Autonomic Computing, International Conference on (2005)
June 13, 2005 to June 16, 2005
DOI Bookmark: http://doi.ieeecomputersociety.org/10.1109/ICAC.2005.45
Anton A. Bougaev , Purdue University
Fault detection is one of the important constituents of fault tolerance, which in turn defines the dependability of autonomic computing. In presented work several pattern recognition tools were investigated in application to early fault detection. The optimal margin classifier technique was utilized to detect the abnormal behavior of software processes. The comparison with the performance of the quadratic classifiers is reported. The optimal margin classifiers were also implemented to the fault detection in hardware components. The impulse parameter probing technique was introduced to mitigate intermittent and transient fault problems. The pattern recognition framework of analysis of responses to a controlled component perturbation yielded promising results.
A. A. Bougaev, "Pattern recognition based tools enabling autonomic computing.," Autonomic Computing, International Conference on(ICAC), Seattle, Washington, 2005, pp. 313-314.