Skip to main content
Przejdź do strony domowej Komisji Europejskiej (odnośnik otworzy się w nowym oknie)
polski pl
CORDIS - Wyniki badań wspieranych przez UE
CORDIS
Zawartość zarchiwizowana w dniu 2024-05-29

Scalable Online Learning Systems

Cel

Learning to input data into a set of classes, given a set of classification examples, is the main subject to a multidisciplinary field of machine learning. The problem can be formulated as a minimization of a given cost function over possible classification functions. The properties of learning systems crucial for a successful application are: performance, i.e. high accuracy of classification, which depends on how well the used cost function corresponds to the application at hand, scalability, i.e ensuring that memory and time complexity of the learning grows gracefully with data size, and ability to process examples online as they come.

The objectives of this proposal are twofold. First, I intend to develop scalable systems that learn online and use structured (hence more natural) costs. These theoretical advances will facilitate development of learning-based systems for various applications. Second, I intend to apply these new learning methods in computer security (Intrusion Detection Systems) and bioinformatics (DNA splice site detection). To achieve the first goal, I will design new learning algorithms able to optimise structured costs common in non-Bayesian decision- making. I will build on recent methods from the Support Vector Machines learning, which transform the task to a Quadratic Programming (QP) optimisation.

The main idea of the proposed methodology is to exploit algorithms from computational geometry to derive online QP optimisation able to process large-scale data. A key to application of the pro posed algorithms to intrusion detection and splice site detection problems is understanding the problem-specific semantic constraints imposed by these applications. To achieve the second goal, I will incorporate the available semantic models into the learning algorithms, building on a large previous experience of the host institution (FhG-FIRST) in these respective problems.

Dziedzina nauki (EuroSciVoc)

Słowa kluczowe

Temat(-y)

Zaproszenie do składania wniosków

FP6-2005-MOBILITY-5
Zobacz inne projekty w ramach tego zaproszenia

System finansowania

EIF - Marie Curie actions-Intra-European Fellowships

Koordynator

FRAUNHOFER-GESELLSCHAFT ZUR FORDERUNG DER ANGEWANDTEN FORSCHUNG E.V.
Wkład UE
Brak danych
Koszt całkowity
Brak danych