Copula-based Kernel Dependency Measures

The paper presents a new copula based method for measuring dependence between random variables. Our approach extends the Maximum Mean Discrepancy to the copula of the joint distribution. We prove that this approach has several advantageous properties. Similarly to Shannon mutual information, the proposed dependence measure is invariant to any strictly increasing transformation of the marginal variables. This is important in many applications, for example in feature selection. The estimator is consistent, robust to outliers, and uses rank statistics only. We derive upper bounds on the convergence rate and propose independence tests too. We illustrate the theoretical contributions through a series of experiments in feature selection and low-dimensional embedding of distributions.

Gamma-ray spectroscopy at TRIUMF-ISAC

The 8π spectrometer at TRIUMF-ISAC consists of 20 Compton-suppressed germanium detectors and various auxiliary devices. The Ge array, once used for studies of nuclei at high angular momentum, has been transformed into the world's most powerful device dedicated to radioactive-decay studies. Many improvements in the spectrometer have been made, including a high-throughput data acquisition system, installation of a moving tape collector, incorporation of an array of 20 plastic scintillators for β-particle tagging, 5 Si(Li) detectors for conversion electrons, and 10 BaF2 detectors for fast-lifetime measurements. Experiments can be performed where data from all detectors are collected simultaneously, resulting in a very detailed view of the nucleus through radioactive decay. A number of experimental programmes have been launched that take advantage of the versatility of the spectrometer, and the intense beams available at TRIUMF-ISAC. © 2006 American Institute of Physics.

Characterising the neutron and gamma responses of the NPL liquid scintillation detectors.

The absolute responses of the NPL liquid scintillation spectrometers to monoenergetic neutrons and gammas were measured at various energies in the ranges 1.2 - 17 MeV approximately for neutrons and 0.28 - 1.8 MeV for gammas. Additional measurements of the proton light output function were also carried out. Calculated responses were then obtained for the larger detector using the programs NRESP7 and PHRESP, and compared with the absolute measurements. Finally, response matrices for this detector were generated using responses calculated at closely spaced energies.

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.

Kernel conditional quantile estimation via reduction revisited

Quantile regression refers to the process of estimating the quantiles of a conditional distribution and has many important applications within econometrics and data mining, among other domains. In this paper, we show how to estimate these conditional quantile functions within a Bayes risk minimization framework using a Gaussian process prior. The resulting non-parametric probabilistic model is easy to implement and allows non-crossing quantile functions to be enforced. Moreover, it can directly be used in combination with tools and extensions of standard Gaussian Processes such as principled hyperparameter estimation, sparsification, and quantile regression with input-dependent noise rates. No existing approach enjoys all of these desirable properties. Experiments on benchmark datasets show that our method is competitive with state-of-the-art approaches. © 2009 IEEE.

Determinantal Clustering Processes - A Nonparametric Bayesian Approach to Kernel Based Semi-Supervised Clustering

Semi-supervised clustering is the task of clustering data points into clusters where only a fraction of the points are labelled. The true number of clusters in the data is often unknown and most models require this parameter as an input. Dirichlet process mixture models are appealing as they can infer the number of clusters from the data. However, these models do not deal with high dimensional data well and can encounter difficulties in inference. We present a novel nonparameteric Bayesian kernel based method to cluster data points without the need to prespecify the number of clusters or to model complicated densities from which data points are assumed to be generated from. The key insight is to use determinants of submatrices of a kernel matrix as a measure of how close together a set of points are. We explore some theoretical properties of the model and derive a natural Gibbs based algorithm with MCMC hyperparameter learning. The model is implemented on a variety of synthetic and real world data sets.

