Fig. 4
This figure illustrates the flow of information between adjacent nodes. Each node belongs to a postivie and a negative plane. The positive information nodes are yellow and the negative information nodes are light blue. A positive edge weight is represented by \(e^{+}_{ij}\), while a negative edge weight is represented by \(e^{-}_{kj}\). The sign of the edge weights determines which information (positive or negative) flows from one node to another. The positive information of a node xj depends on the positive information of xi when the edge is positively weighted (think: positive times positive is positive). The negative information of a node xj depends on the negative information of xi when the edge is positively weighted (think: negative times positive is negative). Similarly, the negative flow of negative information can contribute to positive information. The “dual-channel” RWR algorithm incorporates these edge weights. Note that in practice each node will contain both positive and negative information