DMX Neural Network

• Friday, February 01, 2013 11:10 AM

   

   
  

  0

  I am Developing a Neural Network Model to predict the placement of certain objects on a diagram , by predicting the X-Coordinate and Y-Coordinate of these objects.

  Some predictions for the (X,Y) topple are being returned same for different objects.

  Is  there a way to make sure list of predicted results returned are unique ?

  by saying something like previous predictions are no longer possible results for a prediction ?

  or if there's a way to return the next most probable prediction ?

  If not maybe i can return all possible predictions for an object then by code, or SQL script return a  unique result for each object.

  Which i dont know how to do in DMX.

  Here is the DMX i am currently using:

  SELECT
  t.[FileName],
  t.[SheetName],
  t.[ID],
  t.[Pointer2DX],
  t.[Pointer2DY],
  t.[Pointer3DX],
    t.[Pointer3DY],
    t.[Pointer3DZ],
    t.[Type],
    Predict([Symbol].[Symbol2DX]) AS PredictedSymbol2DX,
    predictprobability([Symbol].[Symbol2DX]) AS ProbSymbol2DX,
    Predict([Symbol].[Symbol2DY]) AS PredictedSymbol2DY,
    predictprobability([Symbol].[Symbol2DY]) AS ProbSymbol2DY
  From [Symbol]
  PREDICTION JOIN
  OpenQuery
  ([7lpt-aghali],
  'SELECT 
    [FileName]
        ,[SheetName]
        ,[Type]
        ,[ID]
        ,[Pointer2DX]
        ,[Pointer2DY]
        ,[Pointer3DX]
        ,[Pointer3DY]
        ,[Pointer3DZ]
        ,[Symbol2DX]
        ,[Symbol2DY]
        ,[SymbolWidth]
        ,[SymbolHeight]
    From [Symbol]
    order by [FileName],[SheetName], [ID]'
    ) AS t
    ON
    [Symbol].[ID]  = t.[ID]  AND
    [Symbol].[Pointer2DX] = t.[Pointer2DX] AND
    [Symbol].[Pointer2DY] = t.[Pointer2DY] AND
    [Symbol].[Pointer3DX] = t.[Pointer3DX] AND
    [Symbol].[Pointer3DY] = t.[Pointer3DY] AND
    [Symbol].[Pointer3DZ] = t.[Pointer3DZ] AND
    [Symbol].[Type] = t.[Type]
   

  Any help is appreciated thnx

  • Reply
  • Quote