Parallel geostatistics for sparse and dense datasets

Very large spatially-referenced datasets, for example, those derived from satellite-based sensors which sample across the globe or large monitoring networks of individual sensors, are becoming increasingly common and more widely available for use in environmental decision making. In large or dense sensor networks, huge quantities of data can be collected over small time periods. In many applications the generation of maps, or predictions at specific locations, from the data in (near) real-time is crucial. Geostatistical operations such as interpolation are vital in this map-generation process and in emergency situations, the resulting predictions need to be available almost instantly, so that decision makers can make informed decisions and define risk and evacuation zones. It is also helpful when analysing data in less time critical applications, for example when interacting directly with the data for exploratory analysis, that the algorithms are responsive within a reasonable time frame. Performing geostatistical analysis on such large spatial datasets can present a number of problems, particularly in the case where maximum likelihood. Although the storage requirements only scale linearly with the number of observations in the dataset, the computational complexity in terms of memory and speed, scale quadratically and cubically respectively. Most modern commodity hardware has at least 2 processor cores if not more. Other mechanisms for allowing parallel computation such as Grid based systems are also becoming increasingly commonly available. However, currently there seems to be little interest in exploiting this extra processing power within the context of geostatistics. In this paper we review the existing parallel approaches for geostatistics. By recognising that diffeerent natural parallelisms exist and can be exploited depending on whether the dataset is sparsely or densely sampled with respect to the range of variation, we introduce two contrasting novel implementations of parallel algorithms based on approximating the data likelihood extending the methods of Vecchia [1988] and Tresp [2000]. Using parallel maximum likelihood variogram estimation and parallel prediction algorithms we show that computational time can be significantly reduced. We demonstrate this with both sparsely sampled data and densely sampled data on a variety of architectures ranging from the common dual core processor, found in many modern desktop computers, to large multi-node super computers. To highlight the strengths and weaknesses of the diffeerent methods we employ synthetic data sets and go on to show how the methods allow maximum likelihood based inference on the exhaustive Walker Lake data set.

Numerical implementation of the Manakov-PMD equation with precomputed M(Ω) matrices

The Manakov-PMD equation can be integrated with the same numerical efficiency as the coarse-step method by using precomputed M(Ω) matrices, which entirely avoids the somewhat ad-hoc rescaling of coefficients necessary in the coarse-step method.

Inference and optimization of real edges on sparse graphs:A statistical physics perspective

Inference and optimization of real-value edge variables in sparse graphs are studied using the Bethe approximation and replica method of statistical physics. Equilibrium states of general energy functions involving a large set of real edge variables that interact at the network nodes are obtained in various cases. When applied to the representative problem of network resource allocation, efficient distributed algorithms are also devised. Scaling properties with respect to the network connectivity and the resource availability are found, and links to probabilistic Bayesian approximation methods are established. Different cost measures are considered and algorithmic solutions in the various cases are devised and examined numerically. Simulation results are in full agreement with the theory. © 2007 The American Physical Society.

Sparsely spread CDMA - A statistical mechanics-based analysis

Sparse code division multiple access (CDMA), a variation on the standard CDMA method in which the spreading (signature) matrix contains only a relatively small number of nonzero elements, is presented and analysed using methods of statistical physics. The analysis provides results on the performance of maximum likelihood decoding for sparse spreading codes in the large system limit. We present results for both cases of regular and irregular spreading matrices for the binary additive white Gaussian noise channel (BIAWGN) with a comparison to the canonical (dense) random spreading code. © 2007 IOP Publishing Ltd.

Modified hydrogel matrices in fibre optic sensors

Common problems encountered in clinical sensing are those of non-biocompatibility, and slow response time of the device. The latter, also applying to chemical sensors, is possibly due to a lack of understanding of polymer support or membrane properties and hence failure to optimise membranes chosen for specific sensor applications. Hydrogels can be described as polymers which swell in water. In addition to this, the presence of water in the polymer matrix offers some control of biocompatibility. They thus provide a medium through which rapid transport of a sensed species to an incorporated reagent could occur. This work considers the feasibility of such a system, leading to the design and construction of an optical sensor test bed. The development of suitable membrane systems and of suitable coating techniques in order to apply them to the fibre optics is described. Initial results obtained from hydrogel coatings implied that the refractive index change in the polymer matrix, due to a change in water content with pH is the major factor contributing to the sensor response. However the presence of the colourimetric reagent was also altering the output signal obtained. An analysis of factors contributing to the overall response, such as colour change and membrane composition were made on both the test bed, via optical response, and on whole membranes via measurement of water content change. The investigation of coatings with low equilibrium water contents, of less than 10% was carried out and in fact a clearer signal response from the test bed was noted. Again these membranes were suprisingly responding via refractive index change, with the reagent playing a primary role in obtaining a sensible or non-random response, although not in a colourimetric fashion. A photographic study of these coatings revealed some clues as to the physical nature of these coatings and hence partially explained this phenomenon. A study of the transport properties of the most successful membrane, on a coated wire electrode and also on the fibre optic test bed, in a series of test environments, indicated that the reagent was possibly acting as an ion exchanger and hence having a major influence on transport and therefore sensor characteristics.

