САНУ - Зоран Обрадовић

Speaker:

Zoran Obradovic, L.H. Carnell Professor of Data Analytics Director, Data Analytics and Biomedical Informatics Center, Professor, Computer and Information Sciences Department, Professor, Statistics Department, Fox School of Business Temple University, Philadelphia, PA, USA.

Abstract:

First, we will show how to improve the representational power of Gaussian Conditional Random Field (GCRF) model for structured regression by (1) introducing an adaptive feature function that can learn nonlinear relationships between inputs and outputs and (2) allowing the weights of feature functions to be dependent on inputs. Experimental evaluation on the remote sensing problem of aerosol estimation from satellite measurements and on the problem of document retrieval showed that the proposed model is more accurate than the benchmark alternatives. Then, we will describe how we used GCRF to estimate unreported hospital charges by utilizing structured regression on a temporal graph of more than 4,000 hospitals observed over 8 years constructed from the US National Inpatient Sample database. The estimates of cost-to-charge ratio obtained using convex optimization of the GCRF parameters on the constructed graph were much better than those relying on group average based cost-to-charge estimates. In addition, cost-to-charge ratio estimates by our GCRF model outperformed regression by nonlinear artificial neural networks.

The first result is obtained in joint research with V. Radosavljevic and S. Vucetic and will be published at the Proc. European Conf. on Machine Learning and Principles and Practice of Knowledge Discovery in Databases, Nancy, France, September, 2014. The second result is obtained jointly with A. Polychronopoulou and will be published at the Proc. 2014 IEEE Int’l Conf, on Bioinformatics and Biomedicine, Belfast, UK, Nov. 2014.