How to know the output in Neural Network Pattern Recognition?

hello guys! i'm a beginner in using matlab neural network pattern recognition and I can't understand the output part. How could I show the intended output after training? For example in the load example data about the type of glass, how can I say if the sample is a window glass or not? I want to show something like this after the data has been recognize if it is indeed a window glass, "The sample is a window glass". What should I do?

I hope someone can answer me or atleast give some idea. Thanks.

[Sivakumaran Chandrasekaran]

after simulating the neural network, you might be getting values like 2 or 5.. consider those values as a threshold values and write the code like if the value for the input image is greater than the threshold, the image is window glass.. else..different

[Aeri]

which values? the performance? there's a lot of values after training so I don't know which one to use. But I kinda get your point. I just don't know which one to use.

[Sivakumaran Chandrasekaran]

what value you got when you use the command sim... use that value as threshold

[Aeri]

[Y,Pf,Af,E,perf] = sim(net,P,Pi,Ai,T) - is this the command sim? Because after I put that, my Y, which is the network output gives a matrix output. Or do I still have to compute for the threshold value?

Sorry for all the questions, still in my beginning stage.

Answer 1

[Greg Heath]

For classification/pattern-recognition into c classes, the columns of the target matrix should be columns of the {0 1 } unit matrix eye(c).

The transformation between classindices and target matrix are obtained via the functions ind2vec and vec2ind.

 trueindices = [ 5 3 1 2 4 ]
target      = full(ind2vec(trueindices))
output      = target + 0.15*randn(5)
predindices  = vec2ind(output)

Hope this helps.

Thank you for formally accepting my answer

Greg

[Aeri]

is trueindices the same with input?

[Aeri]

I think I got it with the use of vec2ind. I think I'm just going to compare the input and the output to get a result.

Thanks for the answers guys!