Learning Systems Group

For our latest research reports, see the CNSE research pages.

Here are the two thrusts of our research work.

Learning Systems

Over the years, the group has developed a theoretical framework for automated learning (the bin model), a technique for augmenting the information given to the learning process (learning from hints), and systems based on learning for financial, medical, and industrial applications. We have worked on the theory and techniques of learning from very noisy data, a more difficult but more realistic type of problem. In addition, we worked on a number of collaborative theoretical and practical projects that have a strong learning component. Recently, we started working on the problem of distributed learning. We continue to be interested in the theoretical issues of learning and in applications of learning models and neural networks in the areas of pattern recognition, time series analysis, and system modeling. (Research projects in learning systems)

Computational Finance

Computational finance is a relatively new field, that explores computational and algorithmic methods to solve some of the problems in the field of finance. It is a truly multi-disciplinary topic, since methods from computer science, mathematics, physics etc find rich application there. Financial data are among the most noisy data available, and financial markets offer a large variety of hints for the learning process. Our initial work in computational finance applied learning from hints to the very noisy foreign exchange market. Our current interests cover several markets and techniques. For instance, we are working on equity (stock markets), interest rates, as well as foreign exchange. We are interested in forecasting and arbitrage, calibration of financial models, pricing of financial instruments, portfolio optimization, and analytics for risk management. (Research projects in computational finance)