readme.txt

用Cross validation的方法建立人工神经网络的模型！

Last modified: May 10, 2005
Modified by: Jingjing Zhang

This zipped folder contains all Matlab source code for Keyblock and PCA experiments.
All copyright @ J. Zhang

To run the experiments, open Matlab 7.0.4, type "main_VQ_PCA" in the command window. 
All experiments results will display at frontend.

There are totally 25 code files in this project and the descprition is as follows:

% avg_distortion.m: compute the average distortion overall partitions based on newly generated cdbk
% calculate_prediction_accuracy.m: calculate the prediction accuracy using neural network on given data and feature subset
% centroid.m: find the central point for a given region
% ClassifierANN.m: ANN neural network classifier, 5-fold cross validation
% divideset.m: divide 'data' to 'train' and 'test' set, how many percents goes to training set is determined by treshold
% eigdec.m: computes the largest N eigenvalues and corresponding eigenvectors of the matrix X in descending order
% get_features.m: partion the 2D slice into certain number of partions, each of given blocksize
% GLAcdbk.m: generate code book with certain blocksize and code book size
% image_encode.m: encode each image in the database using given codebook
% distance.m: returns the distance between two vector x and y
% normalize.m: normalize data using zero-mean normalization
% partition.m: partion the points into 'nwds' partions based on given cdbk
% PCA.m: principle components analysis, projects given data onto N dimensional subspace
% main_VQ.m: Keyblock fMRI data compression
% main_PCA.m: PCA fMRI data compression


For details of each function, please refer to specific comments in the code.