The Or Times Vol 1 Issue 3 Fall 2007

Alumni The following is a question and answer session held with Dr Rajesh Ganesan, an alumna of IMSE-USF. Dr. Ganesan, graduated with a Ph.D. in IE from USF in 2005 and is currently an Assistant Professor in the Systems Engineering Department at George Mason University, Faifax, VA. Dr. Das served as his major Advisor during his Ph.D. and M.S. at USF. Dr. Ganesan recently received a $3 Million grant from the NSF’s GK-12 program. During his stay at USF, Dr. Ganesan served as the project manager for our own GK-12 project at IMSE, USF called STARS. Could you describe the kind of research areas you are involved in at GMU? I specialize in stochastic control. The most interesting aspect of this research area is its wide spectrum of methods and numerous applications. Process control is a crucial aspect of many engineering systems such as air transportation, robotics and nano-manufacturing. Depending on the context, the control problem can be perceived as a sequential decision making process in a stochastic environment. Of particular interest to me is the control of real-world problems that are large-scale, non-linear, complex, adaptive, and stochastic for which process models do not exist. The challenge in these problems is to design innovative model-free methods to predict uncertainties and find solutions that provide the end user with optimal sequential decisions (control laws), which adapt to the changing environment. Two of my major research application areas include control of semiconductor manufacturing processes, and sequential decision making to effectively manage air traffic flow. In my research, such sequential decision making or control problems are cast in the framework of stochastic dynamic programming, and solved using artificial intelligence. Particularly, the solution strategy involves a method known as reinforcement learning, which falls under the class of approximate dynamic programming methods. I have termed such solution algorithms as Intelligent Decision Support (IDS) algorithms.

Describe briefly your doctoral dissertation work. My dissertation research was focused on testing a machine learning based control strategy. The control problem is modelled as a Markov decision process (MDP), and solved using a stochastic approximation method known as reinforcement learning (RL). The approach also incorporates data filtering using wavelet based multiresolution analysis to extract significant process deviations. The controller's performance was tested on a multivariate chemical mechanical planarization (CMP) process of semiconductor wafer polishing. Results showed that the RL based controller outperforms the EWMA based controllers for strongly autocorrelated processes, and disturbances like temporary changes in the mean. The innovative part of this research lies in harnessing the potentials of wavelets and reinforcement learning to enhance the design and use of model-free control systems.

See Alumna on page 2.

Committee column Outsourcing: Is it necessary for a business? Outsourcing has, in the last couple of decades, become a global phenomenon. To a layman, outsourcing means taking away jobs that could be performed by “our own country men”. Lately, to a politician, outsourcing-bashing has become a vote hunting technique. Outsourcing refers to the divesture of non-core operations from internal production to an external entity that specializes in the management of these operations. Outsourcing utilizes experts from outside the entity to perform specific tasks that the entity once performed itself. It is supposed that over the past decade, America has lost an average of 7.71 million jobs every quarter to outsourcing. The most alarmist prediction of jobs lost to outsourcing, by Forrester Research, estimates that 3.3 million service jobs will be outsourced between 2000 and 2015—an average of 55,000 jobs outsourced per quarter, or only 0.71 percent of all jobs lost per quarter. This may sound as pronouncing doom to the labor market. But this leads to the questions: “how much of a factor is outsourcing to the countries employment rate? Is this effect negative or positive?” Reports show that the household employment survey of Americans indicates that there are 1.9 million more Americans employed since the recession ended in November 2001. There are 138.3 million workers in the U.S. economy today— more than ever before. We may infer here that both outsourcing and employment rates have been on the increase in the past decade1. The following are some of the reasons for outsourcing: 1. Outsourcing means economies of scale to both the outsourcer and the outsourced. 2. Sharing of risks, also coined as portfolio effects. A business entity can add leverage to the portfolio by outsourcing the riskfree asset. Markowitz, an influential economist won the 1990 Nobel Prize in Economics in this area. 3. Accommodation of peak loads: Businesses are advised to protect their staff from the fluctuations caused by the peaks and valleys in demand, by staffing the valleys and contracting the peaks. Other advantages include reducing the lead time, access to a larger talent pool, commodification and operating across time zones which ensure 24-7 service provision. 1.http://www.heritage.org/Research/TradeandForeignAid/wm467.cfm#_ftn4.

