Campus Event Calendar
Event Entry
What and Who
Using learning theory in machine learning
Herr Dr. Ingo Steinwart
Los Alamos National Laboratory,USA
MPI-Kolloquium
AG 1, AG 2, AG 3, AG 4, AG 5
AG Audience
Note: We use this to send email in the morning.
Date, Time and Location
Friday, 28 January 2005
16:00
-- Not specified --
45
HS 3
Saarbrücken
Abstract
In this talk I will describe how mathematical learning theory can be used in the understanding and designing of learning algorithms. To this end I will first recall the binary classification problem which is one of the most important learning scenarios in applications. In particular I will describe the learning goals in this scenario and their formalizations. Then I will present an old result of learning theory which essentially states that no learning algorithm can learn all classification problems well. One of the consequences I will then discuss is that establishing learning performance guarantees always require to analyze which implicit assumptions on the data-generating process the algorithm at hand makes. In the second part of the talk I will present a new assumption on the data-generating process and compare this assumption with earlier introduced concepts. Then some corresponding learning performance guarantees for support vector machines, which are the state-of-the-art classification algorithms, will be formulated. In the third part of my talk I will briefly describe how learning theory can be used in the design and analysis of new anomaly detection algorithms. In particular I will present a reliable method which allows us to compare the learning performance of different anomaly detection algorithms.
Contact
--email hidden
passcode not visible
logged in users only
Bahareh Kadkhodazadeh, 01/25/2005 11:02 -- Created document.