Fast algorithms for automatic mapping with space-limited covariance functions

In this paper we discuss a fast Bayesian extension to kriging algorithms which has been used successfully for fast, automatic mapping in emergency conditions in the Spatial Interpolation Comparison 2004 (SIC2004) exercise. The application of kriging to automatic mapping raises several issues such as robustness, scalability, speed and parameter estimation. Various ad-hoc solutions have been proposed and used extensively but they lack a sound theoretical basis. In this paper we show how observations can be projected onto a representative subset of the data, without losing significant information. This allows the complexity of the algorithm to grow as O(n m 2), where n is the total number of observations and m is the size of the subset of the observations retained for prediction. The main contribution of this paper is to further extend this projective method through the application of space-limited covariance functions, which can be used as an alternative to the commonly used covariance models. In many real world applications the correlation between observations essentially vanishes beyond a certain separation distance. Thus it makes sense to use a covariance model that encompasses this belief since this leads to sparse covariance matrices for which optimised sparse matrix techniques can be used. In the presence of extreme values we show that space-limited covariance functions offer an additional benefit, they maintain the smoothness locally but at the same time lead to a more robust, and compact, global model. We show the performance of this technique coupled with the sparse extension to the kriging algorithm on synthetic data and outline a number of computational benefits such an approach brings. To test the relevance to automatic mapping we apply the method to the data used in a recent comparison of interpolation techniques (SIC2004) to map the levels of background ambient gamma radiation. © Springer-Verlag 2007.

Drug/Water interactions in hydrogel matrices

The primary aim of this research has been the investigation of the role of water structuring effects in the widely different extents of irritancy displayed by certain antibiotics. The compounds involved were members of the Lincomycin group of antibiotics. The aqueous solution behaviour of these co~pounds was studied using techniques such as vapour pressure osmometry end differential scanning calorimetry (D.S.C.). The effects of the antibiotics on water structure in hydrogel membrane preparations In which the equilibrium water content (E.W.C.) and constituent amounts of freezing and non-freezing water ware varied were also investigated using D.S.C. The permeability of water swollen hydrogel preparations to aqueous antibiotic solutions as well as other solutes were studied. A series of hydrogel preparations into which the antibiotics had been incorporated during polymerisation were developed and used in studies of the effects of the antibiotics end their water structure modifications on the permeation of a range of solutes.

The controlled release of macromolecules from macroporous hydrophyllic polymer matrices

This work has used novel polymer design and fabrication technology to generate bead form polymer based systems, with variable, yet controlled release properties, specifically for the delivery of macromolecules, essentially peptides of therapeutic interest. The work involved investigation of the potential interaction between matrix ultrastructural morphology, in vitro release kinetics, bioactivity and immunoreactivity of selected macromolecules with limited hydrolytic stability, delivered from controlled release vehicles. The underlying principle involved photo-polymerisation of the monomer, hydroxyethyl methacrylate, around frozen ice crystals, leading to the production of a macroporous hydrophilic matrix. Bead form matrices were fabricated in controllable size ranges in the region of 100µm - 3mm in diameter. The initial stages of the project involved the study of how variables, delivery speed of the monomer and stirring speed of the non solvent, affectedthe formation of macroporous bead form matrices. From this an optimal bench system for bead production was developed. Careful selection of monomer, solvents, crosslinking agent and polymerisation conditions led to a variable but controllable distribution of pore sizes (0.5 - 4µm). Release of surrogate macromolecules, bovine serum albumin and FITC-linked dextrans, enabled factors relating to the size and solubility of the macromolecule on the rate of release to be studied. Incorporation of bioactive macromolecules allowed retained bioactivity to be determined (glucose oxidase and interleukin-2), whilst the release of insulin enabled determination of both bioactivity (using rat epididymal fat pad) and immunoreactivity (RIA). The work carried out has led to the generation of macroporous bead form matrices, fabricated from a tissue biocompatible hydrogel, capable of the sustained, controlled release of biologically active peptides, with potential use in the pharmaceutical and agrochemical industries.

