A nonparametric method for the measurement of size diversity with emphasis on data standardization

The most suitable method for estimation of size diversity is investigated. Size diversity is computed on the basis of the Shannon diversity expression adapted for continuous variables, such as size. It takes the form of an integral involving the probability density function (pdf) of the size of the individuals. Different approaches for the estimation of pdf are compared: parametric methods, assuming that data come from a determinate family of pdfs, and nonparametric methods, where pdf is estimated using some kind of local evaluation. Exponential, generalized Pareto, normal, and log-normal distributions have been used to generate simulated samples using estimated parameters from real samples. Nonparametric methods include discrete computation of data histograms based on size intervals and continuous kernel estimation of pdf. Kernel approach gives accurate estimation of size diversity, whilst parametric methods are only useful when the reference distribution have similar shape to the real one. Special attention is given for data standardization. The division of data by the sample geometric mean is proposedas the most suitable standardization method, which shows additional advantages: the same size diversity value is obtained when using original size or log-transformed data, and size measurements with different dimensionality (longitudes, areas, volumes or biomasses) may be immediately compared with the simple addition of ln k where kis the dimensionality (1, 2, or 3, respectively). Thus, the kernel estimation, after data standardization by division of sample geometric mean, arises as the most reliable and generalizable method of size diversity evaluation

Characterization of Mechanically Alloyed Nanocrystalline Fe(Al): Crystallite Size and Dislocation Density

A nanostructured disordered Fe(Al) solid solution was obtained from elemental powders of Fe and Al using a high-energy ball mill. The transformations occurring in the material during milling were studied with the use of X-ray diffraction. In addition lattice microstrain, average crystallite size, dislocation density, and the lattice parameter were determined. Scanning electron microscopy (SEM) was employed to examine the morphology of the samples as a function of milling times. Thermal behaviour of the milled powders was examined by differential scanning calorimetry (DSC). The results, as well as dissimilarity between calorimetric curves of the powders after 2 and 20 h of milling, indicated the formation of a nanostructured Fe(Al) solid solution

Modeling and Classifying Breast Tissue Density in Mammograms

We present a new approach to model and classify breast parenchymal tissue. Given a mammogram, first, we will discover the distribution of the different tissue densities in an unsupervised manner, and second, we will use this tissue distribution to perform the classification. We achieve this using a classifier based on local descriptors and probabilistic Latent Semantic Analysis (pLSA), a generative model from the statistical text literature. We studied the influence of different descriptors like texture and SIFT features at the classification stage showing that textons outperform SIFT in all cases. Moreover we demonstrate that pLSA automatically extracts meaningful latent aspects generating a compact tissue representation based on their densities, useful for discriminating on mammogram classification. We show the results of tissue classification over the MIAS and DDSM datasets. We compare our method with approaches that classified these same datasets showing a better performance of our proposal

Estimating extreme value cumulative distribution functions using bias-corrected kernel approaches

We propose a new kernel estimation of the cumulative distribution function based on transformation and on bias reducing techniques. We derive the optimal bandwidth that minimises the asymptotic integrated mean squared error. The simulation results show that our proposed kernel estimation improves alternative approaches when the variable has an extreme value distribution with heavy tail and the sample size is small.

Ultralow-density nanocage-based metal-oxide polymorphs

For two important metal oxides (MO, M=Mg, Zn) we predict, via accurate electronic structure calculations, that new low-density nanoporous crystalline phases may be accessible via the coalescence of nanocluster building blocks. Specifically, we consider the assembly of cagelike (MO)12 clusters exhibiting particularly high gas phase stability, leading to new polymorphs with energetic stabilities rivaling (and sometimes higher) than those of known MO polymorphs.

Learning Preferences with Kernel-Based Methods

Learning of preference relations has recently received significant attention in machine learning community. It is closely related to the classification and regression analysis and can be reduced to these tasks. However, preference learning involves prediction of ordering of the data points rather than prediction of a single numerical value as in case of regression or a class label as in case of classification. Therefore, studying preference relations within a separate framework facilitates not only better theoretical understanding of the problem, but also motivates development of the efficient algorithms for the task. Preference learning has many applications in domains such as information retrieval, bioinformatics, natural language processing, etc. For example, algorithms that learn to rank are frequently used in search engines for ordering documents retrieved by the query. Preference learning methods have been also applied to collaborative filtering problems for predicting individual customer choices from the vast amount of user generated feedback. In this thesis we propose several algorithms for learning preference relations. These algorithms stem from well founded and robust class of regularized least-squares methods and have many attractive computational properties. In order to improve the performance of our methods, we introduce several non-linear kernel functions. Thus, contribution of this thesis is twofold: kernel functions for structured data that are used to take advantage of various non-vectorial data representations and the preference learning algorithms that are suitable for different tasks, namely efficient learning of preference relations, learning with large amount of training data, and semi-supervised preference learning. Proposed kernel-based algorithms and kernels are applied to the parse ranking task in natural language processing, document ranking in information retrieval, and remote homology detection in bioinformatics domain. Training of kernel-based ranking algorithms can be infeasible when the size of the training set is large. This problem is addressed by proposing a preference learning algorithm whose computation complexity scales linearly with the number of training data points. We also introduce sparse approximation of the algorithm that can be efficiently trained with large amount of data. For situations when small amount of labeled data but a large amount of unlabeled data is available, we propose a co-regularized preference learning algorithm. To conclude, the methods presented in this thesis address not only the problem of the efficient training of the algorithms but also fast regularization parameter selection, multiple output prediction, and cross-validation. Furthermore, proposed algorithms lead to notably better performance in many preference learning tasks considered.

