Hochbaum to research new techniques to enhance monitoring and quality control
July 31, 2018

IEOR Professor Dorit Hochbaum has just been awarded $400,000 by the National Science Foundation to investigate new techniques aiming to reduce manufacturing costs and enhance quality. Besides reducing costs, applications of the research aim to improve societal welfare with broad applications such as enhancing public health, and detecting spam online. 

Title: Prediction, quality control and data analysis graph theoretic techniques in the presence of spatial or temporal dependencies

Abstract: This project will contribute to the advancement of science and will benefit the national prosperity and welfare, by enhancing manufacturing system monitoring and quality control. Expensive high-tech manufacturing processes require early detection of process problems and accurate yield prediction. Early detection allows for a quick diagnosis of the nature of the faults and their correction to enhance quality and minimizing production costs. This project will devise, test, implement and adapt new prediction methods for diverse applications that exhibit spatial or temporal dependencies. The existence of spatial dependencies is not fully utilized in traditional modeling. The improved prediction capability enabled by the new approach of this project is expected to have a beneficial societal impact: enhanced quality and reduced production costs in manufacturing, limiting the spread of disease to improve public health, and enhanced protection of individuals on social networks by detecting patterns of adverse link behavior, such as spam. The fundamental concepts of this work and the new outlooks on prediction approaches will be incorporated into educational course materials. Both undergraduate and graduate students will be involved in the research and implementation in the areas of manufacturing and health care.

The technique proposed utilizes graph theoretic models for predictive estimation where the goal is to minimize a function of penalties assigned to deviating from the priors, which are the observations, and the penalties assigned to the violations of the spatial dependencies. The efficient combinatorial algorithms devised utilize a form of a parametric cut procedure. The proposed work will impact fundamental research on the complexity of the spatial correlation problems and on improvement of algorithms' efficiency for various models of isotonic regression and machine learning. Specific prominent methodology areas that are expected to benefit include: general Bayesian estimation, prediction, pattern recognition under observed spatio-temporal dynamics; signal extraction; functional fitting; feature segmentation in images or videos; inverse optimization and the general area of fidelity-regularization problems.