Active Selection of Training Examples for Meta-Learning
HIS '07 Proceedings of the 7th International Conference on Hybrid Intelligent Systems
Computers and Electrical Engineering
Particle Swarm Model Selection
The Journal of Machine Learning Research
Distributed modular toolbox for multi-modal context recognition
ARCS'06 Proceedings of the 19th international conference on Architecture of Computing Systems
Hi-index | 0.00 |
The ongoing rapid advance in integration technology combined with the emergence of new sensory elements, computing nodes, and wireless communication delivers unprecedented possibilities for the design and application of intelligent sensor systems. However as the complexity of the designs increase, numerous constraints have to be met. In particular, accuracy and dependability are crucial issues for viable sensory systems. Commonly, similar to the development in the vivid vision market, systems for new applications and novel sensors still have to be designed manually from scratch by expert designers in a tedious process. This requires skills on sophisticated methods ranging from conventional signal processing to computational intelligence. Embedded platforms additionally impose constraints of resource-limitations and reliability issues on the design process. Further, multi-sensor systems with heterogeneous physical principles and dynamics require a rich versatility of methods and calibration techniques. In this research work an emerging tool for automated constrained design is presented along with our methodology for the multi-sensor domain. It will be demonstrated for extrinsic design using gas sensor data in this paper. This will also be extended to the issues of intrinsic optimisation for the compensation of instance related static and environment-induced dynamic deviations. In the latter case, adaptation methods will move on board of the embedded/integrated solution to achieve so called self-x features like, self-monitoring, -trimming, and -repair/healing on various levels of abstraction, seamlessly connecting to reconfigurable and evolvable hardware platforms.