Model based learning of sigma points in unscented Kalman filtering

The unscented Kalman filter (UKF) is a widely used method in control and time series applications. The UKF suffers from arbitrary parameters necessary for sigma point placement, potentially causing it to perform poorly in nonlinear problems. We show how to treat sigma point placement in a UKF as a learning problem in a model based view. We demonstrate that learning to place the sigma points correctly from data can make sigma point collapse much less likely. Learning can result in a significant increase in predictive performance over default settings of the parameters in the UKF and other filters designed to avoid the problems of the UKF, such as the GP-ADF. At the same time, we maintain a lower computational complexity than the other methods. We call our method UKF-L. © 2011 Elsevier B.V.

TMVA, a novel GPIb-binding protein, significantly prevents platelet microthrombi formation and prolongs discordant cardiac xenograft survival

In xenotransplantation, donor endothelium is the first target of immunological attack. Activation of the endothelial cell by preformed natural antibodies leads to platelet binding via the interaction of the glycoprotein (GP) Ib and von Willebrand factor (vWF). TMVA is a novel GPIb-binding protein purified from the venom of Trimeresurus mucrosquamatus. In this study, the inhibitory effect of TMVA on platelet aggregation in rats and the effect on discordant guinea pig-to-rat cardiac xenograft survival were investigated. Three doses (8, 20 or 40 mug/kg) of TMVA were infused intravenously to 30 rats respectively. Platelet aggregation rate was assayed 0.5, 12, and 24 h after TMVA administration. Wister rats underwent guinea pig cardiac cervical heterotopic transplantation using single dosing of TMVA (20 mug/kg, i.v., 0.5 h before reperfusion). Additionally, levels of TXB2 and 6-keto-PGF(1alpha) within rejected graft tissues were determined by radioimmunoassay. Treatment with TMVA at a dose of 20 or 40 mug/kg resulted in complete inhibition of platelet aggregation 0.5 h after TMVA administration. Rats receiving guinea pig cardiac xenografts after TMVA therapy had significantly prolonged xenograft survival. Histologic and immunopathologic analysis of cardiac xenografts in TMVA treatment group showed no intragraft platelet microthrombi formation and fibrin deposition. Additionally, the ratio of 6-keto-PGF(1alpha) to TXB2 in TMVA treatment group was significantly higher than those in control group. We conclude that the use of this novel GPIb-binding protein was very effective in preventing platelet microthrombi formation and fibrin deposition in a guinea pig-to-rat model and resulted in prolongation of xenograft survival. The increased ratio of PGI(2)/TXA(2) in TMVA treatment group may protect xenografts from the endothelial cell activation and contribute to the prolongation of xenograft survival.

Additive Gaussian processes

We introduce a Gaussian process model of functions which are additive. An additive function is one which decomposes into a sum of low-dimensional functions, each depending on only a subset of the input variables. Additive GPs generalize both Generalized Additive Models, and the standard GP models which use squared-exponential kernels. Hyperparameter learning in this model can be seen as Bayesian Hierarchical Kernel Learning (HKL). We introduce an expressive but tractable parameterization of the kernel function, which allows efficient evaluation of all input interaction terms, whose number is exponential in the input dimension. The additional structure discoverable by this model results in increased interpretability, as well as state-of-the-art predictive power in regression tasks.

Gaussian Process training with input noise

In standard Gaussian Process regression input locations are assumed to be noise free. We present a simple yet effective GP model for training on input points corrupted by i.i.d. Gaussian noise. To make computations tractable we use a local linear expansion about each input point. This allows the input noise to be recast as output noise proportional to the squared gradient of the GP posterior mean. The input noise variances are inferred from the data as extra hyperparameters. They are trained alongside other hyperparameters by the usual method of maximisation of the marginal likelihood. Training uses an iterative scheme, which alternates between optimising the hyperparameters and calculating the posterior gradient. Analytic predictive moments can then be found for Gaussian distributed test points. We compare our model to others over a range of different regression problems and show that it improves over current methods.

State-space inference and learning with Gaussian processes

State-space inference and learning with Gaussian processes (GPs) is an unsolved problem. We propose a new, general methodology for inference and learning in nonlinear state-space models that are described probabilistically by non-parametric GP models. We apply the expectation maximization algorithm to iterate between inference in the latent state-space and learning the parameters of the underlying GP dynamics model. Copyright 2010 by the authors.

