Neural

I want 2 neural systems. 1 neural system for 10 categories kai 1 neural system for 3 categories.

The 10 categories are the following:

A	calm sleep
B	REM sleep
C	calm vigilance
D	active vigilance
SH	shift pattern (A or Susp with shifts)
AD	accelerative/decelerative pattern (stress situation)
DE	decelerative pattern (vagal stimulation)
LD	largely decelerative pattern
FS	flat-sinusoidal pattern (pathological state)
SUSP	suspect pattern

The 3 categories are the following:

N=normal;

S=suspect;

P=pathologic

What should be done:

Neural should be done in Matlab2014b.

We will use the 10-Fold Cross Validation.

The data we will use is 2126. And we will separate in 10 smaller pieces. 4 pieces- 212 and 6 pieces-213.

We will use the patternnet(hiddenSizes,trainFcn,performFcn), Indices=crossvalind(‘kfold’,N,K), train=~test; , data=data’;.

Data=attributes*examples; (attributes=fitures)

Using Indices should do 10 test.

testx=data(:,test); They will put in a For Loop that will count the error.

train=data(:,train);

To experiment with different hidden units(5-10-15-…-50) and record the result.

To experiment with different algorithms. trainscg is the default. Do try the trainbr and trainrp and record the result and compare them.

After you declare the neuronal, after paternnet declare:

net.divideParam.trainRatio = 0.8; net.divideParam.valRatio = 0.2; net.divideParam.testRatio = 0;