Create a FeedFoward neural net with 2 input and 1 out with 1 hidden layer?

Questions
, ,
1

Dear all,
i’m looking for doing a Feedfoward neural network with 2 input antd 1 ouput with 1 hidden layer, sigmoid function and retropropagation.
how to do that ?
training and test

data =  {{1657, 0.42} -> 2545, {1466, 0.80} -> 2545, {1410, 0.91} -> 
    1195, {1731, 0.27} -> 4995, {1647, 0.44} -> 3925, {1533, 0.66} -> 
    2765, {1438, 0.85} -> 2115, {1586, 0.56} -> 3670, {1569, 0.59} -> 
    2375, {1317, 1.09} -> 1545, {1560, 0.61} -> 2765, {1490, 0.75} -> 
    2205, {1377, 0.97} -> 1515, {1333, 1.06} -> 1300, {1355, 1.02} -> 
    1570, {1619, 0.49} -> 2935, {1455, 0.82} -> 1715, {1463, 0.80} -> 
    1770, {1296, 1.14} -> 1130};
2

Have a look at the documentation of NetChain, there are many examples that look very much like what you want. I’m not sure what your plans are, but you have very few points; it is probably gonna be tricky to train your network.

What have you tried yourself? and what version of Mathematica do you use? Please read the forum rules!

3

thanks for you answer. i would like to get the weight of each input in function of the output in order to determine which input parameter can be exclude. i have done that with SNNS software but mathematica seem to be more adjustable ? i use mathematica 11. For example you have input data like temperature , pressure and so on.. and you want to know which are influancial on the output.
in fact i don’t know how to start the procedure.

4

JNO – NetExtract will let you see the weights of your network (among other things).