Domain Adaptation in remote sensing: increasing the portability of land-cover classifiers

Among the types of remote sensing acquisitions, optical images are certainly one of the most widely relied upon data sources for Earth observation. They provide detailed measurements of the electromagnetic radiation reflected or emitted by each pixel in the scene. Through a process termed supervised land-cover classification, this allows to automatically yet accurately distinguish objects at the surface of our planet. In this respect, when producing a land-cover map of the surveyed area, the availability of training examples representative of each thematic class is crucial for the success of the classification procedure. However, in real applications, due to several constraints on the sample collection process, labeled pixels are usually scarce. When analyzing an image for which those key samples are unavailable, a viable solution consists in resorting to the ground truth data of other previously acquired images. This option is attractive but several factors such as atmospheric, ground and acquisition conditions can cause radiometric differences between the images, hindering therefore the transfer of knowledge from one image to another. The goal of this Thesis is to supply remote sensing image analysts with suitable processing techniques to ensure a robust portability of the classification models across different images. The ultimate purpose is to map the land-cover classes over large spatial and temporal extents with minimal ground information. To overcome, or simply quantify, the observed shifts in the statistical distribution of the spectra of the materials, we study four approaches issued from the field of machine learning. First, we propose a strategy to intelligently sample the image of interest to collect the labels only in correspondence of the most useful pixels. This iterative routine is based on a constant evaluation of the pertinence to the new image of the initial training data actually belonging to a different image. Second, an approach to reduce the radiometric differences among the images by projecting the respective pixels in a common new data space is presented. We analyze a kernel-based feature extraction framework suited for such problems, showing that, after this relative normalization, the cross-image generalization abilities of a classifier are highly increased. Third, we test a new data-driven measure of distance between probability distributions to assess the distortions caused by differences in the acquisition geometry affecting series of multi-angle images. Also, we gauge the portability of classification models through the sequences. In both exercises, the efficacy of classic physically- and statistically-based normalization methods is discussed. Finally, we explore a new family of approaches based on sparse representations of the samples to reciprocally convert the data space of two images. The projection function bridging the images allows a synthesis of new pixels with more similar characteristics ultimately facilitating the land-cover mapping across images.

Time of ruin in a risk model with generalized Erlang (n) interclaim times and a constant dividend barrier

In this paper we analyze the time of ruin in a risk process with the interclaim times being Erlang(n) distributed and a constant dividend barrier. We obtain an integro-differential equation for the Laplace Transform of the time of ruin. Explicit solutions for the moments of the time of ruin are presented when the individual claim amounts have a distribution with rational Laplace transform. Finally, some numerical results and a compare son with the classical risk model, with interclaim times following an exponential distribution, are given.

Non-constant discounting in finite horizon: The free terminal time case

This paper derives the HJB (Hamilton-Jacobi-Bellman) equation for sophisticated agents in a finite horizon dynamic optimization problem with non-constant discounting in a continuous setting, by using a dynamic programming approach. A simple example is used in order to illustrate the applicability of this HJB equation, by suggesting a method for constructing the subgame perfect equilibrium solution to the problem.Conditions for the observational equivalence with an associated problem with constantdiscounting are analyzed. Special attention is paid to the case of free terminal time. Strotz¿s model (an eating cake problem of a nonrenewable resource with non-constant discounting) is revisited.

Genome-wide arrays in routine diagnostics of hematological malignancies.

Over the last three decades, cytogenetic analysis of malignancies has become an integral part of disease evaluation and prediction of prognosis or responsiveness to therapy. In most diagnostic laboratories, conventional karyotyping, in conjunction with targeted fluorescence in situ hybridization analysis, is routinely performed to detect recurrent aberrations with prognostic implications. However, the genetic complexity of cancer cells requires a sensitive genome-wide analysis, enabling the detection of small genomic changes in a mixed cell population, as well as of regions of homozygosity. The advent of comprehensive high-resolution genomic tools, such as molecular karyotyping using comparative genomic hybridization or single-nucleotide polymorphism microarrays, has overcome many of the limitations of traditional cytogenetic techniques and has been used to study complex genomic lesions in, for example, leukemia. The clinical impact of the genomic copy-number and copy-neutral alterations identified by microarray technologies is growing rapidly and genome-wide array analysis is evolving into a diagnostic tool, to better identify high-risk patients and predict patients' outcomes from their genomic profiles. Here, we review the added clinical value of an array-based genome-wide screen in leukemia, and discuss the technical challenges and an interpretation workflow in applying arrays in the acquired cytogenetic diagnostic setting.

