Topographic mappings and feed-forward neural networks

This thesis is a study of the generation of topographic mappings - dimension reducing transformations of data that preserve some element of geometric structure - with feed-forward neural networks. As an alternative to established methods, a transformational variant of Sammon's method is proposed, where the projection is effected by a radial basis function neural network. This approach is related to the statistical field of multidimensional scaling, and from that the concept of a 'subjective metric' is defined, which permits the exploitation of additional prior knowledge concerning the data in the mapping process. This then enables the generation of more appropriate feature spaces for the purposes of enhanced visualisation or subsequent classification. A comparison with established methods for feature extraction is given for data taken from the 1992 Research Assessment Exercise for higher educational institutions in the United Kingdom. This is a difficult high-dimensional dataset, and illustrates well the benefit of the new topographic technique. A generalisation of the proposed model is considered for implementation of the classical multidimensional scaling (¸mds}) routine. This is related to Oja's principal subspace neural network, whose learning rule is shown to descend the error surface of the proposed ¸mds model. Some of the technical issues concerning the design and training of topographic neural networks are investigated. It is shown that neural network models can be less sensitive to entrapment in the sub-optimal global minima that badly affect the standard Sammon algorithm, and tend to exhibit good generalisation as a result of implicit weight decay in the training process. It is further argued that for ideal structure retention, the network transformation should be perfectly smooth for all inter-data directions in input space. Finally, there is a critique of optimisation techniques for topographic mappings, and a new training algorithm is proposed. A convergence proof is given, and the method is shown to produce lower-error mappings more rapidly than previous algorithms.

A Necessary and Sufficient Condition for the Existence of an (n,r)-arc in PG(2,q) and Its Applications

ACM Computing Classification System (1998): E.4.

Low valence cation doping of bulk Cr2O3: Charge compensation and oxygen vacancy formation

The different oxidation states of chromium allow its bulk oxide form to be reducible, facilitating the oxygen vacancy formation process, which is a key property in applications such as catalysis. Similar to other useful oxides such as TiO2, and CeO2, the effect of substitutional metal dopants in bulk Cr2O3 and its effect on the electronic structure and oxygen vacancy formation are of interest, particularly in enhancing the latter. In this paper, density functional theory (DFT) calculations with a Hubbard + U correction (DFT+U) applied to the Cr 3d and O 2p states, are carried out on pure and metal-doped bulk Cr2O3 to examine the effect of doping on the electronic and geometric structure. The role of dopants in enhancing the reducibility of Cr2O3 is examined to promote oxygen vacancy formation. The dopants are Mg, Cu, Ni, and Zn, which have a formal +2 oxidation state in their bulk oxides. Given this difference in host and, dopant oxidation states, we show that to predict the correct ground state two metal dopants charge compensated with an oxygen vacancy are required. The second oxygen atom removed is termed "the active" oxygen vacancy and it is the energy required to remove this atom that is related to the reduction process. In all cases, we find that substitutional doping improves the oxygen vacancy formation of bulk Cr2O3 by lowering the energy cost.

Analysis and modelling of immune cell behaviour in lymph nodes based on multi-photon imaging

