A unifying view of multiple kernel learning

Recent research on multiple kernel learning has lead to a number of approaches for combining kernels in regularized risk minimization. The proposed approaches include different formulations of objectives and varying regularization strategies. In this paper we present a unifying optimization criterion for multiple kernel learning and show how existing formulations are subsumed as special cases. We also derive the criterion’s dual representation, which is suitable for general smooth optimization algorithms. Finally, we evaluate multiple kernel learning in this framework analytically using a Rademacher complexity bound on the generalization error and empirically in a set of experiments.

A unifying view of multiple kernel learning

Recent research on multiple kernel learning has lead to a number of approaches for combining kernels in regularized risk minimization. The proposed approaches include different formulations of objectives and varying regularization strategies. In this paper we present a unifying general optimization criterion for multiple kernel learning and show how existing formulations are subsumed as special cases. We also derive the criterion's dual representation, which is suitable for general smooth optimization algorithms. Finally, we evaluate multiple kernel learning in this framework analytically using a Rademacher complexity bound on the generalization error and empirically in a set of experiments.

Learning the Kernel Matrix with Semi-Definite Programming

Kernel-based learning algorithms work by embedding the data into a Euclidean space, and then searching for linear relations among the embedded data points. The embedding is performed implicitly, by specifying the inner products between each pair of points in the embedding space. This information is contained in the so-called kernel matrix, a symmetric and positive definite matrix that encodes the relative positions of all points. Specifying this matrix amounts to specifying the geometry of the embedding space and inducing a notion of similarity in the input space -- classical model selection problems in machine learning. In this paper we show how the kernel matrix can be learned from data via semi-definite programming (SDP) techniques. When applied to a kernel matrix associated with both training and test data this gives a powerful transductive algorithm -- using the labelled part of the data one can learn an embedding also for the unlabelled part. The similarity between test points is inferred from training points and their labels. Importantly, these learning problems are convex, so we obtain a method for learning both the model class and the function without local minima. Furthermore, this approach leads directly to a convex method to learn the 2-norm soft margin parameter in support vector machines, solving another important open problem. Finally, the novel approach presented in the paper is supported by positive empirical results.

The rademacher complexity of coregularized kernel classes

In the multi-view approach to semisupervised learning, we choose one predictor from each of multiple hypothesis classes, and we co-regularize our choices by penalizing disagreement among the predictors on the unlabeled data. We examine the co-regularization method used in the co-regularized least squares (CoRLS) algorithm, in which the views are reproducing kernel Hilbert spaces (RKHS's), and the disagreement penalty is the average squared difference in predictions. The final predictor is the pointwise average of the predictors from each view. We call the set of predictors that can result from this procedure the co-regularized hypothesis class. Our main result is a tight bound on the Rademacher complexity of the co-regularized hypothesis class in terms of the kernel matrices of each RKHS. We find that the co-regularization reduces the Rademacher complexity by an amount that depends on the distance between the two views, as measured by a data dependent metric. We then use standard techniques to bound the gap between training error and test error for the CoRLS algorithm. Experimentally, we find that the amount of reduction in complexity introduced by co regularization correlates with the amount of improvement that co-regularization gives in the CoRLS algorithm.

The undirected feedback vertex set problem has a Poly(k) Kernel

Resolving a noted open problem, we show that the Undirected Feedback Vertex Set problem, parameterized by the size of the solution set of vertices, is in the parameterized complexity class Poly(k), that is, polynomial-time pre-processing is sufficient to reduce an initial problem instance (G, k) to a decision-equivalent simplified instance (G', k') where k' � k, and the number of vertices of G' is bounded by a polynomial function of k. Our main result shows an O(k11) kernelization bound.

Residential satisfaction in inner urban higher-density Brisbane, Australia : role of dwelling, neighbours and neighbourhood

Increasing the population density of urban areas is a key policy strategy to sustainably manage growth, but many residents often view higher density living as an undesirable long-term housing option. Thus, this research explores the predictors of residential satisfaction in inner urban higher-density (IUHD) environments, surveying 636 IUHD residents in Brisbane, Australia about the importance of dwelling, neighbours and neighbourhood. Relationships with immediate neighbours did not predict residential satisfaction, but features of the neighbourhood and dwelling were critical, specifically satisfaction with dwelling position, design and facilities, and social contacts (family and friends) in the neighbourhood. Identifying the factors that influence residential satisfaction in IUHD will assist with both planning and design, helping ensure a lower resident turnover rate and greater uptake of high density living.

