We propose to develop black-box techniques for improving the performance of algorithms by adapting them to the sequence of problem instances on which they are run. In particular, we propose to develop effective strategies for solving the following four problems: 1. Combining multiple heuristics online. Given k (randomized) algorithms, we seek to learn online a policy for interleaving and restar...

The optimal configuration of an online store is a challenging task. Nowadays store managers decide on basis of their expert knowledge. This decision making is insufficient because their choices are often nontransparent and the effects of their decisions are difficult to predict. An example of this problem is the optimal configuration of payment methods that should be offered in an online store....

Bipartite ranking aims to maximize the area under the ROC curve (AUC) of a decision function. To tackle this problem when the data appears sequentially, existing online AUC maximization methods focus on seeking a point estimate of the decision function in a linear or predefined single kernel space, and cannot learn effective kernels automatically from the streaming data. In this paper, we first...

We present a Bayesian approach for online portfolio selection, a fundamental problem in computational finance. We pose the problem as the global optimisation of an expensive, time-varying, black-box function. As the optimum is itself dynamic, we use a model that allows us to capture time-dependent patterns of the function and to provide sequential decision processes that enable us to select opt...

This project surveys some recent algorithms and results in the field of online learning theory [5][6][7]. The problem context is that of an online decision problem, where one must choose a point xt from a fixed set K for each round t. After a choice has been made, a cost function ft(x) for that round is revealed and used to compute the cost of that decision. On the next round t + 1, a new point...

We consider the sequential decision problem known as regret minimization, or more precisely its generalization to the vectorial or multi-criteria setup called Blackwell approachability. We assume that Nature, the decision maker, or both, might have some quitting (or terminating) actions so that the stream of payoffs is constant whenever they are chosen. We call those environments “quitting game...

Feeding is the most important cost in the production of growing pigs and has a direct impact on the marketing decisions, growth and the final quality of the meat. In this paper, we address the sequential decision problem of when to change the feed-mix within a finisher pig pen and when to pick pigs for marketing. We formulate a hierarchical Markov decision process with three levels representing...

In an online decision problem, one makes a sequence of decisions without knowledge of the future. Tools from learning such as Weighted Majority and its many variants [13, 18, 4] demonstrate that online algorithms can perform nearly as well as the best single decision chosen in hindsight, even when there are exponentially many possible decisions. However, the naive application of these algorithm...