This thesis presents quantitative studies of T cell and dendritic cell (DC) behaviour in mouse lymph nodes (LNs) in the naive state and following immunisation. These processes are of importance and interest in basic immunology, and better understanding could improve both diagnostic capacity and therapeutic manipulations, potentially helping in producing more effective vaccines or developing treatments for autoimmune diseases. The problem is also interesting conceptually as it is relevant to other fields where 3D movement of objects is tracked with a discrete scanning interval. A general immunology introduction is presented in chapter 1. In chapter 2, I apply quantitative methods to multi-photon imaging data to measure how T cells and DCs are spatially arranged in LNs. This has been previously studied to describe differences between the naive and immunised state and as an indicator of the magnitude of the immune response in LNs, but previous analyses have been generally descriptive. The quantitative analysis shows that some of the previous conclusions may have been premature. In chapter 3, I use Bayesian state-space models to test some hypotheses about the mode of T cell search for DCs. A two-state mode of movement where T cells can be classified as either interacting to a DC or freely migrating is supported over a model where T cells would home in on DCs at distance through for example the action of chemokines. In chapter 4, I study whether T cell migration is linked to the geometric structure of the fibroblast reticular network (FRC). I find support for the hypothesis that the movement is constrained to the fibroblast reticular cell (FRC) network over an alternative 'random walk with persistence time' model where cells would move randomly, with a short-term persistence driven by a hypothetical T cell intrinsic 'clock'. I also present unexpected results on the FRC network geometry. Finally, a quantitative method is presented for addressing some measurement biases inherent to multi-photon imaging. In all three chapters, novel findings are made, and the methods developed have the potential for further use to address important problems in the field. In chapter 5, I present a summary and synthesis of results from chapters 3-4 and a more speculative discussion of these results and potential future directions.

Neurogeometry of reaching

The purpose of the thesis is to develop a model for the functional behaviour of neurons in the primary motor cortex (M1) responsible for arm reaching movements. From Georgopoulos neurophysiological data, we provide a first bundle structure compatible with the hypercolumnar organization and with the position-direction selectivity of motor cortical cells. We then extend this model to encode the direction of arm movement which varies in time, as experimentally measured by Hatsopoulos by introducing the notion of movement fragments. We provide a sub-Riemannian model which describes the time-dependent directional selectivity of cells though integral curves of the geometric structure we set up. The sub-Riemannian distance we define allows to implement a grouping algorithm able to detect a set of hand motor trajectories. These paths, identified by using a kernel defined in terms of kinematic variables, are compatible with the motor primitives obtained from neurophysiological results by spectral analysis applied directly on cortical variables. In a second part of the work, we propose geodesics in this space as an alternative model of models for arm movement trajectories. We define a special class of curves, called admissible, on which to study the geodesics problem: we provide a connectivity property in terms of admissible paths and the existence of normal length minimizers. Admissible geodesics are used as a model of reaching paths, finding a first validation through Flash and Hogan minimizing trajectories.

Variation in mandible shape in Thrichomys apereoides (Mammalia: Rodentia): Geometric analysis of a complex morphological structure

The model of development and evolution of complex morphological structures conceived by Atchley and Hall in 1991 (Biol. Rev. 66:101-157), which establishes that changes at the macroscopic, morphogenetic level can be statistically detected as variation in skeletal units at distinct scales, was applied in combination with the formalism of geometric morphometrics to study variation in mandible shape among populations of the rodent species Thrichomys apereoides. The thin-plate spline technique produced geometric descriptors of shape derived from anatomical landmarks in the mandible, which we used with graphical and inferential approaches to partition the contribution of global and localized components to the observed differentiation in mandible shape. A major pattern of morphological differentiation in T. apereoides is attributable to localized components of shape at smaller geometric scales associated with specific morphogenetic units of the mandible. On the other hand, a clinal trend of variation is associated primarily with localized components of shape at larger geometric scales. Morphogenetic mechanisms assumed to be operating to produce the observed differentiation in the specific units of the mandible include mesenchymal condensation differentiation, muscle hypertrophy, and tooth growth. Perspectives for the application of models of morphological evolution and geometric morphometrics to morphologically based systematic biology are considered.

A soft-tetramer model for diblock copolymer melts: structure formation and geometric characterization

