Graduation date: 1994

In the recent years, Independent Component Analysis (ICA) has become a fundamental tool in adaptive signal and data processing, especially in the field of Blind Source Separation (BSS). Even though there exist some methods for which an algebraic solution to the ICA problem may be found, other iterative methods are very popular. Among them is the class of information-theoretic approaches, laying on entropies. The associated objective functions are maximized based on optimization schemes, and on gradient-ascent techniques in particular. Two major issues in this field are the following: 1) Does the global maximum point of these entropic objectives correspond to a satisfactory solution of BSS ? and 2) as gradient techniques are used, optimization algorithms look in fact for local maximum points, so what about the meaning of these local optima from the BSS problem point of view? Even though there are some partial answers to these questions in the literature, most of them are based on simulation and conjectures; formal developments are often lacking. This thesis aims at filling this lack and providing intuitive justifications, too. We focus the analysis on Rényi's entropy-based contrast functions. Our results show that, generally speaking, Rényi's entropy is not a suitable contrast function for BSS, even though we recover the well-known results saying that Shannon's entropy-based objectives are contrast functions. We also show that the range-based contrast functions can be built under some conditions on the sources. The BSS problem is stated in the first chapter, and viewed under the information (theory) angle. The two next chapters address specifically the above questions. Finally, the last chapter deals with range-based ICA, the only ``entropy-based contrast' which, based on the enclosed results, is also a <i>discriminant</i> contrast function, in the sense that it is theoretically free of spurious local optima. Geometrical interpretations and surprising examples are given. The interest of this approach is confirmed by testing the algorithm on the MLSP 2006 data analysis competition benchmark; the proposed method outperforms the previously obtained results on large-scale and noisy mixture samples obtained through ill-conditioned mixing matrices.

A new method is proposed for correctly modeling the long range interaction between a fluid and a bounding wall in atomistic simulations. This method incorporates the molecular structure of the solid substrate while allowing for a finite interaction cutoff by making a proper estimation of long range correction for the fluid-wall interaction. The method is then applied to a molecular dynamic simulation of a spreading droplet. Conparison to simulations using several other previously used methods shows that the long range correction can be significant in some circumstances. / Singapore-MIT Alliance (SMA)

Understanding the ways in which animals utilize space and obtain food are central themes in modern ecology. Formulating broad principles and elucidating the factors explaining such patterns are limited, however, by the availability of data from a broad range of species and systems. This problem especially true of snakes, a predator group about which even the most basic natural history data are often entirely lacking, even among abundant, widespread, and ecologically-important species. I studied the natural history and ecosystem role of one such species, the fer-de-lance (Bothrops asper) in lowland rainforest in Costa Rica. B. asper is a large, cryptic pitviper that is highly abundant in many Central American ecosystems and is strongly relevant to human health due to high incidence of snakebite, yet its biology under natural conditions is almost entirely undocumented. I used radiotelemetry to quantify home range, movement patterns, habitat usage, and foraging behavior. B. asper was found to have smaller home ranges and reduced movement patterns than similarly-sized temperate pitvipers, likely due to a greater reliance upon ambush foraging in patches of high prey density. Snakes also demonstrated strong selection for swamp habitat, which may reflect efforts to exploit frogs as a primary food source due to low availability of small mammals at the study site. I subsequently addressed the trophic status of this B. asper population using a supplemental-feeding experiment. In comparison to control snakes, individuals receiving supplemental food had smaller home ranges, shorter and less frequent movements, increased mass acquisition, and shifted to primarily forest rather than swamp habitat. These results support the suggestion that B. asper at the study site are strongly food-limited. Finally, I tested the hypothesis that fer-de-lance mediate local seed-predation rates by influencing habitat usage and foraging behavior of rodents. A series of behavioral experiments conflicted with many existing studies in failing to support this idea, as three rodent species demonstrated little snake avoidance, and none of likely ecological relevance. Collectively, this dissertation represents the first comprehensive field study of Bothrops asper and is among the first for any tropical snake, and suggests several avenues for future research.

Range finding are widely used in various industrial applications, such as machine vision, collision avoidance, and robotics. Presently most range finders either rely on active transmitters or sophisticated mechanical controllers and powerful processors to extract range information, which make the range finders costly, bulky, or slowly, and limit their applications. This dissertation is a detailed description of a real-time vision-based range sensing technique and its single-chip CMOS implementation. To the best of our knowledge, this system is the first single chip vision-based range finder that doesn't need any mechanical position adjustment, memory or digital processor. The entire signal processing on the chip is purely analog and occurs in parallel. The chip captures the image of an object and extracts the depth and range information from just a single picture. The on-chip, continuous-time, logarithmic photoreceptor circuits are used to couple spatial image signals into the range-extracting processing network. The photoreceptor pixels can adjust their operating regions, simultaneously achieving high sensitivity and wide dynamic range. The image sharpness processor and Winner-Take-All circuits are characterized and analyzed carefully for their temporal bandwidth and detection performance. The mathematical and optical models of the system are built and carefully verified. A prototype based on this technique has been fabricated and tested. The experimental results prove that the range finder can achieve acceptable range sensing precision with low cost and excellent speed performance in short-to-medium range coverage. Therefore, it is particularly useful for collision avoidance.

