Дисертації з теми "Guided sampling"

Animals must sense their surroundings to update their internal representations of the external environment, and exploratory behaviours such as sensory sampling are influenced by past experiences. This thesis investigates how voluntary sensory sampling activities undergo learning-dependent changes. Studies of freely behaving animals impose two major challenges: 1) the accuracy of biological measurements is compromised by movement-induced artifacts, and 2) large degrees of freedom in unrestrained behaviours confound well-controlled studies. Pulse-type weakly electric fish (WEF) are an ideal choice to study adaptive sensory sampling from unrestrained animals, since they generate readily observable and quantifiable sensory capture events expressed by discrete pulses of electric organ discharges (EODs). To study the voluntarily movements and sensory sampling while animals navigated in darkness, we developed three novel experimental techniques to track movements and detect sensory sampling from a freely behaving WEF: 1) an EOD detector to remotely and accurately measure the sensory sampling rate, 2) an electrical tracking method to track multiple WEF using their own EODs, and 3) visual tracking algorithm for robust body tracking through water under infrared illumination. These techniques were successfully applied to reveal novel sensory sampling behaviours in freely exploring Gymnotus sp. Cortical activity precedes self-initiated movements by several seconds in mammals; this observation has led into inquiries on the nature of volition. Here we demonstrate the sensory sampling enhancement also precedes self-initiated movement by a few seconds in Gymnotus sp. Next, we tested whether these animals can be trained to learn a location of food using electrically detectable landmarks and, if so, whether they can use their past experiences to optimize their sensory sampling. We found that animals revisited the missing food location with high spatial accuracy, and they intensified their sensory sampling near the expected food location by increasing the number of EOD pulses per unit distance travelled.

This thesis describes the motivation and development of Pocket Interview, an easily configurable handheld electronic data collection and diary tool. The system can be used to apply ‘experience sampling’ methods that allow the collection of data in real-time and in the user’s natural environment. Pocket Interview can prompt the user to make diary entries at fixed and/or random intervals. The system is configured via graphical user interfaces that are shown to be easily usable by non-computing users. Pocket Interview includes an option that allows this sampling to be ‘Guided’ whereby inconvenient prompts are temporarily deferred until a more convenient time through the use of contextual audio information. Subjects participating in real-time studies require high levels of commitment and exhibit difficulties maintaining their motivation. Guiding can offer to reduce the perceived burden on the user, improve response rates, increase the quantity of replies and the quality of those replies. This thesis describes a series of studies that investigate the following: • What are the more convenient times for sampling and how can they be detected? • Does Guided real-time sampling improve the data quality and participant compliance rates? • Participants attitudes towards mobile devices automatically gathering their context information. Guiding is a strategy that could be applied to all context-aware computing, phone call or message delivery and indeed all other prompting. As computing power continues to expand and more powerful mobile devices become available we will see an increase in the quantity and sophistication of applications that interrupt their users. This will add to user’s feelings of overload. To maximise user acceptability designers of computing systems require strategies, such as Guiding, to minimise the interruptions caused by proactive prompting.

Thesis: S.M., Massachusetts Institute of Technology, Department of Aeronautics and Astronautics, 2017.
Cataloged from PDF version of thesis.
Includes bibliographical references (pages 79-81).
Aerial resupply can deliver cargo to locations across the globe. A challenge for modern guided parafoil systems is to land accurately in complex terrain, including canyons and cities. This thesis presents the Rewire-RRT algorithm for parafoil terminal guidance. The algorithm uses Rapidly-Exploring Random Trees (RRT) to efficiently search for feasible paths through complex environments. Most importantly, Rewire-RRT provides a mechanism to build and rewire the tree to explicitly minimize the risk of collision with obstacles along each path and to minimize the expected final miss distance from the target. This key adaptation allows for parafoil guidance in urban drop zones not previously considered for airdrop operations. The Rewire-RRT algorithm is first developed and tested in two dimensions and demonstrated to have greater performance than RRT for simple dynamical systems, finding paths that are shorter and safer than those found by RRT. Then, Rewire-RRT is shown to be an effective path planner for a guided parafoil with complex dynamics. Paths planned by Rewire-RRT better meet the performance objectives of guided parafoils than those planned by RRT. Finally, simulation results show that Rewire-RRT performs better than state-of- the-art terminal guidance strategies for guided parafoils when the target location is cluttered with multiple three-dimensional obstacles.
by Brian Le Floch.
S.M.