The ability of block copolymers to spontaneously self-assemble into a variety of ordered nano-structures not only makes them a scientifically interesting system for the investigation of order-disorder phase transitions, but also offers a wide range of nano-technological applications. The architecture of a diblock is the most simple among the block copolymer systems, hence it is often used as a model system in both experiment and theory. We introduce a new soft-tetramer model for efficient computer simulations of diblock copolymer melts. The instantaneous non-spherical shape of polymer chains in molten state is incorporated by modeling each of the two blocks as two soft spheres. The interactions between the spheres are modeled in a way that the diblock melt tends to microphase separate with decreasing temperature. Using Monte Carlo simulations, we determine the equilibrium structures at variable values of the two relevant control parameters, the diblock composition and the incompatibility of unlike components. The simplicity of the model allows us to scan the control parameter space in a completeness that has not been reached in previous molecular simulations.The resulting phase diagram shows clear similarities with the phase diagram found in experiments. Moreover, we show that structural details of block copolymer chains can be reproduced by our simple model.We develop a novel method for the identification of the observed diblock copolymer mesophases that formalizes the usual approach of direct visual observation,using the characteristic geometry of the structures. A cluster analysis algorithm is used to determine clusters of each component of the diblock, and the number and shape of the clusters can be used to determine the mesophase.We also employ methods from integral geometry for the identification of mesophases and compare their usefulness to the cluster analysis approach.To probe the properties of our model in confinement, we perform molecular dynamics simulations of atomistic polyethylene melts confined between graphite surfaces. The results from these simulations are used as an input for an iterative coarse-graining procedure that yields a surface interaction potential for the soft-tetramer model. Using the interaction potential derived in that way, we perform an initial study on the behavior of the soft-tetramer model in confinement. Comparing with experimental studies, we find that our model can reflect basic features of confined diblock copolymer melts.

Technologies for Anatomical and Geometric Characterization of the Corneal Structure and Anterior Segment: A Review

Corneal and anterior segment imaging techniques have become a crucial tool in the clinical practice of ophthalmology, with a great variety of applications, such as corneal curvature and pachymetric analysis, detection of ectatic corneal conditions, anatomical study of the anterior segment prior to phakic intraocular lens implantation, or densitometric analysis of the crystalline lens. From the Placido-based systems that allow only a characterization of the geometry of the anterior corneal surface to the Scheimpflug photography-based systems that provide a characterization of the cornea, anterior chamber, and crystalline lens, there is a great variety of devices with the capability of analyzing different anatomical parameters with very high precision. To date, Scheimpflug photography-based systems are the devices providing the more complete analysis of the anterior segment in a non-invasive way. More developments are required in anterior segment imaging technologies in order to improve the analysis of the crystalline lens structure as well as the ocular structures behind the iris in a non-invasive way when the pupil is not dilated.

The Hierarchical Structure of the Firm: A Geometric Perspective

This paper incorporates hierarchical structure into the neoclassical theory of the firm. Firms are hierarchical in two respects: the organization of workers in production and the wage structure. The firm’s hierarchy is represented as the sector of a circle, where the radius represents the hierarchy’s height, the width of the sector represents the breadth of the hierarchy at a given height, and the angle of the sector represents span of control for any given supervisor. A perfectly competitive firm then chooses height and width, as well as capital inputs, in order to maximize profit. We analyze the short run and long run impact of changes in scale economies, input substitutability and input and output prices on the firm’s hierarchical structure. We find that the firm unambiguously becomes more hierarchical as the specialization of its workers increases or as its output price increases relative to input prices. The effect of changes in scale economies is contingent on the output price. The model also brings forth an analysis of wage inequality within the firm, which is found to be independent of technological considerations, and only depends on the firm’s wage schedule.

On The Consistency Of Monte Carlo Track Structure Dna Damage Simulations.