By Wilkistar Otieno (Ph.D. Student, IMSE) Volume 1, Issue 3 Fall 2007

e: s issu i h t e odels Insid ediction m r Survival p

vent Syste Discrete E

A Newsletter from Student Chapter of INFORMS @ IMSE, USF

ling ms Mode

News Feature The Case for Survival Prediction Models Put yourself in the following situation, you are facing a terminal illness, God (or whoever you believe in or don’t believe in) forbid, and you have the chance of being informed with a 95% confidence or higher about how much longer you will live. If I am asked to choose, I would prefer to know and have time to share with my loved ones and maybe spend some potential savings (if the disease has not left me bankrupt) traveling or doing something that I had always wanted to do. On the other hand, I have heard people saying that they would prefer not to know, and thus, avoid the associated anxiety and stress.

ing heavily in the subjective opinion of physicians). Nowadays, sick people are living longer and the survival variability within patients suffering from the same illness cannot be entrusted to physicians who usually based their judgment in the last case they have seen or in the most common case they have seen. How can we, Industrial Engineers, help? The availability of big databases, namely Cancer Registries at state level and the latest developments regarding the identification of genetic markers for a host of diseases provide a favorable backdrop to make use of data mining techniques to build prediction models based on significant factors (demographic, treatment-related, behavioral, genetic, etc) related to survival for each disease.

Whatever your opinion is, there are more people interested in accurately knowing when you will be passing away in case you got a terminal illness. Who? You are probably wondering. Well, the answer is hospice providers and the usual suspect, the government (which is not a bad thing). The New York Times on November 27th reports that hundreds of hospice providers across the country are facing “the catastrophic financial consequence (a 60 patient hospice had to pay the government $900,000) of what would otherwise seem a positive development: their patients are living longer than expected”.

The road is still bumpy though, databases do not provide all the measurements you would desire, in particular for some interesting biomarkers, and even though there is consensus between practitioners about some factors related to survival for particular diseases, these is no agreement for some others. But, that is exactly the reason they need Industrial Engineers. Therefore, an accurate survival predictive model besides adding an extra decision for terminal patients to make (about knowing or not how longer they will live), it will also help hospice providers and The government-sponsored Medicare Hospice Pro- the government to reassess their budget and make the gram was originally designed for those with less than six necessary adjustments. By Patricio Rocha (Ph.D. Stumonths to live (these six months calculated using actuarial dent, IMSE) survival according to the type of disease and probably relySome tips while interviewing on-site for an academic position:

Alumni (Continued from Page 1) Based on your experience, what courses do you think IMSE Ph.D. students must take from outside our department that will serve them well in the future as academicians or in Industry?

Read the faculty pages before going and connect with the individual faculty. Ask about their research and express your interest to collaborate by finding a common ground between your research and theirs during the conversation. If you know that there is nothing in common then talk about The courses must cover both depth and breadth. Some the university, cost of living, the local school district, comcourses are needed for research purpose and one should go mute etc. Find out from the dept chair about the goals of the into depth in such courses. Courses must cover significant department (new research areas) and any concerns such breadth such as core IE, statistics, mathematics, computer languages, economics (particularly if you are dealing with cost- as enrollment, funding that are presently there. Also check benefit analysis or pricing in your dissertation), and those spe- how supportive the dean is to your new department. Ask cific to research (application-domain-related such as OR, Bio, the dean about tenure process and what his/her advise finance, etc..). would be to a new faculty. Ask about the vision that the dean has for the school and for your dept. Many times the Other than courses, is there something you wish you interview includes dinner(s). Don't go overboard on alcolearnt as a student? holic beverages. It's better to abstain or limit to just a bit of I learnt Matlab which is an important scientific computing it. Your 1 hour presentation is key. Keep some time for Q language, I wished I had also learnt a few others such as C++ and A. Talk about future research goals, teaching, funding and JAVA. I also wish that in addition to the nanomanufactursources for your research and possible collaborators in your ing applications which I studied in my dissertation, I should slides. The research part should be a good balance behave also simultaneously developed at least two other applicatween descriptive and mathematical aspects. tion domains for my research methods. It is possible that the university that hires you may not have the entire necessary lab infrastructure and often you will have to find new application arrears that are of national interest.

Responses for this column were solicited by Vishnu Nanduri (Ph.D. Candidate, IMSE, USF) via email. INFORMS USF would like to thank Dr. Rajesh Ganesan for his contribution to the OR Times.

