MATLAB: How to choose Neural Network Training values

annneural networknewffrecognitiontraining network

I am doing hand gesture recogntion for 30 different signs. I have taken 20 images for each sign (10 for training and 10 for testing). Now I have a total of 300 images both in the training and testing set.
I have taken the following values for ANN, but the recognition is not good. Please tell me how to choose the appropriate values.
net = newff(minmax(P),[no_of_images no_of_images],{'tansig','purelin'},'traingdx','learnp');
net.LW{2,1} = net.LW{2,1}*0.01;
net.b{2} = net.b{2}*0.01;
net.performFcn = 'sse';
net.trainParam.goal =1;    
net.trainParam.show = 50;      
net.trainParam.epochs = 10416;  
net.trainParam.mc = 0.95; 
[net,tr] = train(net,P,T);

Best Answer

Always start with as many defaults as possible. Typically, you only have to MINIMIZE THE NUMBER OF HIDDEN NODES with 10 trials of initial weights for each choice for the number of nodes.
"newff" has been obsolete for years. For classification and pattern recognition it was replaced by "newpr". Years later "newpr" was replaced by "patternnet". You can use the commands "help" and "doc" for the documentation of any function.
I have posted zillions of examples in both the NEWSGROUP and ANSWERS for all of the above. Just include GREG with other search words like PATTERNNET, NEWPR and NEWFF .
Hope this helps.
Thank you for formally accepting my answer
Greg
Related Question