De nombreuses structures utilisées industriellement peuvent être considérées comme des guides d'ondes, comme les plaques, les tuyaux ou encore le rails. La maintenance de ces structures nécessite de pouvoir détecter efficacement des défauts internes par le Contrôle Non Destructif. Nous nous intéressons dans ce manuscrit à l'application d'une méthode d'échantillonnage, la Linear Sampling Method, au CND des guides d'ondes élastiques, qui en particulier impose des sollicitations et des mesures à la surface du guide en régime temporel. La stratégie choisie repose sur une formulation modale et multi-fréquentielle de la LSM, spécifique aux guides d'ondes, qui permet une régularisation efficace et de nature physique du problème inverse, qui est par nature mal posé. Cette stratégie permet par ailleurs une optimisation du nombre et de la position des émetteurs et des récepteurs. Nous nous limitons dans un premier temps au cas scalaire du guide d'ondes acoustiques, pour ensuite s'attaquer au cas vectoriel, et par conséquent plus complexe, du guide d'ondes élastiques.L'efficacité de la méthode inverse est dans un premier temps démontrée sur des données artificielles (obtenues numériquement), puis sur des données réelles obtenues à l'aide d'expériences réalisées sur des plaques métalliques. Ces expériences confirment la faisabilité du CND par méthode d'échantillonnage dans un cadre industriel. Dans le cas où une seule sollicitation est réalisée, l'utilisation de la LSM est exclu. Nous utilisons une approche tout à fait différente et dite "extérieure", couplant une formulation mixte de quasi-réversibilité et une méthode de lignes de niveau, pour reconstruire le défaut
Widely used structures in an industrial context, such as plates, pipes or rails, can be considered as waveguides. Hence efficient Non Destructive Testing techniques are needed in order to detect defects in these structure during their maintenance. This work is about adapting a sampling method, the Linear Sampling Method, to the context of NDT for elastic waveguides. This context implies that the sollicitations and measurements must be on the surface of the waveguide in a time-dependent regime. A modal and multi-frequency formulation of the LSM, specific to waveguides, has been chosen to solve the problem. This formulation allows an efficient and physical regularization of the inverse problem, which is naturally ill-posed. An optimization of the number of sources and measurements and of their positioning is possible thanks to the methodology used to solve the problem. The scalar case of an acoustic waveguide is considered as a first step, while the vectorial case of an elastic waveguide, more complex by nature, is addressed in a second time.The efficiency of the method is at first tested on artificial data (numerically made), and then on real data obtained from experiments on metallic plates. These experiments show the feasibility of using sampling methods for Non Destructive Testing in an industrial context. In the case when only one sollicitation is available, the LSM can not be applied. A completely different approach is then used, which is called the ``exterior'' approach, coupling a mixed formulation of quasi-reversibility and a level-set method in order to recover the shape of the defect