Woody encroachment upon grasslands is a global trend that is cause for concern. In the Lampasas Cut Plain of Texas, Ashe juniper is the dominant woody invasive species. Grassland restoration is dependent upon proper seedbed preparation and seedling establishment. Shearing and bulldozing are common methods of juniper removal. Three hypotheses were tested in this experiment. The first was that bulldozing and shearing results in similar seedbed preparations. The second was that electrical conductivity (ECa) models soil moisture storage heterogeneity across a landscape. The third hypothesis tested was that bulldozing and shearing result in similar seedling establishment. Eighteen plots were selected and three treatments were applied: 1) shearing, 2) dozing, and 3) control. After clearing with both juniper removal methods, the levels of soil disturbance and vegetation regrowth were measured. Point measures were used to describe soil surface disturbance, and at each point, presence and size of ground cover and surface depressions were recorded. After germination, when the plants were large enough to identify, vegetation was sampled at 20-24 locations in each plot. At each sample location a quadrat was placed on the transect, and total herbaceous, grass, forb, rock, litter, and bare soil cover were recorded in cover classes. All species were identified, seeded species were enumerated, and in the second vegetation sample, dominant species were assigned a cover class. There was considerable variation between plots for all parameters measured. Bulldozing created a greater number of depressions in the soil than shearing. It also had a higher percent of large depressions. There was no difference in ground cover change between treatments except in the case of the largest litter category. ECa was effective in modeling soil moisture storage patterns and had a lower coefficient of variation than manual soil depth measurements. There was no difference between sheared and dozed plots for any of the vegetation establishment parameters, but control plots had lower establishment in most categories. Overall, there was little difference between the two treatments in seedbed preparation and seedling establishment. The cost of shearing was 80% of the cost of dozing, which made it more desirable in this scenario.

In this thesis we provide a model for a bufferless optical burst switching (OBS) and an optical packet switching (OPS) network. The thesis is divided in three parts. In the first part we introduce the basic functionality and structure of OBS and OPS networks. We identify the blocking probability as the main performance parameter of interest. In the second part we study the statistical properties of the traffic that will likely run through these networks. We use for this purpose a set of traffic traces obtained from the Universidad Politécnica de Catalunya. Our conclusion is that traffic entering the optical domain in future OBS/OPS networks will be long-range dependent (LRD). In the third part we present the model for bufferless OBS/OPS networks. This model takes into account the results from the second part of the thesis concerning the LRD nature of traffic. It also takes into account specific issues concerning the functionality of a typical bufferless packet-switching network. The resulting model presents scalability problems, so we propose an approximative method to compute the blocking probability from it. We empirically evaluate the accuracy of this method, as well as its scalability.

This thesis aims at a better mechanistic understanding of animal communities. Therefore, an allometry- and individual-based model has been developed which was used to simulate mammal and bird communities in heterogeneous landscapes, and to to better understand their response to landscape changes (habitat loss and fragmentation). / Diese Doktorarbeit strebt ein besseres mechanistisches Verständnis von Tiergemeinschaften an. Dafür wurde ein allometrie- und individuen-basiertes Modell entwickelt und dazu benutzt, Säugetier- und Vogelgemeinschaften in heterogenen Landschaften zu simulieren, und ihre Reaktion auf Landschaftsveränderungen (Habitatverlust und -fragmentierung) besser zu verstehen.

Aircrafttrajectory optimization is traditionally used forminimizing fuel consumption or time when going from one flightstate to another. This thesis presents a possible approach toincorporate tactical constraints in aircraft trajectoryoptimization. The stealth technology of today focuses on making thetactics already in use more effective. Since tactics andstealth are closely interrelated, new and better results may beobtained if both aspects are considered simultaneously. Simplyreducing the radar cross section area in some directionswithout considering tactical aspects may result in little, ifany, improvement. Flight tests have been performed in cooperation withEricsson Microwave Systems and the Swedish Air Force FlightAcademy. The aircraft used was the subsonic jet trainer Saab105, designated SK60 by the Swedish Air Force. The results showa decrease of 40% in the time interval between the instant theaircraft was first detected until it could pass above the radarstation. This corresponds to a reduced radar cross section(RCS) in the direction from the aircraft to the radar of almost90%, if classical RCS reduction techniques would have beenapplied. If a modern aircraft with stealth properties would be used,the proposed methodology is believed to increase the possibleimprovements further. This is because the variation of themagnitude of RCS in different directions is greater for a shapeoptimized aircraft, which is the property exploited by thedeveloped method. The methods presented are indeed an approach utilizing theideas of the network centric warfare (NCW) concept. Themethodology presented depends on accurate information about theadversary, while also providing up-to-date information to theother users in the information network. The thesis focuses on aircraft but the methods are generaland may be adapted for missiles, shipsor land vehicles. Theproposed methods are also economically viable since they areuseful for existing platforms without costly modifications. Themethods presented are not limited to radar threats only. Thereasons for using radar in this thesis are the availablenon-classified data and that radar is known to pose a majorthreat against aircraft.

