition on reactivity constrained length of cycle, and the potential of utilizing burnable poisons (BPs) to alleviate degradation of the reactivity control mechanism and temperature coefficients. Although the subject was addressed in the past, no systematic approach for assessment of BP utilization in FFF cores was published. In this work, we examine all commercially available BP materials in all geometrical arrangements currently used by the nuclear industry with regards to their potential to alleviate the problems associated with the use of FFF in PWRs. The recently proposed MgO-ZrO2 solid-state solution fuel matrix, which appears to be very promising in terms of thermal properties and radiation damage resistance, was used as a reference matrix material in this work. The neutronic impact of the relative amounts of MgO and ZrO2 in the matrix were also studied. The analysis was performed with a neutron transport and fuel assembly burnup code BOXER. A modified linear reactivity model was applied to the two-dimensional single fuel assembly results to approximate the full core characteristics. Based on the results of the performed analyses, the Pu-loaded FFF core demonstrated potential feasibility to be used in existing PWRs. Major FFF core design problems may be significantly mitigated through the correct choice of BP design. It was found that a combination of BP materials and geometries may be required to meet all FFF design goals. The use of enriched (in most effective isotope) BPs, such as 167Er and 157Gd, may further improve the BP effectiveness and reduce the fuel cycle length penalty associated with their use.

Prediction of combustion noise for an aeroengine combustor

Combustion noise may become an important noise source for lean-burn gas turbine engines, and this noise is usually associated with highly unsteady flames. This work aims to compute the broadband combustion noise spectrum for a realistic aeroengine combustor and to compare with available measured noise data on a demonstrator aeroengine. A low-order linear network model is applied to a demonstrator engine combustor to obtain the transfer function that relates to unsteadiness in the rate of heat release, acoustic, entropic, and vortical fluctuations. A spectral model is used for the heat release rate fluctuation, which is the source of the noise. The mean flow of the aeroengine combustor required as input data to this spectral model is obtained from Reynolds-averaged Navier-Stokes simulations. The computed acoustic field for a low-to-medium power setting indicates that the models used in this study capture the main characteristics of the broadband spectral shape of combustion noise. Reasonable agreement with the measured spectral level is achieved. © 2012 AIAA.

Fixed-rank matrix factorizations and Riemannian low-rank optimization

Motivated by the problem of learning a linear regression model whose parameter is a large fixed-rank non-symmetric matrix, we consider the optimization of a smooth cost function defined on the set of fixed-rank matrices. We adopt the geometric framework of optimization on Riemannian quotient manifolds. We study the underlying geometries of several well-known fixed-rank matrix factorizations and then exploit the Riemannian quotient geometry of the search space in the design of a class of gradient descent and trust-region algorithms. The proposed algorithms generalize our previous results on fixed-rank symmetric positive semidefinite matrices, apply to a broad range of applications, scale to high-dimensional problems, and confer a geometric basis to recent contributions on the learning of fixed-rank non-symmetric matrices. We make connections with existing algorithms in the context of low-rank matrix completion and discuss the usefulness of the proposed framework. Numerical experiments suggest that the proposed algorithms compete with state-of-the-art algorithms and that manifold optimization offers an effective and versatile framework for the design of machine learning algorithms that learn a fixed-rank matrix. © 2013 Springer-Verlag Berlin Heidelberg.