In preference-based Evolutionary Multi-objective Optimization (EMO), the decision maker is looking for a diverse, but locally focused non-dominated front in a preferred area of the objective space, as close as possible to the true Pareto-front. Since solutions found outside the area of interest are considered less important or even irrelevant, the optimization can focus its efforts on the preferred area and find the solutions that the decision maker is looking for more quickly, i.e., with fewer simulation runs. This is particularly important if the available time for optimization is limited, as is the case in many real-world applications. Although previous studies in using this kind of guided-search with preference information, for example, withthe R-NSGA-II algorithm, have shown positive results, only very few of them considered the stochastic outputs of simulated systems. In the literature, this phenomenon of stochastic evaluation functions is sometimes called noisy optimization. If an EMO algorithm is run without any countermeasure to noisy evaluation functions, the performance will deteriorate, compared to the case if the true mean objective values are known. While, in general, static resampling of solutions to reduce the uncertainty of all evaluated design solutions can allow EMO algorithms to avoid this problem, it will significantly increase the required simulation time/budget, as many samples will be wasted on candidate solutions which are inferior. In comparison, a Dynamic Resampling (DR) strategy can allow the exploration and exploitation trade-off to be optimized, since the required accuracy about objective values varies between solutions. In a dense, converged population, itis important to know the accurate objective values, whereas noisy objective values are less harmful when an algorithm is exploring the objective space, especially early in the optimization process. Therefore, a well-designed Dynamic Resampling strategy which resamples the solution carefully, according to the resampling need, can help an EMO algorithm achieve better results than a static resampling allocation. While there are abundant studies in Simulation-based Optimization that considered Dynamic Resampling, the survey done in this study has found that there is no related work that considered how combinations of Dynamic Resampling and preference-based guided search can further enhance the performance of EMO algorithms, especially if the problems under study involve computationally expensive evaluations, like production systems simulation. The aim of this thesis is therefore to study, design and then to compare new combinations of preference-based EMO algorithms with various DR strategies, in order to improve the solution quality found by simulation-based multi-objective optimization with stochastic outputs, under a limited function evaluation or simulation budget. Specifically, based on the advantages and flexibility offered by interactive, reference point-based approaches, studies of the performance enhancements of R-NSGA-II when augmented with various DR strategies, with increasing degrees of statistical sophistication, as well as several adaptive features in terms of optimization parameters, have been made. The research results have clearly shown that optimization results can be improved, if a hybrid DR strategy is used and adaptive algorithm parameters are chosen according to the noise level and problem complexity. In the case of a limited simulation budget, the results allow the conclusions that both decision maker preferences and DR should be used at the same time to achieve the best results in simulation-based multi-objective optimization.
Vid preferensbaserad evolutionär flermålsoptimering försöker beslutsfattaren hitta lösningar som är fokuserade kring ett valt preferensområde i målrymden och som ligger så nära den optimala Pareto-fronten som möjligt. Eftersom lösningar utanför preferensområdet anses som mindre intressanta, eller till och med oviktiga, kan optimeringen fokusera på den intressanta delen av målrymden och hitta relevanta lösningar snabbare, vilket betyder att färre lösningar behöver utvärderas. Detta är en stor fördel vid simuleringsbaserad flermålsoptimering med långa simuleringstider eftersom antalet olika konfigurationer som kan simuleras och utvärderas är mycket begränsat. Även tidigare studier som använt fokuserad flermålsoptimering styrd av användarpreferenser, t.ex. med algoritmen R-NSGA-II, har visat positiva resultat men enbart få av dessa har tagit hänsyn till det stokastiska beteendet hos de simulerade systemen. I litteraturen kallas optimering med stokastiska utvärderingsfunktioner ibland "noisy optimization". Om en optimeringsalgoritm inte tar hänsyn till att de utvärderade målvärdena är stokastiska kommer prestandan vara lägre jämfört med om optimeringsalgoritmen har tillgång till de verkliga målvärdena. Statisk upprepad utvärdering av lösningar med syftet att reducera osäkerheten hos alla evaluerade lösningar hjälper optimeringsalgoritmer att undvika problemet, men leder samtidigt till en betydande ökning av antalet nödvändiga simuleringar och därigenom en ökning av optimeringstiden. Detta är problematiskt eftersom det innebär att många simuleringar utförs i onödan på undermåliga lösningar, där exakta målvärden inte bidrar till att förbättra optimeringens resultat. Upprepad utvärdering reducerar ovissheten och hjälper till att förbättra optimeringen, men har också ett pris. Om flera simuleringar används för varje lösning så minskar antalet olika lösningar som kan simuleras och sökrymden kan inte utforskas lika mycket, givet att det totala antalet simuleringar är begränsat. Dynamisk upprepad utvärdering kan däremot effektivisera flermålsoptimeringens avvägning mellan utforskning och exploatering av sökrymden baserat på det faktum att den nödvändiga precisionen i målvärdena varierar mellan de olika lösningarna i målrymden. I en tät och konvergerad population av lösningar är det viktigt att känna till de exakta målvärdena, medan osäkra målvärden är mindre skadliga i ett tidigt stadium i optimeringsprocessen när algoritmen utforskar målrymden. En dynamisk strategi för upprepad utvärdering med en noggrann allokering av utvärderingarna kan därför uppnå bättre resultat än en allokering som är statisk. Trots att finns ett rikligt antal studier inom simuleringsbaserad optimering som använder sig av dynamisk upprepad utvärdering så har inga relaterade studier hittats som undersöker hur kombinationer av dynamisk upprepad utvärdering och preferensbaserad styrning kan förbättra prestandan hos algoritmer för flermålsoptimering ytterligare. Speciell avsaknad finns det av studier om optimering av problem med långa simuleringstider, som t.ex. simulering av produktionssystem. Avhandlingens mål är därför att studera, konstruera och jämföra nya kombinationer av preferensbaserade optimeringsalgoritmer och dynamiska strategier för upprepad utvärdering. Syftet är att förbättra resultatet av simuleringsbaserad flermålsoptimering som har stokastiska målvärden när antalet utvärderingar eller optimeringstiden är begränsade. Avhandlingen har speciellt fokuserat på att undersöka prestandahöjande åtgärder hos algoritmen R-NSGA-II i kombination med dynamisk upprepad utvärdering, baserad på fördelarna och flexibiliteten som interaktiva referenspunktbaserade algoritmer erbjuder. Exempel på förbättringsåtgärder är dynamiska algoritmer för upprepad utvärdering med förbättrad statistisk osäkerhetshantering och adaptiva optimeringsparametrar. Resultaten från avhandlingen visar tydligt att optimeringsresultaten kan förbättras om hybrida dynamiska algoritmer för upprepad utvärdering används och adaptiva optimeringsparametrar väljs beroende på osäkerhetsnivån och komplexiteten i optimeringsproblemet. För de fall där simuleringstiden är begränsad är slutsatsen från avhandlingen att både användarpreferenser och dynamisk upprepad utvärdering bör användas samtidigt för att uppnå de bästa resultaten i simuleringsbaserad flermålsoptimering.

