Machine Learning

Machine learning is a key technology for realizing computers with "intelligent" capabilities, namely, machines that do not just execute fixed programs, but can OBSERVE data, LEARN from data (change their behavior) and MAKE DECISIONS based on the learned information. Learning capabilities are the basis for many new application domains envisioned for computers, such as user-intent-aware services or conversational computer-user interfaces. Machine learning is also an essential component in making systems more robust against malfunctions and malicious attacks.

Machine learning has made great theoretical and practical strides over the last 20 years. Among the most notable is the development of Support Vector Machine (SVM), a universal learning algorithm for generalization in high dimensional spaces. It has proven to be very effective on a number of problems, such as data mining or text analysis. However, further advancement in the current state of the art is necessary to be effective on more complex real-world problems that mandate quick and accurate learning abilities.

We take three approaches for developing powerful learning computers:

• Advance the theoretical understanding of learning through the analysis of new induction principles and through a paradigm shift from traditional inductive inference to non-inductive inference such as transductive and selective inference.


• Accelerate the computation by orders of magnitude through the development of more efficient learning algorithms and through more efficient implementations.


• Apply these advances to innovative applications showcasing their performance.