2

Research Corner Discrete Event System Modeling of Demand Responsive Transportation Systems Demand responsive transportation (DRT) is a variable route service of passengers or freight from specific origin(s) to destination(s) in response to the request of users. Typically, DRT systems are considered complimentary to the existing public transportation systems where passengers from lower density areas use DRT service as transportation to transit centers or to other transfer stations. Other highly dynamical and complex applications of DRT cover the domains of dynamically dispatching of cargo trucks, chartered planes, and courier services. Examples of mission critical applications of DRT are transportation of people with disabilities, military aero-medical evacuation of patients to medical treatment facilities and routing vehicles in large-scale emergencies. DRT operational planning encompasses the methods to provide efficient service to the passengers and to the system operators. These methods cover the assignments of vehicles to transportation requests under various constraints such as environmental conditions, traffic limitations, preferences of the passengers, and operation limitations. Recent approaches of DRT operational planning are based on “closed information loop” and achieve a higher level of automation, increased flexibility and efficiency. Advance in the information and communication technologies, such as the Internet, mobile communication devices, GIS, GPS, Intelligent Transportation Systems has led to a significantly complex and highly dynamical decision making environment. The online service allows real time information gathering and concurrent communication of the customers with several vehicles. These technological advances change the manner in which DRT is planned where passengers’ assignments to the vehicles and the fleet’s routing are made in real time. Intelligent and effective use of the available information in such complex decision making environment requires the use of formal modeling and control approaches which are robust, modular, and decentralized. In my research we propose the representation of DRT systems as a Discrete Event System (DES) where the model captures both the low level dynamics (such as infrastructure conditions, current status of vehicles) and high level dynamics (such as service demand requests) of system evolution in a modular manner. The mathematical foundation of DES Theory facilitates logical analysis of these complex systems and provides the necessary framework for the development of planning tools for real time scheduling and decision making. This study is focused in the application of Supervisory Control Theory based on Finite Automata in DRT real time planning. The developed approach is capable in finding the non-blocking behavior of a DES that represents DRT operation. The algorithm is based on the following three groups of elements: plant – the model of DRT system to be controlled, specifications – the constraints of the passengers’ and fleet’s behaviors, and synthesis of supervisory controller – the required sequence of events of the desired system operation. Two Case Studies are presented based on air-taxi service operation and on emergency aero-medical evacuation. Centralized, modular and decentralized supervisory control architectures are developed with discussions for deriving inferences for obtaining real time solutions. By Daniel Yankov (Ph. D. Student, Daniel is advised by Dr. Ali Yalcin). Column solicited and organized by Diana Prieto, Ph.D. student, IMSE.

IMSE Students at the INFORMS annual conference in Seattle, WA, Nov 3-7, 2007

L to R: Ozan Ozcan (Ph.D. student, IMSE), Dr. Kingsley Reeves (Asst. Prof, IMSE), Arka Bhattacharya, and Swati Verma (IMSE students)

Wilkistar Otieno (Ph.D student IMSE)

3

Did you Know??? USF Academic Computing provides students with high performance computing resources for research purposes. Along with benefits in computing some of the advantages I experienced are: submitting multiple jobs, parallel or series, saves me a lot of time; 50GB storage for students (100 GB for faculty) lets me process my large data files with ease; does not keep my desktop tied up; and access to many software. Information on getting an account and using the resources is provided at the ‘Resource Computing’ website https://rc.usf.edu/ (Contributed by Chaitra Gopalappa Ph.D. Student, IMSE)  US Dollars are not made out of paper, they are made out of cotton.  The Declaration of Independence was written on hemp (marijuana) paper.  The dot over the letter I is called a "tittle."  A raisin dropped in a glass of fresh champagne will bounce up and down continuously from the bottom of the glass to the top.  There are no clocks in Las Vegas gambling casinos.  Guinness Book of Records holds the record for being the Book most often stolen from Public Libraries.” Source: http://www.emmitsburg.net/humor/archives/interesting_facts/interesting_facts_11.htm

Volume 1, Issue 3

1. 2. 3. 4. 5. 6.

JOB POSTINGS

FACULTY POSITIONS at Department of Systems and Industrial Engineering The University of Arizona FACULTY POSITION in the Department of Engineering Management and Systems Engineering --University of Missouri - Rolla FACULTY POSITIONS in the Grado Department of Industrial and Systems Engineering VIRGINIA TECH FACULTY POSITION IN ENGINEERING MANAGEMENT Department of Engineering Management and Systems Engineering The School of Engineering and Applied Science of The George Washington University FACULTY OPENING Rochester Institute of Technology Industrial and Systems Engineering Department (ISE) FACULTY POSITION McCormick School of Engineering and Applied Sciences Department of Industrial Engineering and Management Sciences Northwestern University

For details and more job see http://informs.eng.usf.edu/jobs.htm (compiled by Vishnu Nanduri)

Fall 2007

PUZZLE The committee and members of

Which one is correct; your mind or your calculator? 1. Take 1000 and add 40 to it. 2. Now add another 1000 and then add 30. 3. Add another 1000 and add 20. 4. Now add 1000 and another 10.

INFORMS-USF student chapter wish you happy holidays and a prosperous

2008

Is the answer 4100, or is it 5000? Contributed by Wilkistar

Diana Prieto Laila Cure Publicity

Vishnu Nanduri President Patricio Rocha Vice President Wilkistar Otieno Ozan Ozcan Treasurers Athina Brintaki Secretary Chaitra Gopalappa Dayna Martinez Logistics

Andres Uribe Webmaster

IMSE 4202 E. Fowler Ave. ENB 118 Tampa FL, 33620 Tel: (813) 974-5591 Fax: (813) 974-5953 [email protected] 4

Alcides Santander Shaoqiang Chen Social activities Wilkistar Otieno Laila Cure Editors