Ill-posed inverse problems have wide applications in many fields such as oceanography, signal processing, machine learning, biomedical imaging, remote sensing, geophysics, and others. In this dissertation, we address the problem of solving unstable operator equations with iteratively regularized Newton-type algorithms. Important practical questions such as selection of regularization parameters, construction of generating (filtering) functions based on a priori information available for different models, algorithms for stopping rules and error estimates are investigated with equal attention given to theoretical study and numerical experiments.

In a stopping rule problem, a real-time player decides to stop or continue at stage n based on the observations up to that stage, but in a k-step look-ahead stopping rule problem, we suppose the player knows k steps ahead. The aim of this Ph.D. dissertation is to study this type of prophet problems for simple random walk, determine the optimal stopping rule and calculate the expected return for them. The optimal one-step look-ahead stopping rule for a finite simple random walk is determined in this work. We also study two infinite horizon stopping rule problems, sum with negative drift problems and discounted sum problems. The optimal one, two and three-step look-ahead stopping rules are introduced for the sum with negative drift problem for simple random walk. We also compare the maximum expected returns and calculate the upper bound for the advantage of the prophet over the decision maker. The last chapter of this dissertation concentrates on the discounted sum problem for simple random walk. Optimal one-step look-ahead stopping rule is defined and lastly we compare the optimal expected return for one-step look-ahead prophet with a real-time decision maker.

No description available.

本論文共一冊，分八章八節。 內容：本論文所擬探討之對象為多變量統計分配函數模擬（Simulation）之最佳停止 法則問題（Optimal Stopping Rule Problem ），此類問題之目的在於如何利用盡量 小的樣本數之觀察值來求得未知母數（Unknoron Parameter）的信區間（域）（Co－ nfidence interval ）（Confidence Region），而此信賴區間（域）之寬度（Width ）及包含機率（Coverage Probability）均已事先指定。 以往研究對象多傴限於單變量統計分配函數，而多變量統計分配函數模擬之最佳停止 法則問題，仍尚在研究階段，因此本論文之重點乃在於探討如何求得滿足最佳停止法 則之最小樣本數。在此以多變量常態分配函數為重心，並進而嗜試推廣至其他多數量 統計分配函數。

Neste trabalho consideramos três tópicos principais. Nos dois primeiros generalizamos alguns dos resultados clássicos da Teoria da Confiabilidade na otimização dos procedimentos de burn-in e de políticas de garantia, respectivamente, sob o modelo de tempo de vida geral, quando um sistema coerente é observado ao nível de seus componentes, e estendemos os conceitos de intensidade de falha na forma de banheira e do modelo de falha geral através da definiçâo de processos progressivamente mensuráveis sob a pré-t-história completa dos componentes do sistema. Uma regra de parada monótona é usada na metodologia de otimizaçâo proposta. No terceiro tópico modelamos os custos de garantia descontados por reparo mínimo de um sistema coerente ao nível de seus componentes, propomos o estimador martingal do custo esperado para um período de garantia fixado e provamos as suas propriedades assintóticas mediante o Teorema do Limite Central para Martingais. / In this work we consider three main topics. In the first two, we generalize some classical results on Reliability Theory related to the optimization in burn-in procedures and warranty policies, using the general lifetime model of a coherent system observed on the component level and extending the definitions of bathtub shaped failure rate and general failure model to progressively measurable processes under the complete pre-t-history. A monotone stopping rule is applied within the proposed methodology. In the third topic, we define the discounted warranty cost process for a coherent system minimally repaired on the component level and we propose a martingale estimator to the expected warranty cost for a fixed period and setting its asymptotic properties by means of Martingale Central Limit Theorem.

Neste trabalho consideramos três tópicos principais. Nos dois primeiros generalizamos alguns dos resultados clássicos da Teoria da Confiabilidade na otimização dos procedimentos de burn-in e de políticas de garantia, respectivamente, sob o modelo de tempo de vida geral, quando um sistema coerente é observado ao nível de seus componentes, e estendemos os conceitos de intensidade de falha na forma de banheira e do modelo de falha geral através da definiçâo de processos progressivamente mensuráveis sob a pré-t-história completa dos componentes do sistema. Uma regra de parada monótona é usada na metodologia de otimizaçâo proposta. No terceiro tópico modelamos os custos de garantia descontados por reparo mínimo de um sistema coerente ao nível de seus componentes, propomos o estimador martingal do custo esperado para um período de garantia fixado e provamos as suas propriedades assintóticas mediante o Teorema do Limite Central para Martingais. / In this work we consider three main topics. In the first two, we generalize some classical results on Reliability Theory related to the optimization in burn-in procedures and warranty policies, using the general lifetime model of a coherent system observed on the component level and extending the definitions of bathtub shaped failure rate and general failure model to progressively measurable processes under the complete pre-t-history. A monotone stopping rule is applied within the proposed methodology. In the third topic, we define the discounted warranty cost process for a coherent system minimally repaired on the component level and we propose a martingale estimator to the expected warranty cost for a fixed period and setting its asymptotic properties by means of Martingale Central Limit Theorem.

