Using neural networks for quality management

We present a method for fine grain QoS control of multimedia applications. This method takes as input an application software composed of actions. The execution times are unknown increasing functions of quality level parameters. Our method allows the construction of a quality manager which computes adequate action quality levels, so as to meet QoS requirements for a given platform. These include requirements for safety (action deadlines are met) as well as optimality (maximization and smoothness of quality levels). In this paper, we use learning techniques for computation of quality management policies. Given input parameters of the actions, a neural network is used to refine online pre-computed average execution times. Using refine average execution times allows a better control of the application, which leads to a reduction of fluctuations of CPU load. We present experimental results including the implementation of the method and benchmarks for an MPEG4 video encoder.

[1]  Jukka Saarinen,et al.  Initializing Weights of a Multilayer Perceptron Network by Using the Orthogonal Least Squares Algorithm , 1995, Neural Computation.

[2]  Dennis Shasha,et al.  Skip-Over: algorithms and complexity for overloaded systems that allow skips , 1995, Proceedings 16th IEEE Real-Time Systems Symposium.

[3]  Joseph Sifakis,et al.  QoS control for optimality and safety , 2005, EMSOFT.

[4]  Chenyang Lu,et al.  Feedback control real-time scheduling , 2001 .

[5]  Daniel P. Siewiorek,et al.  A scalable solution to the multi-resource QoS problem , 1999, Proceedings 20th IEEE Real-Time Systems Symposium (Cat. No.99CB37054).

[6]  Alan Burns,et al.  Mechanisms for enhancing the flexibility and utility of hard real-time systems , 1994, 1994 Proceedings Real-Time Systems Symposium.

[7]  John P. Lehoczky,et al.  Algorithms for scheduling hard aperiodic tasks in fixed-priority systems using slack stealing , 1994, 1994 Proceedings Real-Time Systems Symposium.

[8]  Joseph Sifakis,et al.  Fine grain QoS control for multimedia application software , 2005, Design, Automation and Test in Europe.

[9]  Geoffrey E. Hinton,et al.  Learning representations by back-propagating errors , 1986, Nature.

[10]  Kurt Hornik,et al.  Approximation capabilities of multilayer feedforward networks , 1991, Neural Networks.

[11]  P. Werbos,et al.  Beyond Regression : "New Tools for Prediction and Analysis in the Behavioral Sciences , 1974 .

[12]  Yann LeCun,et al.  Une procedure d'apprentissage pour reseau a seuil asymmetrique (A learning scheme for asymmetric threshold networks) , 1985 .

[13]  Sang Hyuk Son,et al.  Feedback Control Real-Time Scheduling: Framework, Modeling, and Algorithms* , 2001, Real-Time Systems.

[14]  Kurt Hornik,et al.  Multilayer feedforward networks are universal approximators , 1989, Neural Networks.

[15]  Reinder J. Bril,et al.  QoS Control Strategies for High-Quality Video Processing , 2004, ECRTS.

[16]  Aggelos K. Katsaggelos,et al.  A Review of the Minimum Maximum Criterion for Optimal Bit Allocation Among Dependent Quantizers , 1999, IEEE Trans. Multim..

[17]  Joseph Sifakis,et al.  Using Speed Diagrams for Symbolic Quality Management , 2007, 2007 IEEE International Parallel and Distributed Processing Symposium.

[18]  Alan Burns,et al.  Scheduling slack time in fixed priority pre-emptive systems , 1993, 1993 Proceedings Real-Time Systems Symposium.

[19]  Giuseppe Lipari,et al.  Elastic task model for adaptive rate control , 1998, Proceedings 19th IEEE Real-Time Systems Symposium (Cat. No.98CB36279).

[20]  Edwin D. Mares,et al.  On S , 1994, Stud Logica.