소프트웨어학부는 외부연사 초청 세미나를 다음과 같이 안내하오니 관심있는 분들의 많은 참여 바랍니다.
- Using Causal Inference to Test Systems that are “Hard To Test”
Speaker: Prof. Neil Walkinshaw (University of Sheffield, UK)
Time: Fri 15 May, 14:00-15:30
Location: Room 332 in S4-1
Abstract
Over the past five years we have, as part of our EPSRC CITCOM project,
focussed on testing systems with particular characteristics that render them
“hard to test”. These can involve large numbers of parameters with limitations on
the ability to control inputs, behaviour can be non-deterministic or subject to
hidden factors, executions can take a long time and require significant
computational resources. Examples include scientific software models,
automated driving systems, and cyber-physical systems such as an artificial pancreas.
With such systems, it can be particularly challenging to check whether or not
the behaviour that is observed from executions is aligned with expectations or
requirements.
Our research has been based on the observation that these problems are not
unique to testing. They are particularly prevalent in fields such as Epidemiology,
where determining the efficacy of some medical treatment can be fraught with
similar problems. Treatments can be expensive to administer, circumstances
can be hard to control, and they can take a long time to take effect.
To address these problems, a family of statistical techniques known as Causal Inference
has risen to prominence. These techniques provide a framework within which it is
possible to combine traditional statistical inference techniques with fragments of
relevant domain knowledge to draw reliable conclusions about causal relationships
from data.
In this presentation I will show how we have applied Causal Inference to software
(and cyber-physical) systems testing.
문의 : hongshin@gmail.com