碩士
國立交通大學
多媒體工程研究所
97
We propose a novel sampling algorithm by considering the importance of visibility in the sampling process. This algorithm extends the bidirectional importance sampling techniques based on SRBF representation by adjusting the weight of each SRBF basis according to the previous history in visibility tests, thus combing the visibility term into importance function. Unlike previous visibility-related researches in importance sampling exploit image-space visibility coherence, we consider visibility in object space by avoiding redrawing samples in invisible directions. Consequently more samples pass the visibility test and contribute to the final rendered result. Considering visibility in object space would make our algorithm more flexible, even for scenes which have heavy occlusion. Our approach successfully reduces the variance over the entire image, not only along the shadow boundaries. Under the same computing performance, we can obtain higher quality than previous bidirectional importance approaches. Although our proposed algorithm is based on the SRBF representation, it can also be applied to other basis such as wavelet or spherical harmonics.

碩士
國立清華大學
資訊工程學系
100
Wireless Sensor Networks (WSNs) have been widely applied to many different areas such as surveillance, healthcare, environmental and utility monitoring, etc. In WSNs, each sensor node has the characteristics of small size, limited power, and connected wirelessly. It is responsible for gathering and delivering sensing data over the network periodically. Thus, the energy consumption problem becomes a challenging issue to prolong the lifetime of WSNs. Several research works utilize data aggregation and/or data compression concept to reduce the quantity of necessary transmission, since it is the primary issue that consumes sensors’ power particularly. However, the implementation of these operations requires high computational power. In this thesis, two approaches adapting to sensing data distribution to largely reduce the amount of required data transmission with limited computation are proposed. They are: Adaptive Sampling with RD Model and Adaptive Sampling in Dynamic Mode. In the first approach, the target distortion is near-optimally distributed (in the rate-distortion sense) to every sensor node corresponding to their relative fluctuation. In the latter one, the possible occurrence of rapid data change in the sensing period is concerned and deliberately manipulated. To combine these two methods, we verify the data trend of each sensor when the prediction function needs to be updated. Then according to the data trend we can decide whether to use Adaptive Sampling with RD Model or Adaptive Sampling in Dynamic Mode. Finally, several real sensed data were gathered and employed to demonstrate the performance of the proposed methods.

