This course is composed of lectures and exercises. The lecture part first introduces linear programming problem, the most fundamental mathematical optimization problem. Then the lecture part overviews the simplex method and theoretical aspects of linear programming. The topic for nonlinear optimization problems includes the optimality conditions, the steepest descent method, the interior-point method. In the exercise part, students apply the simplex method to linear programming problems for understanding the computational steps of the simplex method, and give some parts of mathematical proof on optimization methods.
Mathematical optimization is a mathematical approach to find an optimal candidate from the set of candidates that satisfy specific conditions. It is closely related to scientific problems and it is widely employed to solve practical problems. For example, a diet problem in which we want to find a food recipe that minimizes the calories satisfying enough nutrients can be considered as an example of mathematical optimization problems. In this course , students learn both theoretical and computational aspects of optimization methods.
At the end of this course, students will be able to:
(1) Solve linear programming using the simplex method
(2) Understand theoretical properties of linear programming, for example, the duality theorem
(3) Understand the relation between optimal solutions and optimality conditions
(4) Explain the framework of numerical methods to solve nonlinear optimization problems, for example, the steepest decent method and the interior-point method
Linear programming, Simplex methods, Duality theorem, Sensitive analysis, Shortest path problem, Maximum flow problem, Convex function, Optimality condition for nonlinear optimization methods, Karush-Kuhn-Tucker condition, Steepest descent method, Newton method, Successive quadratic method, Interior-point method.
Intercultural skills | Communication skills | ✔ Specialist skills | Critical thinking skills | ✔ Practical and/or problem-solving skills |
The lecture part overviews numerical methods and theoretical aspects. In the exercise part, exercise problems are assigned. Students apply the simplex method to linear programming problems by hand, and give a proof of optimization methods. For each exercise class, students submit reports.
Course schedule | Required learning | |
---|---|---|
Class 1 | Overview of mathematical optimization, Linear programming | Formulate some problems using the standard form of linear programming |
Class 2 | Simplex method | Solve linear programming problems by the simplex method. |
Class 3 | Exercise: Linear programming and Simplex method | Solve exercise problems related to linear programming and the simplex method |
Class 4 | Bland's rule, Two-phase simplex method | Apply Bland's rule and/or two-phase simplex method to linear programming problems. |
Class 5 | Dual problem, Duality theorem | Derive a dual problem from a linear programming problem. Derive a theoretical properties from weak duality theorem. |
Class 6 | Exercise: Two-phase simplex method and Duality theorem | Solve exercise problems related to two-phase simplex method and duality theorem |
Class 7 | Complementary theorem, Sensitivity analysis | Derive an optimal solution using the complementary theorem. Apply the simplex method in a matrix style. |
Class 8 | Shortest path | Analyze the sensitive of linear programming problems. Apply Dikstra's method to obtain the shortest path. |
Class 9 | Exercise: Complementary theorem and Shortest path | Solve exercise problems related to the complementary problem and shortest path |
Class 10 | Maximum flow problem and question time | Apply the Ford-Fulkerson algorithm to solve the maximum flow. |
Class 11 | Exercise problems to evaluate achievements and review the first part of the course | Evaluate the understanding on the first part of the course. |
Class 12 | Exercise: Maximum flow problem | Solve exercise problems related to maximum flow problems |
Class 13 | Nonlinear optimization, Convex set, Convex function | Formulate some problems as nonlinear optimization problems. Understand the definitions of convex sets and convex functions. |
Class 14 | Steepest descent method | Derive a proof on the convergence of the steepest descent method |
Class 15 | Exercise: Convexity and Optimality conditions | Solve exercise problems related to convexity and optimality conditions |
Class 16 | Newton method, Karush-Kuhn-Tucker condition | Explain the framework of Newton method. Derive relations between the KKT condition and optimal solutions. |
Class 17 | Lagrange function, Duality theorem | Derive relations between Lagrange function and the KKT condition. |
Class 18 | Exercise: Steepest descent method and Karush-Kuhn-Tucker condition, Duality theorem | Solve exercise problems related to steepest descent method, Karush-Kuhn-Tucker condition and duality theorem |
Class 19 | Successive quadratic method | Explain the framework of the successive quadratic method |
Class 20 | Interior-point method | Explain the framework of the interior-point method. |
Class 21 | Exercise: Successive quadratic method | Solve exercise problems related to successive quadratic method |
Class 22 | A review of this course | Re-solve all the excercise problems |
None required. Parts of the course materials are based on the reference books below.
・ Mathematical optimizaiton, Takahito Kuno, Maiko Shigeno, Junnya Goto, Ohmsha, 2012 (in Japanese)
・ Optimization Methods, Akihisa Tamura, Masakazu Muramatsu, Kyouritsu-shuppan, 2002 (in Japanese)
・ Linear and Nonlinear Optimization," Igor Griva, Stephen G. Nash, Ariela Sofer, SIAM,
2009
・ Linear Programming, Vasek Chvatal, Freeman, 1983
・ Linear Optimization, Glenn H. Hurlbert, Springer, 2010
・ Introductory Lectures on Convex Optimizaiton, Yurii Nesterov, Kluwer Academic Publishers, 2004
Students will be assessed on their understanding on the simplex method for linear programming problems, duality theorem, network optimization problems, and numerical methods and optimality conditions for nonlinear optimization problems.
Students' course scores are based on the final exam (80%) and exercise reports (20%).
[In the case that the final exam is not available, a final report will be used instead.)
Students must have studied Linear Algebra and Its Applications (MCS.T203) or have equivalent knowledge.