Dissertation > Excellent graduate degree dissertation topics show

The Improved Artificial Neural Network Algorithm and It’s Image Compression

Author: LiuXiangYang
Tutor: WangRuYun
School: Hohai University
Course: Applied Mathematics
Keywords: Artificial Neural Networks Relative error Cumulative error BP algorithm Adaptive Image Compression
CLC: TP391.41
Type: Master's thesis
Year: 2003
Downloads: 241
Quote: 2
Read: Download Dissertation

Abstract


Taking into account In many cases, people are more concerned with the relative error of the forecast value and the actual value of the prediction model , this paper, the actual output and the desired output relative error sum of squares as the objective function , given a based on relative error sum of squares and for a minimum of BP algorithm . Taking into account the actual network output values ??ranging between 0 and 1 , the ideal output value for the practical problems are given a standardized approach . Cases test confirmed by a large number of operators in the fitted values ??or predictions based on the square of the relative error and for the inspection standards under the premise , the algorithm obtained better than the traditional BP algorithm, the results based on the absolute error sum of squares as the objective function . Final learning outcomes through the accumulated error to the evaluation of artificial neural network , which aimed directly at the accumulated error to study the multi-layer artificial neural network algorithm to learn quickly . We first briefly trapezoidal descent method based on the cumulative error , on this basis , an adaptive learning rate adjustment programs . After a large number of examples inspection , under the same accuracy requirements , the convergence speed of the proposed algorithm greatly accelerate and effectively overcome the general descent method based on the accumulated error ladder with shock in the learning process . Image compression based on the error back propagation algorithm work for many , but there is the problem of the artificial neural network training time is longer, low precision . Taking into account the use of three and more than three BP network for image compression , the effective information is the connection weights between the output value of the intermediate layer unit and the intermediate layer and output layer , the connection weights of the input layer and the middle layer is redundant that have a negative impact on learning speed and compression quality . Based on this , we propose a new two-story back propagation network topology and algorithms , in order to further improve the compression ratio of the image compression and compression quality , we propose a new three-layer back propagation network topology and algorithms . The machine through the compression test , relative to the three-layer BP network , more than three BP network and nested BP network image compression compression ratio , the learning speed and compression quality has greatly improved , and achieved good results .

Related Dissertations

  1. High Speed Frequency Measurment and Non-Linearity Correction of Frequency Modulated Capacitive Displacement Sensor,TH822
  2. Robust Adaptive Control Algorithms Research for the System of Spacecraft Attitude Dynamics,V448.22
  3. Research on Decoding Algorithm for LDPC Codes,TN911.22
  4. Research on Image Compression and Implementation Using TMS320C6713 Based on SPIHT Algorithm,TP391.41
  5. Video Coding Research Based on Texture Characteristics,TP391.41
  6. Research on Sparse Image Representation and Coding Model,TP391.41
  7. Study on Nonlinear Variable Structure Guidance Law,TJ765
  8. Research on Remote Sensing Image Processing Combining High Fidelity Compression and Resolution Enhancement,TP751
  9. Adaptive Control of Post-Boost Vehicle Using Multiple Models,TP273.2
  10. Design of Gyro Stabilized Pod Control System,TP273
  11. Study on Adaptive Fuzzy Control Algorithm and Its Implementation,TP273.4
  12. Research of Anomaly Detection Algorithms of Hyperspectral Imagery Based on Kernel Method,TP751
  13. Active Power Filter and Its Application in Distribution Network,TN713.8
  14. Research on Adaptive Control of Recovery of UUV with Near Wall Constraint,TP273
  15. Robust Tracking Control for Nonlinear Systems,O231
  16. Spectrum Variation of Vegetation in Yanzhou Coal Mine Area and Heavy Metal Stress Characteristic,X173
  17. Adaptive Adjustment of fire emergency plan,X928.7
  18. Remote areas, distribution transformer monitoring system design theft,TM421
  19. Design and Implication of Image Compression System Based on Wavelet and DM6446,TP391.41
  20. Research of Adaptive Active Noise Control Based on Neural Network,TP183
  21. Research on Object Detection and Tracking Method in Active Vision System,TP391.41

CLC: > Industrial Technology > Automation technology,computer technology > Computing technology,computer technology > Computer applications > Information processing (information processing) > Pattern Recognition and devices > Image recognition device
© 2012 www.DissertationTopic.Net  Mobile