The search for the IFN-gamma receptor in fish: Functional and expression analysis of putative binding and signalling chains in rainbow trout Oncorhynchus mykiss

Interferons (IFNs), consisting of three major subfamilies, type I, type II (gamma) and type III (lambda) IFN, activate vertebrate antiviral defences once bound to their receptors. The three IFN subfamilies bind to different receptors, IFNAR1 and IFNAR2 for type I IFNs, IFN gamma R1 and IFN gamma R2 for type II IFN, and IL-28R1 and IL-10R2 for type III IFNs. In fish, although many types I and II IFN genes have been cloned, little is known about their receptors. In this report, two putative IFN-gamma receptor chains were identified and sequenced in rainbow trout (Oncorhynchus mykiss), and found to have many common characteristics with mammalian type II IFN receptor family members. The presented gene synteny analysis, phylogenetic tree analysis and ligand binding analysis all suggest that these molecules are the authentic IFN gamma Rs in fish. They are widely expressed in tissues, with IFN gamma R1 typically more highly expressed than IFN gamma R2. Using the trout RTG-2 cell line it was possible to show that the individual chains could be differentially modulated, with rIFN-gamma and rIL-1 beta down regulating IFN gamma R1 expression but up regulating IFN gamma R2 expression. Overexpression of the two receptor chains in RTG-2 cells revealed that the level of IFN gamma R2 transcript was crucial for responsiveness to rIFN-gamma, in terms of inducing gamma IP expression. Transfection experiments showed that the two putative receptors specifically bound to rIFN-gamma. These findings are discussed in the context of how the IFN gamma R may bind IFN-gamma in fish and the importance of the individual receptor chains to signal transduction. (c) 2009 Elsevier Ltd. All rights reserved.

Comparative study and expression analysis of the interferon gamma gene locus cytokines in Xenopus tropicalis

Using bioinformatics approach, the genome locus containing interleukin (IL)-22, IL-26, and interferon gamma (IFN-gamma) genes has been identified in the amphibian, Xenopus tropicalis. Like that in other vertebrates such as fish, birds, and mammals, the Xenopus IL-22, IL-26, and IFN-gamma are clustered in the same chromosome and the adjacent genes are conserved. The genomic structures of the Xenopus IL-22, IL-26, and IFN-gamma gene were identical to that of their mammalian counterparts. The Xenopus IL-22 and IL-26 genes contained five exons and four introns while the Xenopus IFN-gamma gene consisted of four exons and three introns. The Xenopus IL-22, IL-26, and IFN-gamma share 14.1-41.6%, 14.6-31.2%, and 23.7-36.5% identity to their counterparts in other species, respectively. Reverse-transcription polymerase chain reaction (PCR) and real-time quantitative PCR analyses revealed that the expression of IL-22, IL-26, and IFN-gamma genes was significantly upregulated after simulation with bacterial polyliposaccharide and/or synthetic double-stranded poly(I:C), suggesting these cytokines like those in other vertebrates play an important role in regulating immune response in Xenopus.

The Random Forest Kernel and other kernels for big data from random partitions

We present Random Partition Kernels, a new class of kernels derived by demonstrating a natural connection between random partitions of objects and kernels between those objects. We show how the construction can be used to create kernels from methods that would not normally be viewed as random partitions, such as Random Forest. To demonstrate the potential of this method, we propose two new kernels, the Random Forest Kernel and the Fast Cluster Kernel, and show that these kernels consistently outperform standard kernels on problems involving real-world datasets. Finally, we show how the form of these kernels lend themselves to a natural approximation that is appropriate for certain big data problems, allowing $O(N)$ inference in methods such as Gaussian Processes, Support Vector Machines and Kernel PCA.

Molecular cloning and characterization of common carp (Cyprinus carpio L.) TCR gamma and CD3 gamma/delta chains