Robots already impact the way we understand our world and live our lives. However, their impact and use is limited by the skills they possess. Currently deployed autonomous robots lack the manipulation skills possessed by humans. To achieve general autonomy and applicability in the real world, robots must possess such skills. Autonomous manipulation requires algorithms that rapidly and reliably compute collision-free motion for robotic limbs with many degrees of freedom. Unfortunately, adequate algorithms for this task do not currently exist. Though there are many dimensions of the real-world planning task that require further research. A central problem of reliable real-world planning is that planners must rely on incomplete and inaccurate information about the world in which they are planning. The motion planning problem has exponential complexity in the robot's degrees of freedom. Consequently, the most successful planning algorithms use incomplete information obtained via sampling a subset of all possible movements. Additionally, real-world robots generally obtain information about the state of their environment through lasers, cameras and other sensors. The information obtained from these sensors contains noise and error. Thus the planner's incomplete information about the world is possibly inaccurate as well. Despite such limited information, a planner must be capable of quickly generating collision free motions to facilitate general purpose autonomous robots. This thesis proposes a new utility-guided framework for motion planning that can reliably compute collision-free motions with the efficiency required for real-world planning. The utility-guided approach begins with the observation there is regularity in space of possible motions available to a robot. Further, certain motions are more crucial than others for computing collision free paths. Together these observations form structure in the robot's space of possible movements. This structure provides a guide for the planner's exploration of possible motions. Because a complete understanding of this structure is computationally intractable, the utility-guided framework incrementally develops an approximate model discovered by past exploration. This model of the structure is used to select explorations that maximally benefit the planner. Information provided by each exploration improves the planner's approximation. The process of incremental improvement and further guided exploration iterates until an adequate model of configuration space is constructed. Discovering and exploiting structure in a robot's configuration space enables a utility-guided planner to achieve the performance and reliability required by real-world motion planning. This thesis describes applications of the utility-guided motion-planning framework to multi-query sampling-based roadmap and random-tree motion planning. Additionally, the utility-guided framework is extended to develop a planner that can successfully plan despite inaccuracies in its perception of the environment and to guide further sensing to reduce uncertainty and maximally improve the utility of the path.

The United States Navy is committed to developing technology for an All-Electric Ship (AES) that promises to improve the affordability and capability of its next-generation warships. With the addition of power-intensive 21st century electrical systems, future thermal loads are projected to exceed current heat removal capacity. Furthermore, rising fuel costs necessitate a careful approach to total-ship energy management. Accordingly, the aim of this research is to develop computer tools for early-stage design of shipboard energy distribution systems. A system-level model is developed that enables ship designers to assess the effects of thermal and electrical system configurations on fuel efficiency and survivability. System-level optimization and design exploration, based on these energy system models, is challenging because the models are sometimes computationally expensive and characterized by discrete design variables and discontinuous responses. To address this challenge, a classifier-guided sampling (CGS) method is developed that uses a Bayesian classifier to pursue solutions with desirable performance characteristics. The CGS method is tested on a set of example problems and applied to the AES energy system model. Results show that the CGS method significantly improves the rate of convergence towards known global optima, on average, when compared to genetic algorithms.
text

Digital image acquisition can be a time consuming process for situations where high spatial resolution is required. As such, optimizing the acquisition mechanism is of high importance for many measurement applications. Acquiring such data through a dynamically small subset of measurement locations can address this problem. In such a case, the measured information can be regarded as incomplete, which necessitates the application of special reconstruction tools to recover the original data set. The reconstruction can be performed based on the concept of sparse signal representation. Recovering signals and images from their sub-Nyquist measurements forms the core idea of compressive sensing (CS). In this work, a CS-based data-guided statistical sparse measurements method is presented, implemented and evaluated. This method significantly improves image reconstruction from sparse measurements. In the data-guided statistical sparse measurements approach, signal sampling distribution is optimized for improving image reconstruction performance. The sampling distribution is based on underlying data rather than the commonly used uniform random distribution. The optimal sampling pattern probability is accomplished by learning process through two methods - direct and indirect. The direct method is implemented for learning a nonparametric probability density function directly from the dataset. The indirect learning method is implemented for cases where a mapping between extracted features and the probability density function is required. The unified model is implemented for different representation domains, including frequency domain and spatial domain. Experiments were performed for multiple applications such as optical coherence tomography, bridge structure vibration, robotic vision, 3D laser range measurements and fluorescence microscopy. Results show that the data-guided statistical sparse measurements method significantly outperforms the conventional CS reconstruction performance. Data-guided statistical sparse measurements method achieves much higher reconstruction signal-to-noise ratio for the same compression rate as the conventional CS. Alternatively, Data-guided statistical sparse measurements method achieves similar reconstruction signal-to-noise ratio as the conventional CS with significantly fewer samples.