Monte Carlo track structures (MCTS) simulations have been recognized as useful tools for radiobiological modeling. However, the authors noticed several issues regarding the consistency of reported data. Therefore, in this work, they analyze the impact of various user defined parameters on simulated direct DNA damage yields. In addition, they draw attention to discrepancies in published literature in DNA strand break (SB) yields and selected methodologies. The MCTS code Geant4-DNA was used to compare radial dose profiles in a nanometer-scale region of interest (ROI) for photon sources of varying sizes and energies. Then, electron tracks of 0.28 keV-220 keV were superimposed on a geometric DNA model composed of 2.7 × 10(6) nucleosomes, and SBs were simulated according to four definitions based on energy deposits or energy transfers in DNA strand targets compared to a threshold energy ETH. The SB frequencies and complexities in nucleosomes as a function of incident electron energies were obtained. SBs were classified into higher order clusters such as single and double strand breaks (SSBs and DSBs) based on inter-SB distances and on the number of affected strands. Comparisons of different nonuniform dose distributions lacking charged particle equilibrium may lead to erroneous conclusions regarding the effect of energy on relative biological effectiveness. The energy transfer-based SB definitions give similar SB yields as the one based on energy deposit when ETH ≈ 10.79 eV, but deviate significantly for higher ETH values. Between 30 and 40 nucleosomes/Gy show at least one SB in the ROI. The number of nucleosomes that present a complex damage pattern of more than 2 SBs and the degree of complexity of the damage in these nucleosomes diminish as the incident electron energy increases. DNA damage classification into SSB and DSB is highly dependent on the definitions of these higher order structures and their implementations. The authors' show that, for the four studied models, different yields are expected by up to 54% for SSBs and by up to 32% for DSBs, as a function of the incident electrons energy and of the models being compared. MCTS simulations allow to compare direct DNA damage types and complexities induced by ionizing radiation. However, simulation results depend to a large degree on user-defined parameters, definitions, and algorithms such as: DNA model, dose distribution, SB definition, and the DNA damage clustering algorithm. These interdependencies should be well controlled during the simulations and explicitly reported when comparing results to experiments or calculations.

Determining forest structural attributes using an inverted geometric-optical model in mixed eucalypt forests, Southeast Queensland, Australia

The Montreal Process indicators are intended to provide a common framework for assessing and reviewing progress toward sustainable forest management. The potential of a combined geometrical-optical/spectral mixture analysis model was assessed for mapping the Montreal Process age class and successional age indicators at a regional scale using Landsat Thematic data. The project location is an area of eucalyptus forest in Emu Creek State Forest, Southeast Queensland, Australia. A quantitative model relating the spectral reflectance of a forest to the illumination geometry, slope, and aspect of the terrain surface and the size, shape, and density, and canopy size. Inversion of this model necessitated the use of spectral mixture analysis to recover subpixel information on the fractional extent of ground scene elements (such as sunlit canopy, shaded canopy, sunlit background, and shaded background). Results obtained fron a sensitivity analysis allowed improved allocation of resources to maximize the predictive accuracy of the model. It was found that modeled estimates of crown cover projection, canopy size, and tree densities had significant agreement with field and air photo-interpreted estimates. However, the accuracy of the successional stage classification was limited. The results obtained highlight the potential for future integration of high and moderate spatial resolution-imaging sensors for monitoring forest structure and condition. (C) Elsevier Science Inc., 2000.

Integrating high and moderate spatial resolution image data to estimate forest age structure

The collection of spatial information to quantify changes to the state and condition of the environment is a fundamental component of conservation or sustainable utilization of tropical and subtropical forests, Age is an important structural attribute of old-growth forests influencing biological diversity in Australia eucalypt forests. Aerial photograph interpretation has traditionally been used for mapping the age and structure of forest stands. However this method is subjective and is not able to accurately capture fine to landscape scale variation necessary for ecological studies. Identification and mapping of fine to landscape scale vegetative structural attributes will allow the compilation of information associated with Montreal Process indicators lb and ld, which seek to determine linkages between age structure and the diversity and abundance of forest fauna populations. This project integrated measurements of structural attributes derived from a canopy-height elevation model with results from a geometrical-optical/spectral mixture analysis model to map forest age structure at a landscape scale. The availability of multiple-scale data allows the transfer of high-resolution attributes to landscape scale monitoring. Multispectral image data were obtained from a DMSV (Digital Multi-Spectral Video) sensor over St Mary's State Forest in Southeast Queensland, Australia. Local scene variance levels for different forest tapes calculated from the DMSV data were used to optimize the tree density and canopy size output in a geometric-optical model applied to a Landsat Thematic Mapper (TU) data set. Airborne laser scanner data obtained over the project area were used to calibrate a digital filter to extract tree heights from a digital elevation model that was derived from scanned colour stereopairs. The modelled estimates of tree height, crown size, and tree density were used to produce a decision-tree classification of forest successional stage at a landscape scale. The results obtained (72% accuracy), were limited in validation, but demonstrate potential for using the multi-scale methodology to provide spatial information for forestry policy objectives (ie., monitoring forest age structure).

