Structure Discovery in Nonparametric Regression through Compositional Kernel Search

Despite its importance, choosing the structural form of the kernel in nonparametric regression remains a black art. We define a space of kernel structures which are built compositionally by adding and multiplying a small number of base kernels. We present a method for searching over this space of structures which mirrors the scientific discovery process. The learned structures can often decompose functions into interpretable components and enable long-range extrapolation on time-series datasets. Our structure search method outperforms many widely used kernels and kernel combination methods on a variety of prediction tasks.

Convex relaxation of mixture regression with efficient algorithms

We develop a convex relaxation of maximum a posteriori estimation of a mixture of regression models. Although our relaxation involves a semidefinite matrix variable, we reformulate the problem to eliminate the need for general semidefinite programming. In particular, we provide two reformulations that admit fast algorithms. The first is a max-min spectral reformulation exploiting quasi-Newton descent. The second is a min-min reformulation consisting of fast alternating steps of closed-form updates. We evaluate the methods against Expectation-Maximization in a real problem of motion segmentation from video data.

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).

Regression on fixed-rank positive semidefinite matrices: A Riemannian approach

The paper addresses the problem of learning a regression model parameterized by a fixed-rank positive semidefinite matrix. The focus is on the nonlinear nature of the search space and on scalability to high-dimensional problems. The mathematical developments rely on the theory of gradient descent algorithms adapted to the Riemannian geometry that underlies the set of fixedrank positive semidefinite matrices. In contrast with previous contributions in the literature, no restrictions are imposed on the range space of the learned matrix. The resulting algorithms maintain a linear complexity in the problem size and enjoy important invariance properties. We apply the proposed algorithms to the problem of learning a distance function parameterized by a positive semidefinite matrix. Good performance is observed on classical benchmarks. © 2011 Gilles Meyer, Silvere Bonnabel and Rodolphe Sepulchre.

Application of axial discontinuity factors for high conversion water reactors

The paper shows that generating cross sections using three-dimensional geometry and application of axial discontinuity factors are essential requirements for obtaining accurate prediction of criticality and zone average reaction rates in highly heterogeneous RBWR-type systems using computer codes based on diffusion theory approximation. The same methodology as presented here will be used to generate discontinuity factors for each axial interface between fuel assembly zones to ensure preservation of reaction rates in each zone and global multiplication factor. The use of discontinuity factors and three-dimensional cross sections may allow for a coarser energy group structure which is desirable to simplify and speed up transient calculations.

Axial discontinuity factors for the nodal diffusion analysis of high conversion BWR cores

High conversion LWRs concepts typically rely on a heterogeneous core configuration, where fissile zones are interspersed with fertile blanket zones in order to achieve a high conversion ratio. Modeling such a heterogeneous structure of these cores represents a significant challenge to the conventional reactor analysis methods. It was recently suggested to overcome such difficulties, in particular, for the case of axially heterogeneous reduced moderation BWRs, by introducing an additional set of discontinuity factors in axial direction at the interfaces between fissile and fertile fuel assembly zones. However, none of the existing nodal diffusion core simulators have the capability of accounting for discontinuity of homogeneous nodal fluxes in axial direction since the fuel composition of conventional LWRs is much more axially uniform. In this work, we modified the nodal diffusion code DYN3D by introducing such a capability. The new version of the code was tested on a series of reduced moderation BWR cases with Th-U233 and U-Pu-MA fuel. The library of few-group homogenized cross sections and the data required for the calculation of discontinuity factors were generated using the Monte Carlo transport code Serpent. The results obtained with the modified version of DYN3D were compared with the reference Monte Carlo solutions and were found to be in good agreement. The current analysis demonstrates that high conversion LWRs can in principle be modeled using existing nodal diffusion core simulators. © 2013 Elsevier Ltd. All rights reserved.

Unconstrained monocular 3D human pose estimation by action detection and cross-modality regression forest

This work addresses the challenging problem of unconstrained 3D human pose estimation (HPE) from a novel perspective. Existing approaches struggle to operate in realistic applications, mainly due to their scene-dependent priors, such as background segmentation and multi-camera network, which restrict their use in unconstrained environments. We therfore present a framework which applies action detection and 2D pose estimation techniques to infer 3D poses in an unconstrained video. Action detection offers spatiotemporal priors to 3D human pose estimation by both recognising and localising actions in space-time. Instead of holistic features, e.g. silhouettes, we leverage the flexibility of deformable part model to detect 2D body parts as a feature to estimate 3D poses. A new unconstrained pose dataset has been collected to justify the feasibility of our method, which demonstrated promising results, significantly outperforming the relevant state-of-the-arts. © 2013 IEEE.