Partial cDNA sequences of TCR gamma and CD3 gamma/delta were isolated from the thymus of common carp (Cyprinus carpio L.) by the method of suppression subtractive hybridization (SSH). Subsequently the full length cDNAs of carp TCR gamma and CD3 gamma/delta were obtained by means of 3' RACE and 5' RACE, respectively. The full length of carp TCR gamma chain is 1368 bp and encodes 326 amino acids including a signal peptide region of 19 amino acids and a transmembrane region of 23 amino acids at the C-terminal region from aa 291 to 313. The V region of carp TCR gamma contains 109 amino acids, the core motif FGXG in J segment was also found in carp TCR gamma. The C region of carp TCR gamma contains the characteristic CX6PX6WX45C motif. The CP region of carp TCR C gamma contains 37 amino acids. The full length of carp CD3 gamma/delta is 790 bp and encodes 175 amino acids including a signal peptide region of 17 amino acids and a transmembrane region of 23 amino acids from aa 93 to 115. Similar to other known CD3 gamma/delta s, four cysteine residues in the extracellular domain and an immunoreceptor tyrosine-based activation motif ITAM (YxxL/Ix6-8YxxL/I) in the intracellular domain are also included in carp CD3 gamma/delta. Differing from other known CD3 gamma/delta s, carp CD3 gamma/delta tacks the CXXCXE motif in the extracellular domain. RTPCR analysis demonstrated that the expression of TCR gamma gene was mainly in the thymus and gill of 6-month carp, but in 18-month carp, TCR gamma gene was detected in all the examined tissues. The expression of CD3 gamma/delta gene was detected in all examined tissues of 6 and 18-month carp; among them, the highest expression level was in the thymus of 6-month carp. In situ hybridization showed that CD3 gamma/delta-expressing cells were widely distributed in the head kidney, spleen and kidney of carp, whereas in the thymus, they were densely distributed in the lymphoid outer zone and scattered in the epithelioid inner zone. (c) 2007 Published by Etsevier Ltd.

n-type Ge-SiGe quantum cascade structure utilizing quantum wells for electrons in the L and Gamma valleys

In this letter, we propose an n-type vertical transition bound-to-continuum Ge-SiGe quantum cascade structure utilizing electronic quantum wells in the L and F valleys of the Ge layers. The optical transition levels are located in the quantum wells in the L valley. Under a bias of 80 kV/cm, the carriers in the lower level are extracted by miniband transport and L - Gamma tunneling into the subband in the Gamma well of the next period. And then the electrons are injected into the upper level by ultrafast intervalley scattering, which not only effectively increases the tunneling rate and suppresses the thermal backfilling of electrons, but also enhances the injection efficiency of the upper level. The performance of the laser is discussed.

Characterization of stress induced in SOS and Si/gamma-Al2O3/Si heteroepitaxial thin films by Raman spectroscopy

Raman spectroscopy technique has been performed to investigate the stress induced in as-grown silicon-on-sapphire (SOS), solid-phase-epitaxy (SPE) re-grown SOS, and Si/gamma-Al2O3/Si double-heteroepitaxial thin films. It was demonstrated that the residual stress in SOS film, arising from mismatch and difference of thermal expansion coefficient between silicon and sapphire, was reduced efficiently by SPE process, and that the stress in Si/gamma-Al2O3/Si thin film is much smaller than that of as-grown SOS and SPE upgraded SOS films. The stress decrease for double heteroepitaxial film Si/gamma-Al2O3/Si mainly arises from the smaller lattice mismatching of 2.4% between silicon top layer and the gamma-Al2O3/Si epitaxiial composite substrate, comparing with the large lattice mismatch of 13% for SOS films. It indicated that gamma-Al2O3/Si as a silicon-based epitaxial substrate benefits for reducing the residual stress for further growth of silicon layer, compared with on bulk sapphire substrate. (c) 2005 Elsevier B.V. All rights reserved.

An n-type SiGe/Ge QC structure utilizing the deep Ge quantum well for electron at the Gamma point

In this paper, an n-type Si1-xGex/Ge (x >= 0.85) quantum cascade (QC) structure utilizing a deep Ge quantum well for electrons at the Gamma point is proposed. Based on linear interpolation, a conduction band offset at the Gamma point in a Si1-xGex/Ge ( x >= 0.85) heterostructure is presented, which is suitable for designing a QC laser. This approach has the advantages of a large conduction band offset at the Gamma point, a low lattice mismatch between the Si1-xGex/Ge ( x >= 0.85) active layers and the Si1-yGey ( y > x) virtual substrate, a small electron effective mass in the Gamma band, simple conduction energy band structures and a simple phonon scattering mechanism in the Ge quantum well. The theory predicts that if high-energy electrons are continuously injected into the Gamma band, a quasi-equilibrium distribution of electrons between the Gamma and L bands can be reached and held, i.e., electrons with a certain density will be kept in the Gamma band. This result is supported by the intervalley scattering experiments. In n-type Si1-xGex/Ge ( x >= 0.85) QC structures, population inversion between the laser's upper and lower levels is demonstrated.