This thesis describes a theory for decision-making in a dynamic purchasing environment where one of possibly many bundles of items must be purchased from possibly many suppliers. The online combinatorial purchasing problem is defined as the problem with which a purchase agent in such an environment is faced when deciding which combination of items, from whom and at what time to buy in order to ...

Several ecological factors that could play into species extinction are expected to correlate with species age, i.e., time elapsed since the species arose by speciation. To date, however, statistical tools to incorporate species age into likelihood-based phylogenetic inference have been lacking. We present here a computational framework to quantify age-dependent extinction through maximum likeli...

This paper addresses the problem of nonparametric estimation of the conditional expected shortfall (CES) which has gained popularity in financial risk management. We propose a new nonparametric estimator of the CES. The proposed estimator is defined as a conditional counterpart of the sample average estimator of the unconditional expected shortfall, where the empirical distribution function is ...

A new algorithm named EXPected Similarity Estimation (EXPoSE) was recently proposed to solve the problem of large-scale anomaly detection. It is a non-parametric and distribution free kernel method based on the Hilbert space embedding of probability measures. Given a dataset of n samples, EXPoSE needs only O(n) (linear time) to build a model and O(1) (constant time) to make a prediction. In thi...

This paper presents a new class of variable-structure algorithms, referred to as expected-mode augmentation (EMA), for multiple-model estimation. In this approach, the original model set is augmented by a variable set of models intended to match the expected value of the unknown true mode. These models are generated adaptively in real time as (globally or locally) probabilistically weighted sum...

While neural machine translation (NMT) has made remarkable progress in translating a handful of resource-rich language pairs recently, parallel corpora are not always readily available for most language pairs. To deal with this problem, we propose an approach to zero-resource NMT via maximum expected likelihood estimation. The basic idea is to maximize the expectation with respect to a pivot-to...

In real-world applications, motion and sensor observations of mobile robots are typically subject to noise and errors. A common way to deal with this problem is to represent the states of a robot at different points in time with probability distributions. Being able to calculate the expected distributions of the states of the robot even before the robot starts operation is useful for many robot...