On the surface, the three on-line machine learning problems analyzed in this thesis may seem unrelated. The first is an on-line investment strategy introduced by Tom Cover. We begin with a simple analysis that extends to the case of fixed-percentage transaction costs. We then describe an efficient implementation that runs in time polynomial in the number of stocks. The second problem is k-fold ...

We propose the usefulness of probabilistic relational methods for modeling user behavior at web sites. Web logs (aka "click streams"), server logs, and other data sources, taken as datasets for traditional machine learning algorithms, violate the iid assumption of most algorithms. Requests ("clicks") are not independent within a session, sessions for a visitor are not independent of one another...

RFID applications usually rely on RFID deployments to manage high-level events. A fundamental relation for these purposes is the location of people and objects over time. However, the nature of RFID data streams is noisy, redundant and unreliable and thus streams of low-level tag-reads can be transformed into probabilistic data streams that can reach in practical cases the size of gigabytes in ...

We present an approach to alphabet-independent gestural recognition which differs from existing techniques such as curve-matching in that it attempts to approximate the functionality of the feature-based approaches while allowing the advantages which result from alphabet independence. Our approach uses a probabilistic model to analyze an alphabet based on user supplied samples and to effectivel...

The execution of tasks for a mobile robot embedded in a dynamic environment brings about several challenges, due to the dynamic changes of the environment and the inaccurate perception of the robot. This paper tackles the problem of on-line execution monitoring when the agent has different tasks and several plan to accomplish them, as in the BDI framework. Our method considers uncertainty in th...

One of the most important goals of computer vision research is object recognition. Most current object recognition algorithms assume reliable image segmentation, which in practice is often not available. This research exploits the combination of the Hough method with the geometric hashing technique for model-based object recognition in seriously degraded intensity images. We describe the analys...

 Many efficient interior-point methods (IPMs) are based on the use of a self-concordant barrier function for the domain of the problem that has to be solved. Recently, a wide class of new barrier functions has been introduced in which the functions are not self-concordant, but despite this fact give rise to efficient IPMs. Here, we introduce the notion of locally self-concordant barrier functio...