How to interpret this results after creating neural network ?

9 visualizaciones (últimos 30 días)
afef
afef el 18 de Jun. de 2017
Editada: Greg Heath el 18 de Jun. de 2017
hi everyone i'm trying to create neural network for classification.I begin with the simple script that i got after using nprtool but it dose'nt work so i tried to modify the code with the help of grep patternet and this is a try https://www.mathworks.com/matlabcentral/newsreader/view_thread/348806 and as i trid to correct the code as Dr Greg told me i tried this :
x = patientInputs;
t = patientTargets;
inputs = mapstd(x);
targets = mapstd(t);
N=981
I=9
O=2
[ I N ] = size(x)
[ O N ] = size(t)
net = patternnet(10);
[net,tr] = train(net,inputs,targets);
view(net)
y = net(inputs);
perf = perform(net, targets,y)
net = patternnet;
vart1 = mean(var( targets',1))
net = net
Ntrn = N-2*round(0.15*N) % 687
Ntrneq = Ntrn*O % 1374
tic, Hmin = 0, dH = 1, Hmax = 9, Ntrials = 10
rng( 0 ) % Initialize the rng to facilitate duplication
j =0
for h = Hmin : dH : Hmax
j = j + 1
if h == 0, net = patternnet( [] );
else net = patternnet( h ); end
for i = 1 : Ntrials
s( i, j ) = rng; % Recording rng states
net = configure (net ,x, t );
[ net tr y e ] = train( net, inputs, targets );
NMSEpct(i,j) = 100*mse( e ) / mean(var( targets', 1 ));
end
end
[ minNMSEpct minrow ]= min( NMSEpct )
toc
i correct the code with the help of this link https://www.mathworks.com/matlabcentral/newsreader/view_thread/346318#948388
The results that i got are :
minNMSEpct =
122.0936 122.0409 121.8789 121.4689 120.0578 120.1466 119.7203 119.5837 118.7046 120.0240
minrow =
9 7 5 3 9 8 6 1 4 1
Elapsed time is 144.726362 seconds.
But i don't know how to use this results to improve the accuracy of my neural network ? Please can anyone help me because really i have no idea ? Thanks

Respuestas (1)

Greg Heath
Greg Heath el 18 de Jun. de 2017
Editada: Greg Heath el 18 de Jun. de 2017
This is RIDICULOUS!
You used 3 consecutive NEWSGROUP posts on this problem. For the most part you just copied previous code without understanding what it was doing.
I just spent an inordinate amount of time answering the last one.
Now I come to ANSWERS and see a fourth post.
I will not answer this post unless you move it to where it belongs in your 3rd NEWSGROUP post
Go back to the NEWSGROUP and respond to my answer of your 3rd post.
You do not need multiple posts for a single problem much less use both the NEWSGROUP and ANSWERS.
Greg

Categorías

Más información sobre Sequence and Numeric Feature Data Workflows en Help Center y File Exchange.

Etiquetas

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by