Model reductions for inference: generality of pairwise, binary, and planar factor graphs.

We offer a solution to the problem of efficiently translating algorithms between different types of discrete statistical model. We investigate the expressive power of three classes of model-those with binary variables, with pairwise factors, and with planar topology-as well as their four intersections. We formalize a notion of "simple reduction" for the problem of inferring marginal probabilities and consider whether it is possible to "simply reduce" marginal inference from general discrete factor graphs to factor graphs in each of these seven subclasses. We characterize the reducibility of each class, showing in particular that the class of binary pairwise factor graphs is able to simply reduce only positive models. We also exhibit a continuous "spectral reduction" based on polynomial interpolation, which overcomes this limitation. Experiments assess the performance of standard approximate inference algorithms on the outputs of our reductions.

Consensus on homogeneous manifolds

The present paper considers distributed consensus algorithms for agents evolving on a connected compact homogeneous (CCH) manifold. The agents track no external reference and communicate their relative state according to an interconnection graph. The paper first formalizes the consensus problem for synchronization (i.e. maximizing the consensus) and balancing (i.e. minimizing the consensus); it thereby introduces the induced arithmetic mean, an easily computable mean position on CCH manifolds. Then it proposes and analyzes various consensus algorithms on manifolds: natural gradient algorithms which reach local consensus equilibria; an adaptation using auxiliary variables for almost-global synchronization or balancing; and a stochastic gossip setting for global synchronization. It closes by investigating the dependence of synchronization properties on the attraction function between interacting agents on the circle. The theory is also illustrated on SO(n) and on the Grassmann manifolds. ©2009 IEEE.

Consensus optimization on manifolds

The present paper considers distributed consensus algorithms that involve N agents evolving on a connected compact homogeneous manifold. The agents track no external reference and communicate their relative state according to a communication graph. The consensus problem is formulated in terms of the extrema of a cost function. This leads to efficient gradient algorithms to synchronize (i.e., maximizing the consensus) or balance (i.e., minimizing the consensus) the agents; a convenient adaptation of the gradient algorithms is used when the communication graph is directed and time-varying. The cost function is linked to a specific centroid definition on manifolds, introduced here as the induced arithmetic mean, that is easily computable in closed form and may be of independent interest for a number of manifolds. The special orthogonal group SO (n) and the Grassmann manifold Grass (p, n) are treated as original examples. A link is also drawn with the many existing results on the circle. © 2009 Society for Industrial and Applied Mathematics.

Riemannian geometry of Grassmann manifolds with a view on algorithmic computation

We give simple formulas for the canonical metric, gradient, Lie derivative, Riemannian connection, parallel translation, geodesics and distance on the Grassmann manifold of p-planes in ℝn. In these formulas, p-planes are represented as the column space of n × p matrices. The Newton method on abstract Riemannian manifolds proposed by Smith is made explicit on the Grassmann manifold. Two applications - computing an invariant subspace of a matrix and the mean of subspaces - are worked out.

Optimization Algorithms on Matrix Manifolds

This book shows how to exploit the special structure of such problems to develop efficient numerical algorithms.

What is the weakest topology in which feedback stability is robust?

Mathematical theorems in control theory are only of interest in so far as their assumptions relate to practical situations. The space of systems with transfer functions in ℋ∞, for example, has many advantages mathematically, but includes large classes of non-physical systems, and one must be careful in drawing inferences from results in that setting. Similarly, the graph topology has long been known to be the weakest, or coarsest, topology in which (1) feedback stability is a robust property (i.e. preserved in small neighbourhoods) and (2) the map from open-to-closed-loop transfer functions is continuous. However, it is not known whether continuity is a necessary part of this statement, or only required for the existing proofs. It is entirely possible that the answer depends on the underlying classes of systems used. The class of systems we concern ourselves with here is the set of systems that can be approximated, in the graph topology, by real rational transfer function matrices. That is, lumped parameter models, or those distributed systems for which it makes sense to use finite element methods. This is precisely the set of systems that have continuous frequency responses in the extended complex plane. For this class, we show that there is indeed a weaker topology; in which feedback stability is robust but for which the maps from open-to-closed-loop transfer functions are not necessarily continuous. © 2013 Copyright Taylor and Francis Group, LLC.

