The influence of incoming nodes to the node i could be given by this classical formula (like in neural nets):
Ii = (Σj≠i Aj x Wij) / 100
but it lets an unbalanced influence between nodes with many neighbours (influencing them) and nodes with almost none.
That's the reason why BAsCET use a little more complicated influence.
Average of influences is not a good solution either (one big influence, and many others null would give a quasi null influence). So, one uses a Divider notion, Div = ln( 3 + NbOL) / ln 3, taking into account the number of incoming links (NbOL).
This Divider is logarithmic, and gives a value of one when no links comes to the node, and a value of three when 24 links come to the node.
This behaviour was judged valuable after some experiment. Divider has been fixed in order to have stable parameters on which to found the system's behaviour interpretation. As many parameters of BAsCET, one could change its definition, using for example a sigmoid instead of the logarithm.
Then, the new incoming influence for the node i is:
Ii = (Σj≠i Aj x Wij) / ( 100 x Div )
with a maximal value of 100%.
Note that labeled links compute their influence a bit differently. They also depend on the label node's activation value.