Robust filtering and smoothing with gaussian processes

We propose a principled algorithm for robust Bayesian filtering and smoothing in nonlinear stochastic dynamic systems when both the transition function and the measurement function are described by non-parametric Gaussian process (GP) models. GPs are gaining increasing importance in signal processing, machine learning, robotics, and control for representing unknown system functions by posterior probability distributions. This modern way of system identification is more robust than finding point estimates of a parametric function representation. Our principled filtering/smoothing approach for GP dynamic systems is based on analytic moment matching in the context of the forward-backward algorithm. Our numerical evaluations demonstrate the robustness of the proposed approach in situations where other state-of-the-art Gaussian filters and smoothers can fail. © 2011 IEEE.

Epigenetic remodeling of meiotic crossover frequency in Arabidopsis thaliana DNA methyltransferase mutants.

Meiosis is a specialized eukaryotic cell division that generates haploid gametes required for sexual reproduction. During meiosis, homologous chromosomes pair and undergo reciprocal genetic exchange, termed crossover (CO). Meiotic CO frequency varies along the physical length of chromosomes and is determined by hierarchical mechanisms, including epigenetic organization, for example methylation of the DNA and histones. Here we investigate the role of DNA methylation in determining patterns of CO frequency along Arabidopsis thaliana chromosomes. In A. thaliana the pericentromeric regions are repetitive, densely DNA methylated, and suppressed for both RNA polymerase-II transcription and CO frequency. DNA hypomethylated methyltransferase1 (met1) mutants show transcriptional reactivation of repetitive sequences in the pericentromeres, which we demonstrate is coupled to extensive remodeling of CO frequency. We observe elevated centromere-proximal COs in met1, coincident with pericentromeric decreases and distal increases. Importantly, total numbers of CO events are similar between wild type and met1, suggesting a role for interference and homeostasis in CO remodeling. To understand recombination distributions at a finer scale we generated CO frequency maps close to the telomere of chromosome 3 in wild type and demonstrate an elevated recombination topology in met1. Using a pollen-typing strategy we have identified an intergenic nucleosome-free CO hotspot 3a, and we demonstrate that it undergoes increased recombination activity in met1. We hypothesize that modulation of 3a activity is caused by CO remodeling driven by elevated centromeric COs. These data demonstrate how regional epigenetic organization can pattern recombination frequency along eukaryotic chromosomes.

Uncertainty management for on-line optimisation of a POMDP-based large-scale spoken dialogue system

The optimization of dialogue policies using reinforcement learning (RL) is now an accepted part of the state of the art in spoken dialogue systems (SDS). Yet, it is still the case that the commonly used training algorithms for SDS require a large number of dialogues and hence most systems still rely on artificial data generated by a user simulator. Optimization is therefore performed off-line before releasing the system to real users. Gaussian Processes (GP) for RL have recently been applied to dialogue systems. One advantage of GP is that they compute an explicit measure of uncertainty in the value function estimates computed during learning. In this paper, a class of novel learning strategies is described which use uncertainty to control exploration on-line. Comparisons between several exploration schemes show that significant improvements to learning speed can be obtained and that rapid and safe online optimisation is possible, even on a complex task. Copyright © 2011 ISCA.

Policy optimisation of POMDP-based dialogue systems without state space compression

The partially observable Markov decision process (POMDP) has been proposed as a dialogue model that enables automatic improvement of the dialogue policy and robustness to speech understanding errors. It requires, however, a large number of dialogues to train the dialogue policy. Gaussian processes (GP) have recently been applied to POMDP dialogue management optimisation showing an ability to substantially increase the speed of learning. Here, we investigate this further using the Bayesian Update of Dialogue State dialogue manager. We show that it is possible to apply Gaussian processes directly to the belief state, removing the need for a parametric policy representation. In addition, the resulting policy learns significantly faster while maintaining operational performance. © 2012 IEEE.

Composite electrospun gelatin fiber-alginate gel scaffolds for mechanically robust tissue engineered cornea.

A severe shortage of good quality donor cornea is now an international crisis in public health. Alternatives for donor tissue need to be urgently developed to meet the increasing demand for corneal transplantation. Hydrogels have been widely used as scaffolds for corneal tissue regeneration due to their large water content, similar to that of native tissue. However, these hydrogel scaffolds lack the fibrous structure that functions as a load-bearing component in the native tissue, resulting in poor mechanical performance. This work shows that mechanical properties of compliant hydrogels can be substantially enhanced with electrospun nanofiber reinforcement. Electrospun gelatin nanofibers were infiltrated with alginate hydrogels, yielding transparent fiber-reinforced hydrogels. Without prior crosslinking, electrospun gelatin nanofibers improved the tensile elastic modulus of the hydrogels from 78±19 kPa to 450±100 kPa. Stiffer hydrogels, with elastic modulus of 820±210 kPa, were obtained by crosslinking the gelatin fibers with carbodiimide hydrochloride in ethanol before the infiltration process, but at the expense of transparency. The developed fiber-reinforced hydrogels show great promise as mechanically robust scaffolds for corneal tissue engineering applications.

