We found a match
Your institution may have access to this item. Find your institution then sign in to continue.
- Title
A new approach to risk sensitivity.
- Authors
BACZYNSKI, JACK
- Abstract
A new concept of risk sensitivity is devised, unveiling a class of functions not detected as risk sensitive in the classical scenario. It allows extending known risk sensitive control problems and provides a new pattern concerning the format of the problems to come. We show that the classical concept is insensitive to risk over time and that this may lead to critical exposures regardless how strong the controller's risk aversion is. The new concept, instead, fully covers this aspect. Relations between the classical concept and the corresponding restriction of the new concept are established. Particularly, in the important affine case, we showed that both concepts are equivalent and, parallel to the classical case, risk sensitivity and convexity are equivalent. We provide examples via richer versions of the linear exponential of quadratic Gaussian optimal control problem and the non-linear partially observed optimal control problem given in James et al.
- Subjects
MATHEMATICAL functions; GAUSSIAN processes; STOCHASTIC analysis; PROBLEM solving; COMPUTATIONAL complexity
- Publication
IMA Journal of Mathematical Control & Information, 2017, Vol 34, Issue 2, p425
- ISSN
0265-0754
- Publication type
Article
- DOI
10.1093/imamci/dnv054