The investigator and her colleagues organize a workshop on learning theory to stimulate interactions between the disciplines of probability, approximation theory, and numerical algorithms as they affect the development of learning theory and learning algorithms. The workshop emphasizes the following topics: concentration of measure inequalities and their application to probability bounds in learning, kernel methods of approximation and their role in support vector machines and other learning algorithms, and fast algorithms for the sparse approximation of high-dimensional data.
Understanding intelligence and how it manifests itself in learning and in assimilating information is one of the great scientific challenges. It is a key to designing systems that efficiently analyze data and extract essential information. The scientific discipline that studies this aspect of intelligence is called learning theory. Central tasks in analyzing data are regression, which describes the relationship between variables whose sampled values are part of the data (for example, the locations and speeds of an object and the times when these are observed), and data classification, which categorizes each element of the data set in terms of the values of components of the datum. This workshop emphasizes particular aspects of learning theory: probabilistic bounds to help select good representations of scattered data, particular approximation methods and their relation to a general family of classification techniques called support vector machines, and fast computation methods for the sparse approximation of high-dimensional data. The workshop advances the nation's capabilities in scientific learning theory by bringing together researchers from diverse disciplines to consider these topics. Learning theory sits at a crossroad between aritificial intelligence, statistical inference, and mathematical data analysis. It has a myriad of existing and potential applications in both the defense and civilian sectors. Typical of these are the navigation of unmanned vehicles, the fast classification of data for such applications as homeland security, and modeling the human visual system.
