24 Nov 1130-1200
IDSIA meeting room, Galleria 1, Manno
Nihat Engin Toklu, Okan University, Computer Engineering
Department, Istanbul, Turkey
(joint work with Seda Yanik, Istanbul Technical
University, Industrial Engineering Department, Istanbul
Robust Traveling Salesman Problem Under Dynamic
In the field of robust optimization, we study problems in
which the problem data are not exactly known (i.e. problem
data are under uncertainty). Considering that multiple (or
infinite) scenarios emerge out of the uncertainty, the
goal is to find a robust solution which does not "go bad"
under most or all scenarios. A very popular approach for
robust optimization was proposed by Bertsimas & Sim (2003,
2004), where the uncertain problem data are expressed by
intervals, and there is a Gamma parameter for tuning the
model's conservativeness (i.e. pessimism: on how much of
the coefficients do we assume unluckiness will happen?).
In this talk, we propose a robust traveling salesman
problem model which can be considered as a generalization
of the study of Bertsimas & Sim. Representing the fact
that the traffic on the paths change from time to time,
our model considers that each path has more than one
uncertainty profile, each profile having a different
activation time, and imposing a different cost interval.
Furthermore, we show that our model does not complicate
the process of tuning in terms of conservativeness,
depending on a single Gamma parameter.
I am pleased to announce to all those interested that tomorrow we will host Dr Verena Tiefenbeck, head of the Bits to Energy Lab at ETH Zürich (http://www.bitstoenergy.ch <http://www.bitstoenergy.ch/>)
She will give a talk titled "Digitalization for behavioral control – an empirical study”, which is scheduled to start at 11:30 in IDSIA’s meeting room in Manno, Galleria 1 building
A short abstract is attached.
Networks of low-cost sensors make it possible to track human behavior wherever we are and whatever we do. While behavior change programs in the past were often designed in an paternalistic spirit of “educating the masses”, technology-based behavioral interventions can tap into different mechanisms and be tailored to the specific context and needs of the individual. An increasing number of studies in the health and energy sector already show that technology-based behavioral interventions can cost-effectively induce behavior change on a large scale. In a prior field study, we have shown that feedback on the resource consumption of a specific activity (showering) induced a 22% conservation effect on the target activity. In a follow-up field study with 720 Swiss households, we now investigate why this technology-enabled approach is so powerful: is the large conservation impact due to a) attention drawn to the energy-intensity of the behavior, b) the correction of misconceptions about one’s environmental impact, c) behavioral control (the assessment of one’s performance against reference points) or d) explicit performance appraisal?