MATLAB-based Introduction to Neural Networks for Sensors Curriculum

 

  • Welcome to the webpage of lecture and laboratory on an introduction to Artificial Neural Networks (ANN) that was implemented and assessed in an applications-oriented elective course on smart sensors. The following documents are included:

  • A two class period PowerPoint presentation. The goal of the presentation is to provide basic understanding of the ‘Gradient Descent' back-propagation training algorithm for feed-forward neural networks using step-by-step simple examples with minimal use of mathematical equations.
  • The applications mentioned in the presentation were research projects carried out here at Missouri S&T. The work has been presented at major neural network conferences. Following are the references along with the detailed power point presentations
    • R. Dua, S. E. Watkins, D. C. Wunsch, K. Chandrashekhara, and F. Akhavan, "Detection and Classification of Impact-Induced Damage in Composite Plates using Neural Networks,â€? INNS-IEEE International Joint Conference on Neural Networks, (Mount Royal, NJ: International Neural Network Society) (Washington DC, July 2001) p 51. (ppt1)
    • R. Dua, V. M. Eller, K. M. Isaac, S. E. Watkins, and D. C. Wunsch, "Intelligent Strain Sensing on a Smart Composite Wing using Extrinsic Fabry-Perot Interferometric sensors and Neural Networks,â€? INNS-IEEE International Joint Conference on Neural Networks 2003, 20-24 July 2003, Portland, OR. (ppt2)
  • Homework assignment along with solutions
  • MATLAB® (version 6.1)** based neural network laboratory:
    • The goals of the laboratory are:
      • To understand the working of MATLAB based Graphical User Interface GUI neural network toolbox that involves setting up the network architecture and training it to solve real world problems.
      • To try different network architectures and training algorithms, to solve a simple classification problem.
      • To comment on the performance of the networks trained and results obtained.
    • The laboratory write-up provides a step-by-step procedure to setup, train and simulate a neural network to solve a problem.
    • The laboratory requires the following files. The descriptions of the following files have been included in the write-up.
      • Data files
      • Post-processing code
  • Evaluation for the laboratory. This evaluation will help us to improve the lab for future classes. Please fill out the evaluation and send the evaluation, as a word document attachment, to watkins@mst.edu

** Note: You should have the MATLAB Neural Network Toolbox to do the experiment.