Genesis and classification of Oxisols in a highland toposequence of the upper Jequitinhonha Valley (MG)

The Brazilian System of Soil Classification (SiBCS) is a taxonomic system, open and in permanent construction, as new knowledge on Brazilian soils is obtained. The objective of this study was to characterize the chemical, physical, morphological, micro-morphological and mineralogical properties of four pedons of Oxisols in a highland toposequence in the upper Jequitinhonha Valley, emphasizing aspects of their genesis, classification and landscape development. The pedons occupy the following slope positions: summit - Red Oxisol (LV), mid slope (upper third) - Yellow-Red Oxisol (LVA), lower slope (middle third)- Yellow Oxisol (LA) and bottom of the valley (lowest third) - "Gray Oxisol" ("LAC"). These pedons were described and sampled for characterization in chemical and physical routine analyses. The total Fe, Al and Mn contents were determined by sulfuric attack and the Fe, Al and Mn oxides in dithionite-citrate-bicarbonate and oxalate extraction. The mineralogy of silicate clays was identified by X ray diffraction and the Fe oxides were detected by differential X ray diffraction. Total Ti, Ga and Zr contents were determined by X ray fluorescence spectrometry. The "LAC" is gray-colored and contains significant fragments of structure units in the form of a dense paste, characteristic of a gleysoil, in the horizons A and BA. All pedons are very clayey, dystrophic and have low contents of available P and a pH of around 5. The soil color was related to the Fe oxide content, which decreased along the slope. The decrease of crystalline and low- crystalline Fe along the slope confirmed the loss of Fe from the "LAC". Total Si increased along the slope and total Al remained constant. The clay fraction in all pedons was dominated by kaolinite and gibbsite. Hematite and goethite were identified in LV, low-intensity hematite and goethite in LVA, goethite in LA. In the "LAC", no hematite peaks and goethite were detected by differential X ray diffraction. The micro-morphology indicated prevalence of granular microstructure and porosity with complex stacking patterns.. The soil properties in the toposequence converged to a single soil class, the Oxisols, derived from the same source material. The landscape evolution and genesis of Oxisols of the highlands in the upper Jequitinhonha Valley are related to the evolution of the drainage system and the activity of excavating fauna.

Dielectric constant obtained from TDR and volumetric moisture of soils in southern Brazil

Soil moisture is the property which most greatly influences the soil dielectric constant, which is also influenced by soil mineralogy. The aim of this study was to determine mathematical models for soil moisture and the dielectric constant (Ka) for a Hapludalf, two clayey Hapludox and a very clayey Hapludox and test the reliability of universal models, such as those proposed by Topp and Ledieu and their co-workers in the 80's, and specific models to estimate soil moisture with a TDR. Soil samples were collected from the 0 to 0.30 m layer, sieved through a mesh of 0.002 m diameter and packed in PVC cylinders with a 0.1 m diameter and 0.3 m height. Seven samples of each soil class were saturated by capillarity and a probe composed of two rods was inserted in each one of them. Moisture readings began with the saturated soil and concluded when the soil was near permanent wilting point. In each step, the samples were weighed on a precision scale to calculate volumetric moisture. Linear and polynomial models were adjusted for each soil class and for all soils together between soil moisture and the dielectric constant. Accuracy of the models was evaluated by the coefficient of determination, the standard error of estimate and the 1:1 line. The models proposed by Topp and Ledieu and their co-workers were not adequate for estimating the moisture in the soil classes studied. The adjusted linear and polynomial models for the entire set of data of the four soil classes did not have sufficient accuracy for estimating soil moisture. The greater the soil clay and Fe oxide content, the greater the dielectric constant of the medium for a given volumetric moisture. The specific models, θ = 0.40283 - 0.04231 Ka + 0.00194 Ka² - 0.000022 Ka³ (Hapludox) θ = 0.01971 + 0.02902 Ka - 0.00086 Ka² + 0.000012 Ka³ (Hapludox -PF), θ = 0.01692 - 0.00507 Ka (Hapludalf) and θ = 0.08471 + 0.01145 Ka (Hapludox-CA), show greater accuracy and reliability for estimating soil moisture in the soil classes studied.