Pulsed-field magnetization of drilled bulk high-temperature superconductors: flux front propagation in the volume and on the surface

We present a method for characterizing the propagation of the magnetic flux in an artificially drilled bulk high-temperature superconductor (HTS) during a pulsed-field magnetization. As the magnetic pulse penetrates the cylindrical sample, the magnetic flux density is measured simultaneously in 16 holes by means of microcoils that are placed across the median plane, i.e. at an equal distance from the top and bottom surfaces, and close to the surface of the sample. We discuss the time evolution of the magnetic flux density in the holes during a pulse and measure the time taken by the external magnetic flux to reach each hole. Our data show that the flux front moves faster in the median plane than on the surface when penetrating the sample edge; it then proceeds faster along the surface than in the bulk as it penetrates the sample further. Once the pulse is over, the trapped flux density inside the central hole is found to be about twice as large in the median plane than on the surface. This ratio is confirmed by modelling.

Modification of the trapped field in bulk high-temperature superconductors as a result of the drilling of a pattern of artificial columnar holes

The trapped magnetic field is examined in bulk high-temperature superconductors that are artificially drilled along their c-axis. The influence of the hole pattern on the magnetization is studied and compared by means of numerical models and Hall probe mapping techniques. To this aim, we consider two bulk YBCO samples with a rectangular cross-section that are drilled each by six holes arranged either on a rectangular lattice (sample I) or on a centered rectangular lattice (sample II). For the numerical analysis, three different models are considered for calculating the trapped flux: (i), a two-dimensional (2D) Bean model neglecting demagnetizing effects and flux creep, (ii), a 2D finite-element model neglecting demagnetizing effects but incorporating magnetic relaxation in the form of an E-J power law, and, (iii), a 3D finite element analysis that takes into account both the finite height of the sample and flux creep effects. For the experimental analysis, the trapped magnetic flux density is measured above the sample surface by Hall probe mapping performed before and after the drilling process. The maximum trapped flux density in the drilled samples is found to be smaller than that in the plain samples. The smallest magnetization drop is found for sample II, with the centered rectangular lattice. This result is confirmed by the numerical models. In each sample, the relative drops that are calculated independently with the three different models are in good agreement. As observed experimentally, the magnetization drop calculated in the sample II is the smallest one and its relative value is comparable to the measured one. By contrast, the measured magnetization drop in sample (1) is much larger than that predicted by the simulations, most likely because of a change of the microstructure during the drilling process.

Measurement of the magnetic field inside the holes of a drilled bulk high-Tc superconductor

We use macroscopic holes drilled in a bulk YBCO superconductor to probe its magnetic properties in the volume of the sample. The sample is subjected to an AC magnetic flux with a density ranging from 30mT to 130mT and the flux in the superconductor is probed by miniature coils inserted in the holes. In a given hole, three different penetration regimes can be observed: (i) the shielded regime, where no magnetic flux threads the hole; (ii) the gradual penetration regime, where the waveform of the magnetic field has a clipped sine shape whose fundamental component scales with the applied field; and (iii) the flux concentration regime, where the waveform of the magnetic field is nearly a sine wave, with an amplitude exceeding that of the applied field by up to a factor of two. The distribution of the penetration regimes in the holes is compared with that of the magnetic flux density at the top and bottom surfaces of the sample, and is interpreted with the help of optical polarized light micrographs of these surfaces. We show that the measurement of the magnetic field inside the holes can be used as a local characterization of the bulk magnetic properties of the sample.

Magnetic properties of drilled bulk high-temperature superconductors filled with a ferromagnetic powder

It is shown that filling the holes of a drilled bulk high-temperature superconductor (HTS) with a soft ferromagnetic powder enhances its trapping properties. The magnetic properties of the trapped field magnet are characterized by Hall probe mapping and magnetization measurements. This analysis is completed by a numerical model based on a 3D finite-element method where the conductivity of the superconducting material is described by a power law while the permeability of the ferromagnetic material is fixed to a given value and is considered uniform. Numerical results support the experimental observations. In particular, they confirm the increase of trapped flux that is observed with Hall probe mapping after impregnation. © 2011 IOP Publishing Ltd.