Design of a laminated composite multi-cel structure subjected to torsion

Laminate composite multi-cell structures have to support both axial and shear stresses when sustaining variable twist. Thus the properties and design of the laminate may not be the most adequate at all cross-sections to support the torsion imposed on the cells. In this work, the effect of some material and geometric parameters on the optimal mechanical behaviour of a multi-cell composite laminate structure is studied when torsion is present. A particle swarm optimization technique is used to maximize the multi-cell structure torsion constant that can be used to obtain the angle of twist of the composite laminate profile.

Vertex component analysis: a geometric-based approach to unmix hyperspectral data

Hyperspectral remote sensing exploits the electromagnetic scattering patterns of the different materials at specific wavelengths [2, 3]. Hyperspectral sensors have been developed to sample the scattered portion of the electromagnetic spectrum extending from the visible region through the near-infrared and mid-infrared, in hundreds of narrow contiguous bands [4, 5]. The number and variety of potential civilian and military applications of hyperspectral remote sensing is enormous [6, 7]. Very often, the resolution cell corresponding to a single pixel in an image contains several substances (endmembers) [4]. In this situation, the scattered energy is a mixing of the endmember spectra. A challenging task underlying many hyperspectral imagery applications is then decomposing a mixed pixel into a collection of reflectance spectra, called endmember signatures, and the corresponding abundance fractions [8–10]. Depending on the mixing scales at each pixel, the observed mixture is either linear or nonlinear [11, 12]. Linear mixing model holds approximately when the mixing scale is macroscopic [13] and there is negligible interaction among distinct endmembers [3, 14]. If, however, the mixing scale is microscopic (or intimate mixtures) [15, 16] and the incident solar radiation is scattered by the scene through multiple bounces involving several endmembers [17], the linear model is no longer accurate. Linear spectral unmixing has been intensively researched in the last years [9, 10, 12, 18–21]. It considers that a mixed pixel is a linear combination of endmember signatures weighted by the correspondent abundance fractions. Under this model, and assuming that the number of substances and their reflectance spectra are known, hyperspectral unmixing is a linear problem for which many solutions have been proposed (e.g., maximum likelihood estimation [8], spectral signature matching [22], spectral angle mapper [23], subspace projection methods [24,25], and constrained least squares [26]). In most cases, the number of substances and their reflectances are not known and, then, hyperspectral unmixing falls into the class of blind source separation problems [27]. Independent component analysis (ICA) has recently been proposed as a tool to blindly unmix hyperspectral data [28–31]. ICA is based on the assumption of mutually independent sources (abundance fractions), which is not the case of hyperspectral data, since the sum of abundance fractions is constant, implying statistical dependence among them. This dependence compromises ICA applicability to hyperspectral images as shown in Refs. [21, 32]. In fact, ICA finds the endmember signatures by multiplying the spectral vectors with an unmixing matrix, which minimizes the mutual information among sources. If sources are independent, ICA provides the correct unmixing, since the minimum of the mutual information is obtained only when sources are independent. This is no longer true for dependent abundance fractions. Nevertheless, some endmembers may be approximately unmixed. These aspects are addressed in Ref. [33]. Under the linear mixing model, the observations from a scene are in a simplex whose vertices correspond to the endmembers. Several approaches [34–36] have exploited this geometric feature of hyperspectral mixtures [35]. Minimum volume transform (MVT) algorithm [36] determines the simplex of minimum volume containing the data. The method presented in Ref. [37] is also of MVT type but, by introducing the notion of bundles, it takes into account the endmember variability usually present in hyperspectral mixtures. The MVT type approaches are complex from the computational point of view. Usually, these algorithms