Routine therapeutic drug monitoring in patients treated with 10-360 mg/day citalopram

From data collected during routine TDM, plasma concentrations of citalopram (CIT) and its metabolites demethylcitalopram (DCIT) and didemethylcitalopram (DDCIT) were measured in 345 plasma samples collected in steady-state conditions. They were from 258 patients treated with usual doses (20-60 mg/d) and from patients medicated with 80-360 mg/d CIT. Most patients had one or several comedications, including other antidepressants, antipsychotics, lithium, anticonvulsants, psychostimulants and somatic medications. Dose-corrected CIT plasma concentrations (C/D ratio) were 2.51 +/- 2.25 ng mL-1 mg-1 (n = 258; mean +/- SD). Patients >65 years had significantly higher dose-corrected CIT plasma concentrations (n = 56; 3.08 +/- 1.35 ng mL-1 mg-1) than younger patients (n = 195; 2.35 +/- 2.46 ng mL-1 mg-1) (P = 0.03). CIT plasma concentrations in the generally recommended dose range were [mean +/- SD, (median)]: 57 +/- 64 (45) ng/mL (10-20 mg/d; n = 64), 117 +/- 95 (91) ng/mL (21-60 mg/d; n = 96). At higher than usual doses, the following concentrations of CIT were measured: 61-120 mg/d CIT, 211 +/- 103 (190) ng/mL (n = 93); 121-200 mg/d: 339 +/- 143 (322) ng/mL (n = 70); 201-280 mg/d: 700 +/- 408 (565) ng/mL (n = 18); 281-360 mg/d: 888 +/- 620 (616) ng/mL (n = 4). When only one sample per patient (at the highest daily dose if repeated dosages) is considered, there is a linear and significant correlation (n = 48, r = 0.730; P < 0.001) between daily dose (10-200 mg/d) and CIT plasma concentrations. In experiments with dogs, DDCIT was reported to affect the QT interval when present at concentrations >300 ng/mL. In this study, DDCIT concentration reached 100 ng/mL in a patient treated with 280 mg/d CIT. Twelve other patients treated with 140-320 mg/d CIT had plasma concentrations of DDCIT within the range 52-73 ng/mL. In a subgroup comprised of patients treated with > or =160 mg/d CIT and with CIT plasma concentrations < or =300 ng/mL, and patients treated with < or =200 mg/d CIT and CIT plasma concentrations > or = 600 ng/mL, the enantiomers of CIT and DCIT were also analyzed. The highest S-CIT concentration measured in this subgroup was 327 ng/mL in a patient treated with 140 mg/d CIT, but the highest S-CIT concentration (632 ng/mL) was measured in patient treated with 360 mg/d CIT. In conclusion, there is a highly linear correlation between CIT plasma concentrations and CIT doses, well above the usual dose range.

Effective mass and dielectric constant mismatch effects in spherical multishell quantum dots

The role of effective mass and dielectric mismatches on chemical potentials and addition energies of many-electron multishell quantum dots (QDs) is explored within the framework of a recent extension of the spin density functional theory. It is shown that although the gross electronic density is located in the wells of these multishell QDs, taking position-dependent effective mass and dielectric constant into account can lead to the appearance of relevant differences in chemical potential and addition energies as compared to standard calculations in which the effective mass and the dielectric constant of the well is assumed for the whole multishell structure.