# Generalization of Binomial Coefficients to Numbers on the Nodes of Graphs

A. Khmelnitskaya, G. van der Laan, Dolf Talman

Research output: Working paperDiscussion paperOther research output

### Abstract

The triangular array of binomial coefficients, or Pascal's triangle, is formed by starting with an apex of 1. Every row of Pascal's triangle can be seen as a line-graph, to each node of which the corresponding binomial coefficient is assigned. We show that the binomial coefficient of a node is equal to the number of ways the line-graph can be constructed when starting with this node and adding subsequently neighboring nodes one by one. Using this interpretation we generalize the sequences of binomial coefficients on each row of Pascal's
triangle to so-called Pascal graph numbers assigned to the nodes of an arbitrary (connected) graph. We show that on the class of connected cycle-free graphs the Pascal graph numbers have properties that are very similar to the properties of binomial coefficients.
We also show that for a given connected cycle-free graph the Pascal graph numbers, when normalized to sum up to one, are equal to the steady state probabilities of some Markov process on the nodes. Properties of the Pascal graph numbers for arbitrary connected graphs are also discussed. Because the Pascal graph number of a node in a connected graph is defined as the number of ways the graph can be constructed by a sequence of increasing connected subgraphs starting from this node, the Pascal graph numbers can be seen as a measure of centrality in the graph.
Original language English Tilburg CentER, Center for Economic Research 26 2016-007 Published - 15 Feb 2016

### Publication series

Name CentER Discussion Paper 2016-007

### Keywords

• binomal coefficient
• Pascal's triangle
• graph
• Markov process
• centrality measure

• ## Cite this

Khmelnitskaya, A., van der Laan, G., & Talman, D. (2016). Generalization of Binomial Coefficients to Numbers on the Nodes of Graphs. (CentER Discussion Paper; Vol. 2016-007). CentER, Center for Economic Research.