[09] September 13-14 Short Course | Marco Campi: Scenario-Based Optimization and Generalization Theory

Scenario-Based Optimization: Heuristics and Certificates in Decision Making Shortcourse

Instructor: Dr. Marco Campi, University of Brescia (Italy) 

September 13 -14, 2016 

National Institute of Aerospace – Room 101

100 Exploration Way, Hampton, VA 23666



If you are interested in attending this course and/or obtaining more information as it becomes available please contact Mary Catherine Bunde at mary.bunde@nianet.org or by calling (757) 325-6731.

Course Overview:

This course will review scenario-based optimization and its powerful generalization theory. Specific application domains covered include identification, machine learning, and robust control.

Who Should Attend: 

The course will be open to qualified faculty, research staff and graduate students who are both from NASA and non-NASA communities.

Course Outline:

Optimization with uncertain elements is ubiquitous, and application domains range from robust and predictive control to management, from engineering design to quantitative finance. The scenario approach is a general methodology to address uncertain optimization based on previous examples. The scenario solution stands on a solid mathematical footing and one is given certificates that guarantee the achievable performance when the scenario solution is applied to a new, yet unseen, case. In this course, we give an introduction to the scenario approach, followed by some new perspectives and open problems that provide an opportunity of further investigation and research.

Content of the Course:

Section 1

Introduction to scenario optimization:

1.1  Data-based decision making

1.2  Performance guarantees for when the decision is applied to a new case

1.3  Examples

Section 2


2.1 Probability box

2.2 Optimization with constraints. Example: control with saturation limits

2.3 L1 – regularization. Example: regression

2.4 Conditional Value at Risk (CVaR). Examples in finance

Section 3

Beyond the classical set-up:

3.1 The wait-and judge approach. Example: antenna array design.

3.2 Non-convex scenario optimization. Extension of the wait-and-judge philosophy.

Examples: Open problems

3.3 Alternative approaches to non-convex scenario optimization. Examples: non-linear system identification and mixed integer problems for control.

Daily Format:

Course will run 9:00am – 5:00pm both days with breaks and time for lunch (not provided).

Instructor Recommended Hotel:

Courtyard Marriott-Hampton, 1917 Coliseum Drive, Hampton, VA 23666 Ph: (757) 838-3300.

Recommended Airports:

Norfolk (ORF), Newport News (PHF) or Richmond (RIC)