SMILEARN: Introduction

From DSL

One of the most challenging tasks related to graphical models is the step of creating them. In general, two distinct approaches are possible: a model can be created by a human expert, or created automatically from a database. In practice, usually some combination of these two approaches is used, for example the causal structure of a model is acquired from an expert, while the numerical parameters of the model are learned from a database. GeNIe, which is a graphical user interface to SMILE, contains a number of tools that are designed to help building models by an expert. SMILEARN, described in this document, provides tools for building models from data.

This document describes the application programming interface (API) to the SMILEARN – a specialized module of SMILE* that provides learning and data mining functionality. SMILE is a platform independent library of C++ classes for reasoning in graphical probabilistic models, such as Bayesian networks and influence diagrams. SMILEARN extends functionality provided by SMILE by providing a set of specialized classes that implement learning algorithms and other useful tools for automated building graphical models from data. We assume that the reader has reasonable level of familiarity with graphical models, data mining, and machine learning. Foundations of graphical models are covered in GeNIe and SMILE manuals.

* For more information on the original SMILE please refer to the SMILE On-line Documentation.