在生態學與生物學上，物種數常作為生物多樣性的指標，以估計單一群體物種數為例，較知名的方法首推Good (1953)以在樣本中出現一次的物種為基礎，提出的物種數估計方法堪稱的先驅，隨後許多文獻延伸Good的想法，發展出許多的估計方法，例如Burham and Overton (1978)的摺刀估計法，Chao and Lee (1992)則以涵蓋機率方式估計。相對而言，兩群體的共有物種數的研究少有人探討，目前以Chao et al. (2000)的估計式較為知名。 本研究參考Good (1953)提出估計未發現物種出現機率的想法，估計未發現共有物種的機率，並以Burham and Overton (1978)中應用摺刀法估計物種數的概念，建立一階摺刀估計式與變異數，且另行以多項分配公式推導變異數估計式，進行電腦模擬與實際資料驗證並與Chao et al. (2000)提出的共有物種估計式比較。最後根據Rasmussen and Starr (1979)以抽樣成本建立最適停止規則的概念，應用於本研究所提出的估計式，並經由電腦模擬找出抽樣成本與物種分佈均勻程度的關聯，可作為設定停止規則的依據。 / The number of species is often used to measure the biodiversity of a population in ecology and biology. Good (1953) proposed a famous estimate for the number of species based on the probability of unseen species. Subsequently, many studies applied Good’s idea to create new estimation methods, For example, the Jackknife estimate by Burham and Overton (1978), and the estimate by using the sample coverage probability in Chao and Lee (1992) are two famous examples. However, not many studies focus on estimating the number of shared species of two populations, except the method by Chao et al. (2000). In this study, we modify Good’s idea and extend the Jackknife method of Burham and Overton (1978) to develop the estimate for the number of shared species of two populations. In addition, we also establish the variance formula of the estimator by using the multinomial distribution. Subsequently, we use computer simulation and real data sets to evaluate the proposed method, and compare them with the estimator by Chao et al. (2000). Finally, we adapt the idea of optimal stopping rule by Rasmussen and Starr (1979) and combine it with the proposed jackknife estimate. We found that using the sampling cost as the stopping rule is a feasible approach for estimating the number of shared species.

Gegenstand dieser Dissertation ist die non-parametrische Schätzung der Kovarianz in multi-dimensionalen Lévy-Prozessen auf der Basis von Hochfrequenzbeobachtungen. Im ersten Teil der Arbeit wird eine modifizierte Version der von Jacod und Reiß vorgeschlagenen Methode der Hochfrequenzbeobachtung für die Ermittlung der Kovarianz multi-dimensionaler Lévy-Prozesse gegeben. Es wird gezeigt, dass der Kovarianzschätzer optimal im Minimaxsinn ist. Darüber hinaus demonstrieren wir, dass die Indexaktivität der co-jumps durch das harmonische Mittel der Sprungaktivitätsinzidenzen der Komponenten von unten beschränkt wird. Der zweite Teil behandelt das Problem der adaptiven Schätzung. Ausgehend von einer Familie asymptotischer Minimax-Schätzer der Kovarianz, erhalten wir einen datenbasierten Schätzer. Wir wenden Lepskii’s Methode an, um die Kovarianz an die unbekannte Aktivität des co-jumps Indexes des Sprungteils anzupassen. Da wir es mit einem Adaptierungsproblem zu tun haben, müssen wir eine Schätzung der charakteristischen Funktion des multi-dimensionalen Lévy-Prozesses konstruieren, damit die charakteristische Funktion weder von einer semiparametrischen Annahme abhängt noch schnell abfällt. Aus diesem Grund wird auf Basis von Neumanns Methode ein trunkierter Schätzer für die empirische charakteristische Funktion konstruiert. Die Anwesenheit der trunkierten, empirischen charakteristischen Funktion im Zähler führt jedoch zu einer Situation, die auch bei der Deconvolution auftritt, d.h. einem irregulären Verhalten des stochastischen Fehlers. Dieser U-förmige stochastische Fehler verhindert die Anwendung von Lepskii’s Grundsatz. Um diesem Problem, entgegenzuwirken, entwickeln wir eine Strategie, welche zu einem Orakelstart von Lepskii's Methode führt, mit deren Hilfe ein monoton steigender stochastischer Fehler konstruiert wird. Dies erlaubt uns, ein Balancing Principle einzuführen und einen adaptiven Schätzer für die Kovarianz zu erhalten, der fast-optimale Raten erzeugt. / In this thesis, we consider the problem of nonparametric estimation for the continuous part of the covariation of a multi-dimensional Lévy process from high-frequency observations. This continuous part of covariation is also called covariance. The first part modifies the high-frequency estimation method, proposed by Jacod and Reiss, to cover estimation of the covariance of multi-dimensional Lévy processes. The covariance estimator is shown to be optimal in the minimax-sense. Moreover, the co-jump index activity is proved to be bounded from below by the harmonic mean of the jump activity indices of the components. In the second part, we address the problem of the adaptive estimation. Starting from an asymptotically minimax family of estimators for the covariance, we derive a data-driven estimator. Lepskii's method is applied to adapt the covariance to the unknown co-jump index activity of the jump part. Faced with an adaptation problem, we need to secure an estimation for the characteristic function of the multi-dimensional Lévy process so that it does not depend on a semiparametric assumption and, at the same time, does not decay fast. For this reason, a truncated estimator for the empirical characteristic function is constructed based on Neumann's method. The presence of the truncated empirical characteristic function in the denominator leads to a situation similar to the deconvolution problem, i.e., an irregular behavior of the stochastic error. This U-shaped stochastic error does not permit us to apply Lepskii's principle. To counteract this problem, we establish a strategy to obtain an oracle start of Lepskii's method, according to which a monotonically increasing stochastic error is constructed. This enables us to apply a balancing principle and build an adaptive estimator for the covariance which obtains near-optimal rates.