matlab中dividevec函数,求助dividevec

function [trainV,valV,testV,trainInd,valInd,testInd] = dividerand(allV,trainRatio,valRatio,testRatio)

%DIVIDERAND Divide vectors into three sets using random indices.

%

% Syntax

%

%   [trainV,valV,testV,trainInd,valInd,testInd] =

%     dividerand(allV,trainRatio,valRatio,testRatio)

%

% Description

%

%   DIVIDERAND is used to separate input and target vectors into three

%   sets: training, validation and testing.

%

%   DIVIDERAND takes the following inputs,

%     allV       - RxQ matrix of Q R-element vectors.

%     trainRatio - Ratio of vectors for training, default = 0.6.

%     valRatio   - Ratio of vectors for validation, default = 0.2.

%     testRatio  - Ratio of vectors for testing, default = 0.2.

%   and returns:

%     trainV   - Training vectors

%     valV     - Validation vectors

%     testV    - Test vectors

%     trainInd - Training indices

%     valInd   - Validation indices

%     testInd  - Test indices

%

% Examples

%

%     p = rands(3,1000);

%     t = [p(1,:).*p(2,:); p(2,:).*p(3,:)];

%     [trainP,valP,testV,trainInd,valInd,testInd] = dividerand(p,0.6,0.2,0.2);

%     [trainT,valT,testT] = divideind(t,trainInd,valInd,testInd);

%

%  Network Use

%

%   Here are the network properties that defines which data division function

%   to use, and what its parameters are, when TRAIN is called.

%

%     net.divideFcn

%     net.divideParam

%

% See also divideblock, divideind, divideint.

% Copyright 2006-2008 The MathWorks, Inc.

%% ERROR CHECKING

if nargin < 1, error('NNET:Arguments','Not enough arguments.'),end

%% FUNCTION INFO

if ischar(allV)

switch (allV)

case 'info'

info.name = mfilename;

info.title = 'Random';

info.type = 'Data Division';

info.version = 6;

trainV = info;

case 'name'

trainV = 'Random';

case 'fpdefaults'

defaults = struct;

defaults.trainRatio = 0.6;

defaults.valRatio = 0.2;

defaults.testRatio = 0.2;

trainV = defaults;

otherwise

error('NNET:Arguments','Unrecognized string: %s',allV)

end

return

end

%% DEFAULTS

if (nargin == 1), trainRatio = dividerand('fpdefaults'); end

if isstruct(trainRatio)

valRatio = trainRatio.valRatio;

testRatio = trainRatio.testRatio;

trainRatio = trainRatio.trainRatio;

else

if nargin < 3, testRatio = trainRatio * 0.2/0.6; end

if nargin < 4, valRatio = (trainRatio+testRatio) * 0.2/0.8; end

end

%% DIVIDE DATA

totalRatio = trainRatio + testRatio + valRatio;

testPercent = testRatio/totalRatio;

valPercent = valRatio/totalRatio;

[allV,mode] = nnpackdata(allV);

Q = size(allV{1,1},2);

numValidate = round(valPercent * Q);

numTest = round(testPercent * Q);

numTrain = Q - numValidate - numTest;

allInd = randperm(Q);

trainInd = sort(allInd(1:numTrain));

valInd = sort(allInd(numTrain+(1:numValidate)));

testInd = sort(allInd(numTrain+numValidate+(1:numTest)));

[trainV,valV,testV] = divideind(allV,trainInd,valInd,testInd);

trainV = nnunpackdata(trainV,mode);

valV = nnunpackdata(valV,mode);

testV = nnunpackdata(testV,mode);