Driving speed is altered by monocular Neutral Density filters : the Enright phenomenon

Introduction: An observer, looking sideways from a moving vehicle, while wearing a neutral density filter over one eye, can have a distorted perception of speed, known as the Enright phenomenon. The purpose of this study was to determine how the Enright phenomenon influences driving behaviour. Methods: A geometric model of the Enright phenomenon was developed. Ten young, visually normal, participants (mean age = 25.4 years) were tested on a straight section of a closed driving circuit and instructed to look out of the right side of the vehicle and drive at either 40 Km/h or 60 Km/h under the following binocular viewing conditions: with a 0.9 ND filter over the left eye (leading eye); 0.9 ND filter over the right eye (trailing eye); 0.9 ND filters over both eyes, and with no filters over either eye. The order of filter conditions was randomised and the speed driven recorded for each condition. Results: Speed judgments did not differ significantly between the two baseline conditions (no filters and both eyes filtered) for either speed tested. For the baseline conditions, when subjects were asked to drive at 60 Km/h they matched this speed well (61 ± 10.2 Km/h) but drove significantly faster than requested (51.6 ± 9.4 Km/h) when asked to drive at 40 Km/h. Subjects significantly exceeded baseline speeds by 8.7± 5.0 Km/h, when the trailing eye was filtered and travelled slower than baseline speeds by 3.7± 4.6 Km/h when the leading eye was filtered. Conclusions: This is the first quantitative study demonstrating how the Enright effect can influence perceptions of driving speed, and demonstrates that monocular filtering of an eye can significantly impact driving speeds, albeit to a lesser extent than predicted by geometric models of the phenomenon.

Sub-tropical high-density, low rise residential buildings: case study 3

The case study 3 team viewed the mitigation of noise and air pollution generated in the transport corridor that borders the study site to be a paramount driver of the urban design solution. These key urban planning strategies were adopted: * Spatial separation from transport corridor pollution source. A linear green zone and environmental buffer was proposed adjacent to the transport corridor to mitigate the environmental noise and air quality impacts of the corridor, and to offer residents opportunities for recreation * Open space forming the key structural principle for neighbourhood design. A significant open space system underpins the planning and manages surface water flows. * Urban blocks running on east-west axis. The open space rationale emphasises an east-west pattern for local streets. Street alignment allows for predominantly north-south facing terrace type buildings which both face the street and overlook the green courtyard formed by the perimeter buildings. The results of the ESD assessment of the typologies conclude that the design will achieve good outcomes through: * Lower than average construction costs compared with other similar projects * Thermal comfort; A good balance between daylight access and solar gains is achieved * The energy rating achieved for the units is 8.5 stars.

Homo- and heteronuclear meso,meso-(E)-Ethene-1,2-diyl-Linked Diporphyrins : preparation, X-ray crystal structure, electronic absorption and emission spectra and density functional theory calculations

Homo-and heteronuclear meso,meso-(E)-ethene-1,2-diyl-linked diporphyrins have been prepared by the Suzuki coupling of porphyrinylboronates and iodovinylporphyrins. Combinations comprising 5,10,15-triphenylporphyrin (TriPP) on both ends of the ethene-1,2-diyl bridge M 210 (M 2=H 2/Ni, Ni 2, Ni/Zn, H 4, H 2Zn, Zn 2) and 5,15-bis(3,5-di-tert-butylphenyl)porphyrinato-nickel(II) on one end and H 2, Ni, and ZnTriPP on the other (M 211), enable the first studies of this class of compounds possessing intrinsic polarity. The compounds were characterized by electronic absorption and steady state emission spectra, 1H NMR spectra, and for the Ni 2 bis(TriPP) complex Ni 210, single crystal X-ray structure determination. The crystal structure shows ruffled distortions of the porphyrin rings, typical of Ni II porphyrins, and the (E)-C 2H 2 bridge makes a dihedral angle of 50° with the mean planes of the macrocycles. The result is a stepped parallel arrangement of the porphyrin rings. The dihedral angles in the solid state reflect the interplay of steric and electronic effects of the bridge on interporphyrin communication. The emission spectra in particular, suggest energy transfer across the bridge is fast in conformations in which the bridge is nearly coplanar with the rings. Comparisons of the fluorescence behaviour of H 410 and H 2Ni10 show strong quenching of the free base fluorescence when the complex is excited at the lower energy component of the Soret band, a feature associated in the literature with more planar conformations. TDDFT calculations on the gas-phase optimized geometry of Ni 210 reproduce the features of the experimental electronic absorption spectrum within 0.1 eV. © 2012 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim.

Sensitivity of the NMR density matrix to pulse sequence parameters : a simplified analytic approach

We present a formalism for the analysis of sensitivity of nuclear magnetic resonance pulse sequences to variations of pulse sequence parameters, such as radiofrequency pulses, gradient pulses or evolution delays. The formalism enables the calculation of compact, analytic expressions for the derivatives of the density matrix and the observed signal with respect to the parameters varied. The analysis is based on two constructs computed in the course of modified density-matrix simulations: the error interrogation operators and error commutators. The approach presented is consequently named the Error Commutator Formalism (ECF). It is used to evaluate the sensitivity of the density matrix to parameter variation based on the simulations carried out for the ideal parameters, obviating the need for finite-difference calculations of signal errors. The ECF analysis therefore carries a computational cost comparable to a single density-matrix or product-operator simulation. Its application is illustrated using a number of examples from basic NMR spectroscopy. We show that the strength of the ECF is its ability to provide analytic insights into the propagation of errors through pulse sequences and the behaviour of signal errors under phase cycling. Furthermore, the approach is algorithmic and easily amenable to implementation in the form of a programming code. It is envisaged that it could be incorporated into standard NMR product-operator simulation packages.