find in the first place the convex hull defined by the observed data and then fit a minimum volume simplex to it. For example, the gift wrapping algorithm [38] computes the convex hull of n data points in a d-dimensional space with a computational complexity of O(nbd=2cþ1), where bxc is the highest integer lower or equal than x and n is the number of samples. The complexity of the method presented in Ref. [37] is even higher, since the temperature of the simulated annealing algorithm used shall follow a log( ) law [39] to assure convergence (in probability) to the desired solution. Aiming at a lower computational complexity, some algorithms such as the pixel purity index (PPI) [35] and the N-FINDR [40] still find the minimum volume simplex containing the data cloud, but they assume the presence of at least one pure pixel of each endmember in the data. This is a strong requisite that may not hold in some data sets. In any case, these algorithms find the set of most pure pixels in the data. PPI algorithm uses the minimum noise fraction (MNF) [41] as a preprocessing step to reduce dimensionality and to improve the signal-to-noise ratio (SNR). The algorithm then projects every spectral vector onto skewers (large number of random vectors) [35, 42,43]. The points corresponding to extremes, for each skewer direction, are stored. A cumulative account records the number of times each pixel (i.e., a given spectral vector) is found to be an extreme. The pixels with the highest scores are the purest ones. N-FINDR algorithm [40] is based on the fact that in p spectral dimensions, the p-volume defined by a simplex formed by the purest pixels is larger than any other volume defined by any other combination of pixels. This algorithm finds the set of pixels defining the largest volume by inflating a simplex inside the data. ORA SIS [44, 45] is a hyperspectral framework developed by the U.S. Naval Research Laboratory consisting of several algorithms organized in six modules: exemplar selector, adaptative learner, demixer, knowledge base or spectral library, and spatial postrocessor. The first step consists in flat-fielding the spectra. Next, the exemplar selection module is used to select spectral vectors that best represent the smaller convex cone containing the data. The other pixels are rejected when the spectral angle distance (SAD) is less than a given thresh old. The procedure finds the basis for a subspace of a lower dimension using a modified Gram–Schmidt orthogonalizati on. The selected vectors are then projected onto this subspace and a simplex is found by an MV T pro cess. ORA SIS is oriented to real-time target detection from uncrewed air vehicles using hyperspectral data [46]. In this chapter we develop a new algorithm to unmix linear mixtures of endmember spectra. First, the algorithm determines the number of endmembers and the signal subspace using a newly developed concept [47, 48]. Second, the algorithm extracts the most pure pixels present in the data. Unlike other methods, this algorithm is completely automatic and unsupervised. To estimate the number of endmembers and the signal subspace in hyperspectral linear mixtures, the proposed scheme begins by estimating sign al and noise correlation matrices. The latter is based on multiple regression theory. The signal subspace is then identified by selectin g the set of signal eigenvalue s that best represents the data, in the least-square sense [48,49 ], we note, however, that VCA works with projected and with unprojected data. The extraction of the end members exploits two facts: (1) the endmembers are the vertices of a simplex and (2) the affine transformation of a simplex is also a simplex. As PPI and N-FIND R algorithms, VCA also assumes the presence of pure pixels in the data. The algorithm iteratively projects data on to a direction orthogonal to the subspace spanned by the endmembers already determined. The new end member signature corresponds to the extreme of the projection. The algorithm iterates until all end members are exhausted. VCA performs much better than PPI and better than or comparable to N-FI NDR; yet it has a computational complexity between on e and two orders of magnitude lower than N-FINDR. The chapter is structure d as follows. Section 19.2 describes the fundamentals of the proposed method. Section 19.3 and Section 19.4 evaluate the proposed algorithm using simulated and real data, respectively. Section 19.5 presents some concluding remarks.