The direction of a connection shows the direction of the transfer of activation. The unit from which the connection starts is called the source unit, or source for short, while the other is called the target unit, or target. Connections where source and target are identical (recursive connections) are possible. Multiple connections between one unit and the same input port of another unit are redundant, and therefore prohibited. This is checked by SNNS.
Each connection has a weight (or strength) assigned to it. The effect of the output of one unit on the successor unit is defined by this value: if it is negative, then the connection is inhibitory, i.e. decreasing the activity of the target unit; if it is positive, it has an excitatory, i.e. activity enhancing, effect.
The most frequently used network architecture is built hierarchically
bottom-up. The input into a unit comes only from the units of
preceding layers. Because of the unidirectional flow of
information within the net they are also called feed-forward nets
(as example see the neural net classifier introduced in chapter
). In many models a full connectivity between all
units of adjoining levels is assumed.
Weights are represented as floats with nine decimal digits of precision.