Possibilites about obtaining mathematical equation from neural network toolbox after training?
1 view (last 30 days)
Show older comments
Hi, I was searching for a mathematical equation just like in the regression analysis where coefficient are formed for each independent valuable. Is it possible to find out an equation after performing the training parameter in neural network. As training is done by levenberg marquardt algorithum. As it is shown in the architecture of the neural network manual.
0 Comments
Accepted Answer
Greg Heath
on 4 Feb 2013
Edited: Greg Heath
on 4 Feb 2013
Standard Equations for a single hidden layer feedforward multilayer perceptron:
h = tansig( IW*x + b1 ); % -1 < h <1
Regression:
y1 = b2 + LW*h; % Equivalent to using PURELIN
y2 = tansig( y1 ); % -1 < y2 < 1
Classification:
y3 = logsig( y1 ); % 0 < y3 < 1
y4 = softmax( y1 ); % 0 < y4 < 1 , sum(y4) = ones(1,c), c classes
0 Comments
More Answers (1)
Shashank Prasanna
on 31 Jan 2013
Neural Networks are non parametric methods, which means there are no parametrized equations. The model is expressed it as a weighted sum of several sigmoids or other transfer function depending on how many layers or nodes you have.
In short there is not small equation like in parametric regression analysis.
Wolfram has a nice explanation:
0 Comments
See Also
Categories
Find more on Define Shallow Neural Network Architectures in Help Center and File Exchange
Community Treasure Hunt
Find the treasures in MATLAB Central and discover how the community can help you!
Start Hunting!