The durability of glass fibre reinforced cements made with new cementitious matrices

Widespread use of glass fibre reinforced cement (GRC) has been impeded by concerns over its durability. Three degradation mechanisms are proposed - fibre corrosion, Ca(OHh precipitation and matrix densification - although their relative importance is debated. Matrices with reduced alkalinities and Ca(OH)2 contents are being developed; the aim of this study was to investigate their hydration and interaction with alkali-resistant fibres to determine the factors controlling their long-term durability, and assess the relevancy of accelerated ageing. The matrices studied were: OPC/calcium-sulphoaluminate cement plus metakaolin (C); OPC plus metakaolin (M); blast-furnace slag cement plus a micro-silica based additive (D); and OPC (O). Accelerated ageing included hot water and cyclic regimes prior to tensile testing. Investigations included pore solution expression, XRD, DTA/TG, SEM and optical petrography. Bond strength was determined from crack spacings using microstructural parameters obtained from a unique image analysis technique. It was found that, for the new matrices - pore solution alkalinities were lower; Ca(OH)2 was absent or quickly consumed; different hydrates were formed at higher immersion temperatures; degradation under 65°C immersion was an order of magnitude slower, and no interfilamental Ca(OH)2 was observed .It was concluded that: fibre weakening caused by flaw growth was the primary degradation mechanism and was successfully modelled on stress corrosion/static fatigue principles. OPC inferiority was attributed partly to its higher alkalinity but chiefly to the growth of Ca(OH)2 aggravating the degradation; and hot water ageing although useful in model formulation and contrasting the matrices, changed the intrinsic nature of the composites rather than simply accelerating the degradation mechanisms.

Binary sparse nonnegative matrix factorization

This paper presents a fast part-based subspace selection algorithm, termed the binary sparse nonnegative matrix factorization (B-SNMF). Both the training process and the testing process of B-SNMF are much faster than those of binary principal component analysis (B-PCA). Besides, B-SNMF is more robust to occlusions in images. Experimental results on face images demonstrate the effectiveness and the efficiency of the proposed B-SNMF.

Gabor-based region covariance matrices for face recognition

This paper presents a new method for human face recognition by utilizing Gabor-based region covariance matrices as face descriptors. Both pixel locations and Gabor coefficients are employed to form the covariance matrices. Experimental results demonstrate the advantages of this proposed method.

Spectra of modular and small-world matrices

We compute spectra of symmetric random matrices describing graphs with general modular structure and arbitrary inter- and intra-module degree distributions, subject only to the constraint of finite mean connectivities. We also evaluate spectra of a certain class of small-world matrices generated from random graphs by introducing shortcuts via additional random connectivity components. Both adjacency matrices and the associated graph Laplacians are investigated. For the Laplacians, we find Lifshitz-type singular behaviour of the spectral density in a localized region of small |?| values. In the case of modular networks, we can identify contributions of local densities of state from individual modules. For small-world networks, we find that the introduction of short cuts can lead to the creation of satellite bands outside the central band of extended states, exhibiting only localized states in the band gaps. Results for the ensemble in the thermodynamic limit are in excellent agreement with those obtained via a cavity approach for large finite single instances, and with direct diagonalization results.

Sparse signal representation by adaptive non-uniform B-spline dictionaries on a compact interval

Non-uniform B-spline dictionaries on a compact interval are discussed in the context of sparse signal representation. For each given partition, dictionaries of B-spline functions for the corresponding spline space are built up by dividing the partition into subpartitions and joining together the bases for the concomitant subspaces. The resulting slightly redundant dictionaries are composed of B-spline functions of broader support than those corresponding to the B-spline basis for the identical space. Such dictionaries are meant to assist in the construction of adaptive sparse signal representation through a combination of stepwise optimal greedy techniques.

Precipitation casting of drug-loaded microporous PCL matrices:incorporation of progesterone by co-dissolution

Microporous, poly(ε-caprolactone) (PCL) matrices were loaded with progesterone by precipitation casting using co-solutions of PCL and progesterone in acetone. Progesterone loadings up to 32% w/w were readily achieved by increasing the drug content of the starting PCL solution. The kinetics of steroid release in PBS at 37°C over 10 days could be described effectively by a diffusional release model although the Korsmeyer-Peppas model indicated the involvement of multiple release phenomena. The diffusion rate constant (D) increased from 8 to 24 μg/mg matrix/day0.5 as the drug loading increased from 3.6 to 12.4% w/w. A total cumulative release of 75%-95% indicates the high efficiency of steroid delivery. Increasing the matrix density from 0.22 to 0.39 g/cm3, by increasing the starting PCL solution concentration, was less effective in changing drug release kinetics. Retention of anti-proliferative activity of released steroid was confirmed using cultures of breast cancer epithelial (MCF-7) cells. Progesterone released from PCL matrices into PBS at 37°C over 14 days retarded the growth of MCF-7 cells by a factor of at least 3.5 compared with progesterone-free controls. These findings recommend further investigation of precipitation-cast PCL matrices for delivery of bioactive molecules such as anti-proliferative agents from implanted, inserted or topical devices.