Mathematics Colloquium

Department of Mathematics
University of Nebraska at Omaha

WHEN:
On Friday, April 02, 2004 at 2:30 PM

WHERE:
Durham Science Center, Room 255

WHAT:

ABSTRACT:
Results of statistical analysis of medical data shall model medical thinking, to be acceptable and useful for medical researchers. Induction of statistical dependencies in a logical form is, essentially, a logical problem. However, traditional logics, such as FOL, second order logic or fuzzy logic, are logics of deduction and cannot be used to formulate it. SI logic, which we introduced, is intended to study broad possibilities of inductive learning with ambiguous, redundant and insufficient data, typical for medical problems. One of the features of this multi-sort logic is that it includes two agents: one represents a researcher, with certain language and ability to validate the logical statements on the models, datasets. Another agent signifies nature with the laws in the form of unknown statistical distributions, which define the probability of models in this logic. We define and prove consistency of the induction in this logic as ability to learn these laws with increasing datasets.

    Back to the Mathematics Colloquium Page