Knowledge acquisition planning for inference from large databases

Abstract

It is shown that by combining techniques from machine learning and robot planning, knowledge acquisition (KA) planning provides a framework for automatically applying diverse and complex analytical tools to extremely large collections of data. KA planning is the process of automatically combining inferential tools such as induction, search, database lookup, and statistical analysis into methods for addressing complex query statements. This process depends on having both a domain model that supports subgoal decomposition and a library of KA actions annotated with the type and form of required input data, expected outcomes of the action, estimates of computational resources needed, and so on. When analytical tools are annotated in this way, it is possible to automatically select and deploy them and combine their results, even if they run on incompatible platforms. KA planning was implemented in a program that learned about diagnosing lung tumors. A larger project is underway in the domain of molecular biology.< <ETX xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">&gt;</ETX>