readme

很经典的国外模式识别教材MATLAB例程

Thursday, October 26, 1995

README for Neural Network Design Demonstrations


The book "Neural Network Design" is accompanied by 58 demonstrations which can be run from within MATLAB. References to the demonstrations in the book are indicated by the MATLAB icon (a 3-D surface).
 
OVERVIEW OF DEMONSTRATION FILES 
 
The demonstration files consist of two directories, NNDESIGN and MININNET. The first directory NNDESIGN contains all the demonstrations and functions which the demonstrations utilize to do their work. 
 
The second directory MININNET contains a few key functions borrowed from the Neural Network Toolbox (NNT). These functions allow many of the demonstrations to run without the Toolbox. However, you should only install this directory if you do not have the NNT. Having both the Toolbox and the MININNET directory installed may result in unpredictable results due to multiple versions of the borrowed functions. 
 
DEMO REQUIREMENTS 

MATLAB version 4.0 or later, or the student edition of MATLAB version 4.0, should be installed on your hard drive in a directory named MATLAB. To create this directory and complete the installation process, follow the instructions given in the MATLAB documentation. Take care to follow the guidelines for setting the path. 

Many of the demonstrations do not require either the MININNET directory or the Neural Network Toolbox. Some functions require either the MININNET directory or the Toolbox, and a few require the Toolbox. 
 
The last section of this file lists all the demonstrations and indicates the requirements for each. You can see the same list from within MATLAB by typing "help nndesign" after installing the NNDESIGN directory. 
 
RUNNING THE DEMONSTRATIONS 

After the software has been loaded into the MATLAB directory on your computer, you can run the demonstrations directly by typing their names at the MATLAB prompt. Typing help nndesign brings up a list of all the demos you can choose from. 
 
Alternatively, you can run the Neural Network Design splash window (nnd) and then click the Contents button. This will take you to a graphical Table of Contents (nndtoc). From there you can select chapters with buttons at the bottom of the window, and individual demonstrations with popup menus. 
 
SOUND 

Many of the demonstrations use sound. In many cases the sound adds to the understanding of a demonstration. In other cases it is there simply for fun. 
 
If you need to turn the sound off you can give MATLAB the following command and all demonstrations will run quietly: 
 
  nnsound off 
 
To turn sound back on: 
 
  nnsound on 
 
You may note that demonstrations that utilize sound often run faster when sound is off. 
 
KNOWN BUGS

The Neural Network Design Demonstrations have been tested on PC and Macintosh personal computers and on Sun, IBM and HP workstations. Most of the demonstrations work reliably on all platforms, however, a few minor problems remain that are described in this section. Most of these problems occur on PCs with certain combinations of processor, operating system, sound card and graphics card.

On some computers with certain versions of MATLAB, some text may not appear in the demonstration window or may overwrite other text. The user can see the correct text by hiding the window and then redisplaying it. This problem occurs in the Chapter 3 and Chapter 4 demonstrations, and in individual demos in Chapter 6 and Chapter 14 (nnd6eg and nnd14cc).

On some computers the sound commands that are used in some demonstrations can cause general protection faults. This can be avoided by turning the sound off using the command "nnsound off" before using the demonstrations.

On some computers, parts of certain figures in a few demonstrations may disappear when the user clicks on a control button or a slider. The figure will sometimes be redrawn correctly if the window is hidden and then redisplayed.

An early version of MATLAB 4.0 for Windows did not show patches of color properly, making some of the demos difficult to use. Current versions of MATLAB do not have this problem.

On some computers there is a delay as the windows are being created. After clicking on a button, give the window some time to draw before clicking again.

LIST OF DEMONSTRATIONS 
 
Many of the demonstrations are followed by one of two symbols to indicate the resources required to run them: 
 
 + Requires MININNET functions or the Neural Network Toolbox. 
  * Requires the Neural Network Toolbox. 
 

 General
   nnd      - Splash screen.
   nndtoc   - Table of contents.
   nnsound  - Turn Neural Network Design sounds on and off.

