none
Neural Network Hidden layer

    Question

  • Hi

    How to calculate the Hidden Node 1, Hidden Node 2 ........

     bogdancrivat . net /dm/archives/36

    Input Node                Hidden Node 1   Hidden Node 2
    60000000000000000 0,271975739    0,633542494
    60000000000000001 0,461254404   -0,669184916
    60000000000000002      -0,991132321    0,038144284
    60000000000000003 -0,093255775    0,756049303
    Intercept 0,112857411  0,094328628

    Thanks

    Thursday, July 11, 2013 2:06 PM

Answers

  • You might like to review Querying a Neural Network Model, and specifically the Sample Query 4: Retrieving Weights from the Hidden Layer:

    SELECT FLATTENED TOP 1 NODE_UNIQUE_NAME, 
    (SELECT ATTRIBUTE_NAME, ATTRIBUTE_VALUE, VALUETYPE
    FROM NODE_DISTRIBUTION) as t
    FROM  [Call Center Default NN -- Predict Service and Orders].CONTENT
    WHERE NODE_TYPE = 22
    AND [PARENT_UNIQUE_NAME] = '40000000200000000' FROM [Call Center Default NN].CONTENT

    The weights are multiplied by values from the previous layer and the result typically goes through a sigmoid function that determines the value for the hidden node.  That's how a set of input values get turned into output values, after going through hidden layers.  The weights are determined by training the neural network on a set of input data.  The individual hidden layer weights and values will usually be less informative than the behavior of the net as a whole.

    Thursday, July 11, 2013 2:43 PM