Loading ...
Sorry, an error occurred while loading the content.

Re: A Query abt 'Association Strength'

Expand Messages
  • Stephen Wolstenholme
    ... Strength is not based on a scale. All the columns are set to be Inputs. Each column in turn is changed to Output and a network is created. The networks are
    Message 1 of 1 , Mar 5, 2008
    • 0 Attachment
      On Wed, 5 Mar 2008 11:59:07 +0530, Rajeev Upadhye wrote:

      >
      >Hello
      >
      >I have not understood the concept of association strength in EasyNN. Here
      >are specific questions -
      >o The strength is indicated in numbers. Is it based on a scale? If yes then
      >what is the range of this scale?

      Strength is not based on a scale. All the columns are set to be
      Inputs. Each column in turn is changed to Output and a network is
      created. The networks are trained for 2000 cycles. The Output is
      paired with the most important Input. The strength of the association
      is the importance of the most important input which is the sum of the
      absolute weights of the connections from the input node to all the
      nodes in the first hidden layer.

      >o I did this simple exercise of associating two columns containg same data,
      >i.e. one column is an exact image of the other. I was expecting a strong
      >association as there is one-to-one mapping. But I got answer 3.5 etc. I dont
      >know what it means...
      >

      3.5 is the input importance after training for 2000 cycles.

      >Can anyone clarify?
      >
      >
      Steve

      --
      Neural Planner Software Ltd
      EasyNN-plus. Build Neural Networks. http://www.easynn.com
      SwingNN. Forecast the Future. http://www.swingnn.com
    Your message has been successfully submitted and would be delivered to recipients shortly.