AIB2 Components - Neural Network

The neural network component is a complex non-linear decision logic component. It self-constructs based on the imported data vectors specified by the user. Following a successful data import, each column of data is assumed to be a unique variable in the feature vector. Each variable will be represented by a single input on the component.

Known classification of each feature vector (row) in the imported data file is expected to be in the last column. The number of outputs created and exposed by the component will represent each of the possible classifications (or decisions) the neural network must make.

Once the import is completed, the neural network is then trained to compute the correct classification or decision when presented the original feature vector (row) of data. When computation is made, each output will have a floating point value between 0.0 and 1.0. The decision is simply the output with the highest value.

Use the properties dialog box to change the outputs from detailed floating point to 1 and zeros. There will only be one high value of 1, (the winner or computed decision) and the rest of the outputs will be set to 0.

Check in the video section of the online help system for how-to videos for the neural network.

Make sure to check the box next to Neural Network Vectors in the View Tab of the ribbon menu to display the feature vectors spreadsheet.