Hybrid core carbon fiber composite sandwich panels: Fabrication and mechanical response

Carbon fiber reinforced polymer (CFRP) composite sandwich panels with hybrid foam filled CFRP pyramidal lattice cores have been assembled from a carbon fiber braided net, 3D woven face sheets and various polymeric foams, and infused with an epoxy resin using a vacuum assisted resin transfer process. Sandwich panels with a fixed CFRP truss mass have been fabricated using a variety of closed cell polymer and syntactic foams, resulting in core densities ranging from 44-482kgm-3. The through thickness and in-plane shear modulus and strength of the cores increased with increasing foam density. The use of low compressive strength foams within the core was found to result in a significant reduction in the compressive strength contributed by the CFRP trusses. X-ray tomography led to the discovery that the trusses develop an elliptical cross-section shape during pressure assisted resin transfer. The ellipticity of the truss cross-sections increased, and the lattice contribution to the core strength decreased as the foam density was reduced. Micromechanical modeling was used to investigate the relationships between the mechanical properties and volume fractions of the core materials and truss topology of the hybrid core. The specific strength and moduli of the hybrid cores lay between those of the CFRP lattices and foams used to fabricate them. However, their volumetric and gravimetric energy absorptions significantly exceeded those of the materials from which they were fabricated. They compare favorably with other lightweight energy absorbing materials and structures. © 2013.

Cost-effectiveness of alternative powertrains for reduced energy use and CO2 emissions in passenger vehicles

This work analysed the cost-effectiveness of avoiding carbon dioxide (CO2) emissions using advanced internal combustion engines, hybrids, plug-in hybrids, fuel cell vehicles and electric vehicles across the nine UK passenger vehicles segments. Across all vehicle types and powertrain groups, minimum installed motive power was dependent most on the time to accelerate from zero to 96.6km/h (60mph). Hybridising the powertrain reduced the difference in energy use between vehicles with slow (t z - 60 > 8 s) and fast acceleration (t z - 60 < 8 s) times. The cost premium associated with advanced powertrains was dependent most on the powertrain chosen, rather than the performance required. Improving non-powertrain components reduced vehicle road load and allowed total motive capacity to decrease by 17%, energy use by 11%, manufacturing cost premiums by 13% and CO2 emissions abatement costs by 15%. All vehicles with advanced internal combustion engines, most hybrid and plug-in hybrid powertrains reduced net CO2 emissions and had lower lifetime operating costs than the respective segment reference vehicle. Most powertrains using fuel cells and all electric vehicles had positive CO2 emissions abatement costs. However, only vehicles using advanced internal combustion engines and parallel hybrid vehicles may be attractive to consumers by the fuel savings offsetting increases in vehicle cost within two years. This work demonstrates that fuel savings are possible relative to today's fleet, but indicates that the most cost-effective way of reducing fuel consumption and CO2 emissions is by advanced combustion technologies and hybridisation with a parallel topology. © 2014 Elsevier Ltd.

Application of fast oxygen sensors for investigations into air-path dynamics and EGR distribution in a diesel engine

The control of NOX emissions by exhaust gas recirculation (EGR) is of widespread application. However, despite dramatic improvements in all aspects of engine control, the subtle mixing processes that determine the cylinder-to-cylinder distribution of the recirculated gas often results in a mal-distribution that is still an issue for the engine designer and calibrator. In this paper we demonstrate the application of a relatively straightforward technique for the measurement of the absolute and relative dilution quantity in both steady state and transient operation. This was achieved by the use of oxygen sensors based on standard UEGO (universal exhaust gas oxygen) sensors but packaged so as to give good frequency response (∼ 10 ms time constant) and be completely insensitivity to the sample pressure and temperature. Measurements can be made at almost any location of interest, for example exhaust and inlet manifolds as well as EGR path(s), with virtually no flow disturbance. At the same time, the measurements yield insights into air-path dynamics. We argue that "dilution", as indicated by the deviation of the oxygen concentration from that of air, is a more appropriate parameter than EGR rate in the context of NOX control, especially for diesel engines. Experimental results are presented for the EGR distribution in a current production light duty 4-cylinder diesel engine in which significant differences were found in the proportion of the recirculated gas that reached each cylinder. Even the individual inlet runners of the cylinders exhibited very different dilution rates - differences of nearly 50% were observed at some conditions. An application of such data may be in the improvement of calibration and validation of CFD and other modelling techniques. Copyright © 2014 SAE International.

