I have 6 parameters each a row matrix of dimension 1 x 5, now i want to consider these as inputs and define 2 outputs. For each parameter i want to set a cut-off and give the appropriate rules to get classified into either of the two outputs. After feeding in the neural network and training it, how to given a new unknown input (i.e. all 6 parameters) and it should be classified into either of the one category. How to design and do this ? My problem case is somewhat similar to the breast-cancer example in the matlab neural network toolbox examples.
hi,
You can see an example in this website:
http://www.mathworks.com/help/nnet/examples/training-a-deep-neural-network-for-digit-classification.html
I currently work in breast cancer images, Maybe I can help you more.
Dear Vyza,
I believe in your case that there is no need to use ANN you have an input of 6 variables with each has a dimension of 1X5 and you want to classify them into 2 classes based on a cut-off you mean a threshold. In this case there is no need for ANN you can solve it using any easier method.
Best regards
Dear Vyza,
you should provide us with further details on the input parameters (type, dimension ecc.).
However, generally speaking, when you face a classification problem by means of an ANN, trained, for example, by a back-propagation procedure, you need to:
Define the architecture of your ANN. In your case, you could try a 6 input, 10 hidden nodes (it depends on your dataset) and 1 output network.
Next, you set the output neuron activation function to logistic sigmoid (for classification tasks).
You start the training (actually I'm ignoring other network's parameters) by providing the ANN with the couples (input,targets).
When the training phase is finished, you test the generalization performance of your network giving it the input data only (with no targets). You'll observe the output (a value
Try using CGPANN. for further details you can see my publications in my Research gate profile.
Regards.
Arbab Masood Ahmad
Mr. M. Awad Just for a simplified case i have asked a question citing fewer inputs. But in reality i have many number of samples roughly about 4,50,000 samples and for each of these samples i found out some of the parameters. Basing on this i would like to classify them. Also i have successfully implemented this situation by using fuzzy logic. But i'm unable to correspond the same using ANN.
Dear Antonino Staiano
Thanks for the reply. I have actually gathered EEG samples for epileptic subjects as well as normal subjects. I have correspondingly found some statistical parameters based on it like Approximation Entropy, Energy, Std. Deviation, variance. Basing on all these parameters, i want to feed them as inputs to the neural network train all the samples and i want the end result to be classified if it is epileptic or non-epileptic. Also to find the corresponding acuracies at different level's of decomposition as i have used Discrete wavelet transform.
Anyhow i shall try out the method you have mentioned.
Regards
- Badges
- Science topic