You are here

Modeling and Learning Methods; A Report to the Board of Scientific Counselors

Printer-friendly versionPrinter-friendly version
May 2004 Technical Report to the LHNCBC Board of Scientific Counselors.
Abstract: 

The Modeling and Learning Methods (MLM) project at the Lister Hill Center of the National Library of Medicine (NLM) seeks to develop new modeling methods that enable researchers to rapidly construct effective computational models from large datasets. The objectives of the project are to develop machine learning methods that automate the process of constructing probabilistic models for (1) identifying relevant information among large datasets and corpora, (2) mapping identified information to networks of ontologies, (3) accessing queried information accurately, and (4) answering user queries through mining the data located in heterogeneous information sources. Interest in probabilistic models ranges over a wide spectrum of biomedical fields, including computational biology; bio, clinical, and healthcare informatics; and epidemiology. The objectives of the project will be evaluated with a set of suitable metrics such as receiver operating characteristics (ROC) that measure the performance of prospective models in terms of sensitivity and specificity in reaching their target functions. Depending on the domain of the models and the problems of interest, domain subjects and/or experts might be needed to determine the gold standards or the target functions for the performance evaluations of the models if such gold standards or target functions are not readily available.

Kayaalp M. Modeling and Learning Methods; A Report to the Board of Scientific Counselors May 2004 Technical Report to the LHNCBC Board of Scientific Counselors.