Experimental study of the effects of temperature variation on droplet size and wetness fraction in a low-pressure model steam turbine

The three-stage low-pressure model steam turbine at the Institute of Thermal Turbomachinery and Machinery Laboratory (ITSM) was used to study the impact of three different steam inlet temperatures on the homogeneous condensation process and the resulting wetness topology. The droplet spectrum as well as the particle number concentration were measured in front of the last stage using an optical-pneumatic probe. At design load, condensation starts inside the stator of the second stage. A change in the steam inlet temperature is able to shift the location of condensation onset within the blade row up- or downstream and even into adjoining blade passages, which leads to significantly different local droplet sizes and wetness fractions due to different local expansion rates. The measured results are compared to steady three-dimensional computational fluid dynamics calculations. The predicted nucleation zones could be largely confirmed by the measurements. Although the trend of measured and calculated droplet size across the span is satisfactory, there are considerable differences between the measured and computed droplet spectrum and wetness fractions. © IMechE 2013 Reprints and permissions: sagepub.co.uk/ journalsPermissions.nav.

Linear dimensionality reduction: Survey, insights, and generalizations

© 2015 John P. Cunningham and Zoubin Ghahramani. Linear dimensionality reduction methods are a cornerstone of analyzing high dimensional data, due to their simple geometric interpretations and typically attractive computational properties. These methods capture many data features of interest, such as covariance, dynamical structure, correlation between data sets, input-output relationships, and margin between data classes. Methods have been developed with a variety of names and motivations in many fields, and perhaps as a result the connections between all these methods have not been highlighted. Here we survey methods from this disparate literature as optimization programs over matrix manifolds. We discuss principal component analysis, factor analysis, linear multidimensional scaling, Fisher's linear discriminant analysis, canonical correlations analysis, maximum autocorrelation factors, slow feature analysis, sufficient dimensionality reduction, undercomplete independent component analysis, linear regression, distance metric learning, and more. This optimization framework gives insight to some rarely discussed shortcomings of well-known methods, such as the suboptimality of certain eigenvector solutions. Modern techniques for optimization over matrix manifolds enable a generic linear dimensionality reduction solver, which accepts as input data and an objective to be optimized, and returns, as output, an optimal low-dimensional projection of the data. This simple optimization framework further allows straightforward generalizations and novel variants of classical methods, which we demonstrate here by creating an orthogonal-projection canonical correlations analysis. More broadly, this survey and generic solver suggest that linear dimensionality reduction can move toward becoming a blackbox, objective-agnostic numerical technology.

Manopt, a matlab toolbox for optimization on manifolds

Optimization on manifolds is a rapidly developing branch of nonlinear optimization. Its focus is on problems where the smooth geometry of the search space can be leveraged to design effcient numerical algorithms. In particular, optimization on manifolds is well-suited to deal with rank and orthogonality constraints. Such structured constraints appear pervasively in machine learning applications, including low-rank matrix completion, sensor network localization, camera network registration, independent component analysis, metric learning, dimensionality reduction and so on. The Manopt toolbox, available at www.manopt.org, is a user-friendly, documented piece of software dedicated to simplify experimenting with state of the art Riemannian optimization algorithms. By dealing internally with most of the differential geometry, the package aims particularly at lowering the entrance barrier. © 2014 Nicolas Boumal.

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.