Self-adaptation via concurrent multi-action evaluation for unknown context

Nimalasena, A. 2017. Self-adaptation via concurrent multi-action evaluation for unknown context. PhD thesis University of Westminster Engineering https://doi.org/10.34737/q32qv

TitleSelf-adaptation via concurrent multi-action evaluation for unknown context
TypePhD thesis
AuthorsNimalasena, A.
Abstract

Context-aware computing has been attracting growing attention in recent years. Generally, there are several ways for a context-aware system to select a course of action for a particular change of context. One way is for the system developers to encompass all possible context changes in the domain knowledge. Other methods include system inferences and adaptive learning whereby the system executes one action and evaluates the outcome and self-adapts/self-learns based on that. However, in situations where a system encounters unknown contexts, the iterative approach would become unfeasible when the size of the action space increases. Providing efficient solutions to this problem has been the main goal of this research project.

Based on the developed abstract model, the designed methodology replaces the single action implementation and evaluation by multiple actions implemented and evaluated concurrently. This parallel evaluation of actions speeds up significantly the evolution time taken to select the best action suited to unknown context compared to the iterative approach.

The designed and implemented framework efficiently carries out concurrent multi-action evaluation when an unknown context is encountered and finds the best course of action. Two concrete implementations of the framework were carried out demonstrating the usability and adaptability of the framework across multiple domains.

The first implementation was in the domain of database performance tuning. The concrete implementation of the framework demonstrated the ability of concurrent multi-action evaluation technique to performance tune a database when performance is regressed for an unknown reason.

The second implementation demonstrated the ability of the framework to correctly determine the threshold price to be used in a name-your-own-price channel when an unknown context is encountered.

In conclusion the research introduced a new paradigm of a self-adaptation technique for context-aware application. Among the existing body of work, the concurrent multi-action evaluation is classified under the abstract concept of experiment-based self-adaptation techniques.

Year2017
File
PublisherUniversity of Westminster
Digital Object Identifier (DOI)https://doi.org/10.34737/q32qv

Related outputs

Context-aware Approach for Determining the Threshold Price in Name-Your-Own-Price Channels
Nimalasena, A. and Getov, Vladimir 2016. Context-aware Approach for Determining the Threshold Price in Name-Your-Own-Price Channels. in: Context-Aware Systems and Applications Springer. pp. 83-93

Context-Aware Framework for Performance Tuning via Multi-action Evaluation
Nimalasena, A. and Getov, Vladimir 2015. Context-Aware Framework for Performance Tuning via Multi-action Evaluation. in: 2015 IEEE 39th Annual Computer Software and Applications Conference (COMPSAC) Taichung, Taiwan IEEE . pp. 318 - 323

Performance Tuning of Database Systems Using a Context-aware Approach
Nimalasena, A. and Getov, Vladimir 2014. Performance Tuning of Database Systems Using a Context-aware Approach. in: Proc. 9th International Conference on Computer Engineering & Systems (ICCES) Cairo IEEE . pp. 98 - 103

System evolution for unknown context through multi-action evaluation
Nimalasena, A. and Getov, Vladimir 2013. System evolution for unknown context through multi-action evaluation. in: Proceedings of 2013 IEEE 37th Annual Computer Software and Applications Conference Workshops (COMPSACW) IEEE . pp. 271-276

Permalink - https://westminsterresearch.westminster.ac.uk/item/q32qv/self-adaptation-via-concurrent-multi-action-evaluation-for-unknown-context


Share this

Usage statistics

144 total views
184 total downloads
These values cover views and downloads from WestminsterResearch and are for the period from September 2nd 2018, when this repository was created.