Vira Chankong
Associate Professor, Department of Electrical, Computer and Systems Engineering
Develops rapid learning and optimization methods for intensity modulated radiation therapy, and develops new paradigm for combinatorial optimization
Education
Ph.D., Systems Engineering, Case Western Reserve University, 1977
M.S., Systems Engineering, Asian Institute of Technology, 1974
B.E., Electrical Engineering, University of New South Wales, 1971
Awards and Recognitions
2018,
Srinivasa P. Gutti Memorial Teaching Award Nominee,
Ohio Alpha chapter of Tau Beta Pi National Engineering Honor Society
2016,
Nominee for the John S. Diekhoff award,
CWRU
2011,
Nominee for the Diekoff Award ,
CWRU
2005,
Nominee of CWRU Wittke Award ,
CWRU
1997,
Semi-Finalist, INFORMS Franz Edelman Award ,
INFORMS
Research Interests
Current interests are theory and algorithms for large-combinatorial optimization and application to important problems in energy (Smart grid planning and Management), and technology for healthcare (compressing sensing and intensity modulated radiation therapy.<br>1. My student (Natachai Naviroj) and I have developed a system for warehouse-managed inventory of multiple products for multiple retailers’ problem faced by department store chains and/or grocery chains. These problems are characterized by a large number of products (in tens of thousand) and complicated supply chain and logistics. To solve the MILP problems of such magnitude, decomposition and relaxation are necessary, a specialized bounded knapsack solver is required, and a network-based optimization method is needed to handle the associated logistics and vehicle routing subproblems. A decomposition framework based on Lagrangian relaxation, a specialized algorithm for solving bounded knapsack subproblems, and a network-based method to solve the warehouses-to-retailers ordering and shipment subproblems have all been developed and tested. The integrated methodology has now been applied to management replenishment ordering and delivery of products to the grocery stores chains owned by the MALL conglomerate in Thailand. We believe a warehouse-managed inventory and logistics system such as this is the way of the future in the supply chain management of large retailer chains with a very large number of products.<br>2. I have served as a consultant in an NIH 4-year R01 project entitled Knowledge-based Models to Enable Rapid Learning in Radiation Therapy; with the PI--Professor Q Jackie Wu of the Department of Radiation Oncology, Duke Medical Center, Duke University. The goal of this project is to mine enormous data of past patients from several major medical centers and combine those with modern-day treatment planning methods so that a personalized IMRT treatment plan for a particular patient can be developed rapidly and more effectively. This will help improve the prognosis with regards to the survival rate and quality of life of cancer patients undergoing radiation treatment. My role as a consultant is to assist in the knowledge-based modeling of a bi-objective IMRT fluence map optimization (determining how the effective dose can be delivered to the planned target volume to kill cancer cells, while minimizing toxicity to organs-at-risk (OARs) and surrounding normal tissues), and to develop a system for a rapid generation a set of Pareto-optimal treatment plan with key trade-off information to allow clinician to quickly choose the best compromised plans to treat the patient. <br>3. During the last few years, I have concentrated on developing an algorithm to solve the “0-1 minimax IP” master problem of Bender Decomposition in an attempt to remove the main stumbling block for using Bender Decomposition as the new paradigm for solving general mixed 0-1 integer programming problems. I have succeeded in developing a branch-and-bound based algorithm which is as efficient as one can make it without incorporating any “cuts”. We have completed stress tests on the algorithm using randomly-generated large-scale problems as well as problems from publically available libraries of MILP test problems. We have found that the algorithm works very efficiently and quickly when the numbers of variables and constraints are 40 or less. This means that for larger problems, a suitable decomposition scheme and/or incorporation of appropriate “cuts” would be needed. This will be one of the most interesting challenges to pursue. These are precisely the tasks that I will continue to investigate.<br>4. In the past few years, I have been working on developing a battery of algorithms to solve the subset sum problem with a dual aim (i) to resolve the P v.s. NP question, and (2) to combine it with various decomposition and relaxation techniques to develop a new paradigm for solving integer programming problems that does not rely too heavily on cuts generation and branch and bound-based procedures. This is the challenge that I will continue to try to tackle.Teaching Interests
Main teaching interests are optimization, dynamic optimization, large-scale optimization, and computational engineering such as simulation and computational mathematics. Can also teach Signals and Systems, Engineering Economics, and Systems Thinking and Systems Approach to Problem SolvingConsulting
2016 -
PRESENT,
Duke Medical Center, Duke University