Variable batch sizes for Learning Feed Forward Control using Key Sample Machines

Kamphuis, Job (2005) Variable batch sizes for Learning Feed Forward Control using Key Sample Machines.

[img]
Preview
PDF
1MB
Abstract:Recently a new algorithm for function approximation in LFFC has been developed, called ´key sample machine´ (KSM). This algorithm has a number of advantages compared to algorithms like ´ordinary least squares´, ´dual least squares´ or ´support vector machines´. KSM´s excel especially in using less computing recourses while being more or about as accurate in their approximations. The KSM algorithm has two variants; one approximates the function recursively (´online´), while the ´offline´ KSM needs all the data in advance. The main goal of this project is to develop a KSM that forms the midst between the ´offline´ and ´online´ algorithms. At first some data will be collected and given to the KSM to be able to make a rough approximation much alike the offline variant. This approximation is used as a starting point for the recursive part of the algorithm. The expected advantages of this approach are that there is probably less computing power needed and the chances of incorrect steering behaviour caused by bad starting approximations will be significantly smaller. To test the new KSM variant, it will be applied in learning feed forward controller that controls a physical system.
Item Type:Essay (Master)
Faculty:EEMCS: Electrical Engineering, Mathematics and Computer Science
Subject:53 electrotechnology
Programme:Electrical Engineering MSc (60353)
Link to this item:http://purl.utwente.nl/essays/56915
Export this item as:BibTeX
EndNote
HTML Citation
Reference Manager

 

Repository Staff Only: item control page