반복영역 바로가기
주메뉴로 바로가기
좌측메뉴로 바로가기
본문으로 바로가기

Home News&ActivitySeminars

Seminars

프린트페이스북

(대학원생 마일리지 적용) 학과 세미나 안내 (3/25(월), Taewoo Lee at University of Houston / Seminar Notice: 3/25 at 4:00pm)

2019.03.20 14:25

교수님 및 학생분들께

 

안녕하세요

학과 세미나가 다음과 같이 진행될 예정입니다.

 

1. 일시: 3 25(월) 오후 4

 

2. 장소: 산업경영학동(E2-2) 지하 계단강의실(B105)

 

3. 주제Addressing Model-Data Fit in Optimization: Inverse Optimization Approach    

 

4. 연사Taewoo Lee, Assistant Professor at University of Houston    

 

5. 언어: 미정

 

많은 관심과 참석 부탁드립니다.

 

감사합니다.

 

 

 

----------------------------

 

Dear ISysE Professors and Students,

 

ISysE department invites you to attend the following seminar.

 

Speaker: Taewoo Lee, Assistant Professor at University of Houston     

Title: Addressing Model-Data Fit in Optimization: Inverse Optimization Approach     

 

Date & Time: March 25th(Mon), 4:00 pm

Place: E2-2 Bldg, #B105

Language: TBA

 

We look forward to your attendance and encourage you to forward this invitation to colleagues who may be interested in the topic.

           

With best regards,

 

 

 

 

----------------

 

 

Given observations from a system, inverse optimization determines parameter values for an optimization model that describe the system dynamics and are consistent with the original observations, which then can be used for future decision making. In classical inverse optimization, observations are typically assumed to represent optimal system behaviors. However, data is often imperfect and noisy, so there is no guarantee this assumption is satisfied. We develop a unified inverse optimization framework that can accommodate noisy data and closely replicate the system, and highlight the role of inverse optimization as a model-fitting tool. We derive a closed-form solution for this framework and discuss its geometric interpretations. Inspired by its connection to regression, we propose a goodness-of-fit metric for inverse optimization, termed the coefficient of complementarity, which shares similar properties with R2 in linear regression and thus is an intuitive, interpretable measure for model-data fit. The proposed framework of model estimation and evaluation will be demonstrated in various application domains.

List