Many animals use vision to guide their behaviour and to collect relevant information about their environment. The diversity of visual environments and of visually guided tasks has led to a large variety of specialisations of eyes and visual systems. Our knowledge, however, about how the anatomical and physiological properties of eyes and the behavioural strategies of animals relate to the visual signals that are important to them in their natural environment, is extremely limited. In this thesis, I make use of optical, physiological and behavioural analyses to reconstruct the flow of visual information that the fiddler crab Uca vomeris experiences during its daily life on the mudflat. I present a detailed analysis of the first stage of visual processing, the sampling by the ommatidial array of the crabs' compound eye and demonstrate how regional specialisations of optical and sampling resolution reflect the information content and behavioural relevance of different parts of the visual field. Having developed the first intracellular electrophysiological preparation in fiddler crabs, I then examine the spectral sensitivities of photoreceptors - the basis for colour vision. I show that the crabs possess an unusual trichromatic colour vision system featuring a UV-sensitive and a variety of short-wavelength receptor types based on the coexpression of two short-wavelength sensitive pigments. Finally, the natural visual signals that predatory and non-predatory birds present to fiddler crabs are described. The visual cues the crabs use when deciding whether and when to respond to these potential predators are analysed and compared to those used in dummy predator experiments. The crabs use a decision criterion that combines multiple visual cues - including retinal speed, elevation and visual flicker. Neither of these cues accurately predicts risk, but together they reflect the statistical properties of the natural signals the crabs experience. The complex interactions between the design of the crabs' visual system, the stimuli they experience in their natural context and their behaviour demonstrate that neither of them can be understood without knowledge of the other two.
Research School of Biological Sciences (RSBS, now RSB), and the Australian National University for providing funding through an ANU PhD scholarship; the Australian Department of Education, Employment and Workplace Relations for an International Postgraduate Research Scholarship; the German National Academic Foundation and the Zeiss Foundation for support through a Heinz-Dürr Scholarship; and the Australian Institute of Marine Sciences for providing accommodation and facilities during fieldwork in Queensland.

In many computer vision applications, the task of robustly estimating the set of parameters of a geometric model is a fundamental problem. Despite the longstanding research efforts on robust model fitting, there remains significant scope for investigation. For a large number of geometric estimation tasks in computer vision, maximum consensus is the most popular robust fitting criterion. This thesis makes several contributions in the algorithms for consensus maximization. Randomized hypothesize-and-verify algorithms are arguably the most widely used class of techniques for robust estimation thanks to their simplicity. Though efficient, these randomized heuristic methods do not guarantee finding good maximum consensus estimates. To improve the randomize algorithms, guided sampling approaches have been developed. These methods take advantage of additional domain information, such as descriptor matching scores, to guide the sampling process. Subsets of the data that are more likely to result in good estimates are prioritized for consideration. However, these guided sampling approaches are ineffective when good domain information is not available. This thesis tackles this shortcoming by proposing a new guided sampling algorithm, which is based on the class of LP-type problems and Monte Carlo Tree Search (MCTS). The proposed algorithm relies on a fundamental geometric arrangement of the data to guide the sampling process. Specifically, we take advantage of the underlying tree structure of the maximum consensus problem and apply MCTS to efficiently search the tree. Empirical results show that the new guided sampling strategy outperforms traditional randomized methods. Consensus maximization also plays a key role in robust point set registration. A special case is the registration of deformable shapes. If the surfaces have the same intrinsic shapes, their deformations can be described accurately by a conformal model. The uniformization theorem allows the shapes to be conformally mapped onto a canonical domain, wherein the shapes can be aligned using a M¨obius transformation. The problem of correspondence-free M¨obius alignment of two sets of noisy and partially overlapping point sets can be tackled as a maximum consensus problem. Solving for the M¨obius transformation can be approached by randomized voting-type methods which offers no guarantee of optimality. Local methods such as Iterative Closest Point can be applied, but with the assumption that a good initialization is given or these techniques may converge to a bad local minima. When a globally optimal solution is required, the literature has so far considered only brute-force search. This thesis contributes a new branch-and-bound algorithm that solves for the globally optimal M¨obius transformation much more efficiently. So far, the consensus maximization problems are approached mainly by randomized algorithms, which are efficient but offer no analytical convergence guarantee. On the other hand, there exist exact algorithms that can solve the problem up to global optimality. The global methods, however, are intractable in general due to the NP-hardness of the consensus maximization. To fill the gap between the two extremes, this thesis contributes two novel deterministic algorithms to approximately optimize the maximum consensus criterion. The first method is based on non-smooth penalization supported by a Frank-Wolfe-style optimization scheme, and another algorithm is based on Alternating Direction Method of Multipliers (ADMM). Both of the proposed methods are capable of handling the non-linear geometric residuals commonly used in computer vision. As will be demonstrated, our proposed methods consistently outperform other heuristics and approximate methods.
Thesis (Ph.D.) (Research by Publication) -- University of Adelaide, School of Computer Science, 2018