Fragmentation of massive dense cores down to <~ 1000 AU: relation between fragmentation and density structure

In order to shed light on the main physical processes controlling fragmentation of massive dense cores, we present a uniform study of the density structure of 19 massive dense cores, selected to be at similar evolutionary stages, for which their relative fragmentation level was assessed in a previous work. We inferred the density structure of the 19 cores through a simultaneous fit of the radial intensity profiles at 450 and 850 μm (or 1.2 mm in two cases) and the spectral energy distribution, assuming spherical symmetry and that the density and temperature of the cores decrease with radius following power-laws. Even though the estimated fragmentation level is strictly speaking a lower limit, its relative value is significant and several trends could be explored with our data. We find a weak (inverse) trend of fragmentation level and density power-law index, with steeper density profiles tending to show lower fragmentation, and vice versa. In addition, we find a trend of fragmentation increasing with density within a given radius, which arises from a combination of flat density profile and high central density and is consistent with Jeans fragmentation. We considered the effects of rotational-to-gravitational energy ratio, non-thermal velocity dispersion, and turbulence mode on the density structure of the cores, and found that compressive turbulence seems to yield higher central densities. Finally, a possible explanation for the origin of cores with concentrated density profiles, which are the cores showing no fragmentation, could be related with a strong magnetic field, consistent with the outcome of radiation magnetohydrodynamic simulations.

Model for interevent times with long tails and multifractality in human communications: An application to financial trading

Social, technological, and economic time series are divided by events which are usually assumed to be random, albeit with some hierarchical structure. It is well known that the interevent statistics observed in these contexts differs from the Poissonian profile by being long-tailed distributed with resting and active periods interwoven. Understanding mechanisms generating consistent statistics has therefore become a central issue. The approach we present is taken from the continuous-time random-walk formalism and represents an analytical alternative to models of nontrivial priority that have been recently proposed. Our analysis also goes one step further by looking at the multifractal structure of the interevent times of human decisions. We here analyze the intertransaction time intervals of several financial markets. We observe that empirical data describe a subtle multifractal behavior. Our model explains this structure by taking the pausing-time density in the form of a superstatistics where the integral kernel quantifies the heterogeneous nature of the executed tasks. A stretched exponential kernel provides a multifractal profile valid for a certain limited range. A suggested heuristic analytical profile is capable of covering a broader region.

New approximation to the third-order density : application to the calculation of correlated multicenter indices

In this work we present the formulas for the calculation of exact three-center electron sharing indices (3c-ESI) and introduce two new approximate expressions for correlated wave functions. The 3c-ESI uses the third-order density, the diagonal of the third-order reduced density matrix, but the approximations suggested in this work only involve natural orbitals and occupancies. In addition, the first calculations of 3c-ESI using Valdemoro's, Nakatsuji's and Mazziotti's approximation for the third-order reduced density matrix are also presented for comparison. Our results on a test set of molecules, including 32 3c-ESI values, prove that the new approximation based on the cubic root of natural occupancies performs the best, yielding absolute errors below 0.07 and an average absolute error of 0.015. Furthemore, this approximation seems to be rather insensitive to the amount of electron correlation present in the system. This newly developed methodology provides a computational inexpensive method to calculate 3c-ESI from correlated wave functions and opens new avenues to approximate high-order reduced density matrices in other contexts, such as the contracted Schrödinger equation and the anti-Hermitian contracted Schrödinger equation

Is contextual-potentiated eating dependent on caloric density of food?

One experiment tested whether a specific context could elicit eating in rats as a result of Pavlovian conditioning and whether this effect depended on the caloric density of food. Thirty two deprived rats experienced two contexts. They had access to food in context A, but no food was available in context B. During conditioning, half of the animals received high density caloric food (HD groups) whereas the other half, low density caloric food (LD groups). Then, half of the rats in each type of food group was tested in context A and the other half in context B. The results demonstrated an effect of context conditioning only in HD groups. These findings suggest the relevance of both contextual conditioning and caloric density of food in eating behaviour. Implications for the aetiology of binge eating will be discussed.

A Theoretical investigation of a potential high energy density compound 3,6,7,8-tetranitro-3,6,7,8-tetraaza-tricyclo[3.1.1.1(2,4)]octane

The B3LYP/6-31G (d) density functional theory (DFT) method was used to study molecular geometry, electronic structure, infrared spectrum (IR) and thermodynamic properties. Heat of formation (HOF) and calculated density were estimated to evaluate detonation properties using Kamlet-Jacobs equations. Thermal stability of 3,6,7,8-tetranitro-3,6,7,8-tetraaza-tricyclo [3.1.1.1(2,4)]octane (TTTO) was investigated by calculating bond dissociation energy (BDE) at the unrestricted B3LYP/6-31G(d) level. Results showed the N-NO2 bond is a trigger bond during the thermolysis initiation process. The crystal structure obtained by molecular mechanics (MM) methods belongs to P2(1)/C space group, with cell parameters a = 8.239 Å, b = 8.079 Å, c = 16.860 Å, Z = 4 and r = 1.922 g cm-3. Both detonation velocity of 9.79 km s-1 and detonation pressure of 44.22 GPa performed similarly to CL-20. According to the quantitative standards of energetics and stability, TTTO essentially satisfies this requirement as a high energy density compound (HEDC).