IT 879

(maybe renumbered as IT 735 or OR 735)

Advanced Stochastic Simulation

Spring 2007 (maybe, please come back and check)

Important Announcements & Deadlines

Instructor: Dr. Chun-Hung Chen
Email: cchen9@gmu.edu
Office: Science & Tec II, Room 319
Phone: 703-993-3572
Fax: 703-993-1521
Office Hours: Monday & Wednesday 4:30 - 5:30 PM

Course Description:

This class basically is an advanced version and an extension of the basic simulation class OR 635 Discrete System Simulation.  We will cover some important topics which OR 635 does not have time to include. In particular, the focus this semester will be on "efficient simulation-based decision making".

Simulation is a popular tool for designing large, complex, stochastic systems, since closed-form analytical solutions generally do not exist for such problems.  While the advance of new technology has dramatically increased computational power, efficiency is still a big concern when using simulation for stochastic optimization, in which case many alternative designs must be simulated.  A decision maker is forced to compromise on simulation accuracy, modeling accuracy, and the optimality of the selected design.  This class will discuss different approaches to address this issue.  Students will have to investigate and/or develop efficient simulation-based optimization techniques in the term projects.

Prerequisite: Students in the this class are assumed to have the background of an introductory simulation class such as OR 635 Discrete System Simulation, or permission from the instructor.

Grading: Homework 15%; Special Topic Presentation 25%; Project Proposal 5%; Project Presentations 17%; Term Project Report 32%; Class Participation 5%.

Required Text: A. M. Law & W. D. Kelton, "Simulation Modeling & Analysis," 2000 (same as OR 635)

General Rules:

  1. Late homework and term project report is always allowed. No need to get advanced permission. However, the penalty for late homework and term project report is 25% for the first day and then 5% per day. No exemption.
  2. Turning in HW through email is subject to a 20% penalty.
  3. No collaborations are allowed for homework, although discussions are encouraged.
  4. Comments are strongly encouraged.
  5. No cheating.

Course Outline (will be completely revised)

1. Introduction of Simulation:

Give an overview of stochastic simulation by quickly reviewing the materials covered in OR 635.

2. Efficient Simulation Techniques and Optimal Computing Budget Allocation (OCBA):

We will review some well-know techniques for improving simulation efficiency, such as importance sampling. Another focus is the OCBA which is a new control-theoretic simulation technique invented by Dr. Chen.  OCBA advances the state-of-the-art by optimally allocating a computing budget to the candidate alternatives under evaluation.  Intuitively, to ensure that the best alternative is correctly selected, a larger portion of the computing budget should be allocated to those alternatives that are critical in the process of identifying the best alternative.  Overall simulation efficiency is improved as less computational effort is spent on simulating non-critical alternatives.  In particular, OCBA is employed to provide the smallest number of simulation runs for a desired accuracy by not only considering variances of different alternatives - as the standard theory now provides - but also their relative profiles.  Numerical testing demonstrates that OCBA is about 3 to 10 times faster than traditional techniques for problems with 10 to 100 alternative designs.

3. Efficient Simulation-based Decision Making:

Some useful optimization techniques will be introduced. Examples include genetic algorithm, nested partition, coordinate search, and dynamic programming.  The goal herein is to help students develop efficient simulation-based optimization techniques for their projects.

4. Special Topic Study & Presentation

Each student has to select a topic to study and present in the class (30 minutes).  The presentation will be graded by the instructor and the class.  Please identify a paper which can give a good introduction and overview of the topic you study, and email the paper to the class at least one week before your presentation. You can also consider to send another paper which give more in-depth discussions.  Some possible topics are list below; but not limited to the list.  Good resources for selecting a topic are our text book and the Winter Simulation Conference Proceedings)

·        Markov Chain Simulation

·        Petri Net

·        Agent-based Modeling & Simulation

·        Advanced Random Number Generation and/or Testing

·        Advanced Input/Output Analysis

·        Variance Reduction Techniques

·        Importance Sampling

·        Rare Event Simulation

·        Validation & Verification

·        Fluid Dynamic Simulation

·        Experimental Design

·        Quasi Monte Carlo Simulation

·        Latin Hypercube Sampling

·        Parallel Simulation

·        HLA

·        Advanced Event Simulation Methodologies

Please email your presentation to the instructor at least 24 hours in advance.  In your presentation, please consider to include the following items:

·        Introduction

·        Basic ideas & fundamentals

·        What are the strengths?

·        What are the weaknesses?

·        Where and how to apply?

5. Term Project and Presentation: Development of Efficient Simulation-based Optimization Techniques:

Students are expected to investigate a technique for efficient simulation-based decision making. One possible approach is to integrate the efficient simulation techniques (given in topic 2) with optimization methodology (given in topic 3).  Students have to meet with the instructor personally in the projects to ensure right progress and discuss potential research questions. Students will give presentations to the class about their techniques at the end of the semester.

Homework Assignments & Others:

Useful Links:

Go to Professor Chun-Hung Chen's Page