<p>Southern Africa harbours one-third of the world&rsquo / s Testudinid species, many of which inhabit arid or semi-arid areas, but ecological information on these species is scant. I studied the habitat, morphology and ecology of Kalahari tent tortoises over 13 months in semi-arid Savanna at Benfontein farm, Northern Cape Province, South Africa. In order to allow continuous monitoring of individuals, I attached radiotransmitters to males and females, split equally between two habitats, sites E (east) and W (west), with apparent differences in vegetation structure. Results of the study were based on data obtained from 27 telemetered tortoises and 161 individuals encountered opportunistically. Female Kalahari tent tortoises were larger than males and the sex ratio did not differ from 1:1. Based on person-hours to capture tortoises, the population appeared to have a low density, with more time required to capture a juvenile (35 hours) than an adult (10-11 hours). The frequency distribution of body size ranges was indicative of recruitment. Relative age, based on annuli counts, suggested that males were younger than females, perhaps because males as the smaller sex are more predation-prone than females. Linear relationships between annuli counts and shell volume indicated that, after reaching sexual maturity, female body size increased faster in volume than did male body size, possibly because a larger volume may enhance female reproductive success. Body condition differed between sites, sexes and among seasons. The hot and dry summer may account for low summer body condition, whereas vegetation differences and size effects, respectively, may account for the low body condition of tortoises in site W and in males. Site E was sandy with grasses, particularly Schmidtia pappophoroides, being the prevalent growth form. This habitat resembled a Savanna vegetation type Schmidtia pappophoroides &ndash / Acacia erioloba described for a neighbouring reserve. Site W was stonier, dominated by shrubs, and was reminiscent of Northern Upper Karoo vegetation (NKu3). Neither site resembled Kimberley Thornveld (SVk4), the designated vegetation type of the area. Differences in substrate and grazing intensity may have contributed to site vegetation differences. Rainfall had an important influence on seasonal vegetation. Short grass abundance correlated with rainfall and annual plants sprouted after spring rain. Refuge use changed according to season and sex. Males selected denser refuges than females did, perhaps because males were smaller and more vulnerable to predation and solar heat. Tortoises selected sparse, short grass as refuges in cool months, probably to maximise basking whilst remaining in protective cover. During hot periods, mammal burrows were preferred to vegetation as refugia. The smaller males spent more time in cover than females, which may be related to predator avoidance or thermoregulation.&nbsp / Females spent more time basking than males, perhaps due to their larger size and to facilitate reproductive processes. Tortoises did not brumate, but through a combination of basking, and orientation relative to the sun in their refuges, managed to attain body temperatures that allowed small bouts of activity. Body temperature for active tortoises was similar among seasons, and was higher for more specialised active behaviours, such as feeding and socialising, than for walking. Increased activity by males in spring could relate to mating behaviour while females were more active in autumn, when they foraged more than males, perhaps due to the high cost of seasonal reproductive requirements. Males displaced further per day than did females, but home range estimates did not differ between sexes. Annual home range estimates varied substantially among individuals: 0.7&ndash / 306 ha for minimum convex polygons and 0.7&ndash / 181 ha for 95% fixed kernel estimates. The ability to&nbsp / cover large areas would assist tortoises in finding resources, e.g., food, in an area where resource distribution may be patchy. Differences among seasonal home ranges and movements probably reflect seasonal climatic change / activity areas shrinking when temperatures were extreme. In order to assess the effects of a semi-arid environment on the morphology of P. oculifer, I compared its morphology to that of its &lsquo / cool-adapted&rsquo / sister taxon Psammobates geometricus, using live and museum specimens. Both P. oculifer and P. geometricus are sexually dimorphic and differences between the two species could indicate environmental or sexual selection effects, or a combination of the two. The shorter bridge length, which allowed more leg space, and wider front feet in P. oculifer cohorts probably represent traits for manoeuvring in a sandy habitat, while wider heads in P. oculifer possibly relate to interspecific differences in diet. The flatter shell in female P. oculifer, relative to P. geometricus, may represent a trade-off between space for reproductive structures, e.g., eggs, and the need to fit into small refuges, e.g., mammal burrows. Male P. oculifer had wider shells, more space around their hind legs, and wider hind feet than P. geometricus males had, all characteristics which may assist males to fight and mate in a sandy environment.</p>