System Identification using CMAC

Artificial neural networks [6] give the advantage of execution improvement through learning using parallel and distributed processing. These networks are carried out using huge connections among processing units with changeable brawn, and they are enticing, of applications in System Identification and control [1]. It uses a neural network called CMAC [4], which stands for Cerebellar Model Arithmetic Computer. The CMAC net appertains for system identification. This network is second to inseminate multi-layer networks in real-time applications. CMAC has properties of conjecture, rapid algorithmic estimation based on LMS [2] training, practical representation, and output incrustation. CMAC network has hundreds of thousands of convertible weights that can be direct to imprecise nonlinearities which are not written out or foggiest idea. CMAC can learn nonlinear association from a very comprehensive category of functions. CMAC is an associative memory [3] that has an integral conjecture. The architecture of CMAC is alike to that of the cerebellum (a part of the brain).

System Identification complications can be observed as delineate betwixt the inputs and outputs of the identification block. The identification block lay hold of the current input and output variables of the process as its inputs and gives the estimates of the system parameters. The ply of CMAC network in understanding the required delineate. Its precedence in system identification of both linear and nonlinear systems mentioned [5].

The simulated network is carried out to approximate the damping coefficient of a nonlinear pendulum subjected to changeable driving force and viscous friction. The network is imitated for dissimilar learning rates and different number of training data points [7]. It is appearing that CMAC network can appear for dynamic systems too. The network meets smoothly if the learning rate is average. The network’s staging is supercilious when the training data points are analogously distributed over the whole input scale instead of random points.

 

 

 

 

 

 

 

 

 

 

REFERENCES

 

  1. G. Horvath, R. Dunay, B. Pataki: “Recurrent CMAC: A Powerful Neural network for System Identification” IEEE Instrumentation and Measurement Technology Conference Brussels, Belgium, June 4-6, 1996

 

  1. Ali Ozen: “A novel variable step size adjustment method based on channel output autocorrelation for the LMS training algorithm” International Journal Of Communication Systems Int. J. Commun. Syst. 2011; 24:938–949

3.      A.J. Krijgsman: ” Associative Memories: The CMA approach ” Lecture Notes Erasmus Intensive Course 1992, Pages 403-421

 

  1. Si-Zhao Qin, Hong-Te Su, T. J. McAvoy: “Comparison of Four Neural Net Learning Methods for Dynamic System Identification” IEEE Trans. on Neural Networks, Vol. 3. No. 1 Jan. 1992. pp. 122- 130.

 

  1. S. Chen, S. Billings, P. Grant: “Non-linear System Identification Using Neural Networks” International Journal of Control Vol. 51.

 

  1. KL Priddy, PE Keller: “Artificial Neural Network: an introduction” SPIE press 2005

 

  1. Sergios Theodoridis, Konstantinos Koutroumbas: “Training Data Points” in Pattern Recognition (Fourth Edition), 2009

 

 

 

Like this article?

Share on Facebook
Share on Twitter
Share on Linkdin
Share on Pinterest

Leave a Comment

Your email address will not be published. Required fields are marked *

Kalinga Plus is an initiative by Kalinga University, Raipur. The main objective of this to disseminate knowledge and guide students & working professionals.
This platform will guide pre – post university level students.
Pre University Level – IX –XII grade students when they decide streams and choose their career
Post University level – when A student joins corporate & needs to handle the workplace challenges effectively.
We are hopeful that you will find lot of knowledgeable & interesting information here.
Happy surfing!!

  • Free Counseling!