 Chapter 2, Neuron Model and Network Architectures
   nnd2n1   - One-input neuron demonstration.+
   nnd2n2   - Two-input neuron demonstration.+

 Chapter 3, An Illustrative Example
   nnd3pc   - Perceptron classification demonstration.+
   nnd3hamc - Hamming classification demonstration.+
   nnd3hopc - Hopfield classification demonstration.+

 Chapter 4, Perceptron Learning Rule
   nnd4db   - Decision boundaries demonstration.+
   nnd4pr   - Perceptron rule demonstration.+

 Chapter 5, Signal and Weight Vector Spaces
   nnd5gs   - Gram-Schmidt demonstration.
   nnd5rb   - Reciprocal basis demonstration.

 Chapter 6, Linear Transformations for Neural Networks
   nnd6lt   - Linear transformations demonstration.
   nnd6eg   - Eigenvector game.

 Chapter 7, Supervised Hebbian Learning
   nnd7sh   - Supervised Hebb demonstration.

 Chapter 8, Performance Surfaces and Optimum Points
   nnd8ts1  - Taylor series demonstration #1.
   nnd8ts2  - Taylor series demonstration #2.
   nnd8dd   - Directional derivatives demonstration.
   nnd8qf   - Quadratic function demonstration.

 Chapter 9, Performance Optimization
   nnd9sdq  - Steepest descent for quadratic function demonstration.
   nnd9mc   - Method comparison demonstration.
   nnd9nm   - Newton's method demonstration.
   nnd9sd   - Steepest descent demonstration.

 Chapter 10, Widrow-Hoff Learning
   nnd10nc  - Adaptive noise cancellation demonstration.
   nnd10eeg - Electroencephelogram noise cancellation demonstration.
   nnd10lc  - Linear pattern classification demonstration.

 Chapter 11, Backpropagation
   nnd11nf  - Network function demonstration.+
   nnd11bc  - Backpropagation calculation demonstration.*
   nnd11fa  - Function approximation demonstration.*
   nnd11gn  - Generalization demonstration.*

 Chapter 12, Variations on Backpropagation
   nnd12sd1 - Steepest descent backpropagation demonstration #1.*
   nnd12sd2 - Steepest descent backpropagation demonstration #2.*
   nnd12mo  - Momentum backpropagation demonstration.*
   nnd12vl  - Variable learning rate backpropagation demonstration.*
   nnd12ls  - Conjugate gradient line search demonstration.*
   nnd12cg  - Conjugate gradient backpropagation demonstration.*
   nnd12ms  - Maquardt step demonstration.*
   nnd12m   - Marquardt backpropagation demonstration.*
	
 Chapter 13, Associative Learning
   nnd13uh  - Unsupervised Hebb demonstration.+
   nnd13edr - Effects of decay rate demonstration.+
   nnd13hd  - Hebb with decay demonstration.+
   nnd13gis - Graphical instar demonstration.+
   nnd13is  - Instar demonstration.+
   nnd13os  - Outstar demonstration.+

 Chapter 14, Competitive Networks
   nnd14cc  - Competitive classification demonstration.+
   nnd14cl  - Competitive learning demonstration.+
   nnd14fm1 - 1-D Feature map demonstration.*
   nnd14fm2 - 2-D Feature map demonstration.*
   nnd14lv1 - LVQ1 demonstration.*
   nnd14lv2 - LVQ2 demonstration.*

 Chapter 15, Grossberg Network
   nnd15li  - Leaky integrator demonstration.
   nnd15sn  - Shunting network demonstration.
   nnd15gl1 - Grossberg layer 1 demonstration.
   nnd15gl2 - Grossberg layer 2 demonstration.
   nnd15aw  - Adaptive weights demonstration.

 Chapter 16, Adaptive Resonance Theory
   nnd16al1 - ART1 layer 1 demonstration.
   nnd16al2 - ART1 layer 2 demonstration.
   nnd16os  - Orienting subsystem demonstration.
   nnd16a1  - ART1 algorithm demonstration.

 Chapter 17, Stability
   nnd17ds  - Dynamical system demonstration.

 Chapter 18, Hopfield Network
   nnd18hn  - Hopfield network demonstration.