GraphMemory
====Initial Scribblings
==
Link decay and reenforcement.
Node death and creation.
Friendships must be maintained.
Erdos numbers; death of an author.
Economics: worth more to link to older links.
Analogies: - human memory: easier to remember things you have thought about the most - friendships: best friends with the people you talk to the most (often)
====Erdos Growing Graph
==
Erdos numbers represent how many steps an author is from Erdos (in terms of authorship). Create a graph where nodes are authors and links are co-authorship. How do the Erdos numbers of each node change over time if we include node death and add the incentive that it is more prestigious to author with people with low Erdos numbers.
Each new node (grad student) is associated with one other node (advisor). A node can coauthor with its neighbors and its neighbors neigbors. (As an added complexity the probability with coauthoring with your neighbors can be higher than coauthoring with your neighbors neighbors). Coauthorship implies a new link. (As an added complexity we can have links decay and nodes die).
Questions: Can this type of graph be grown from scratch or do we need a preexisting structure to work on? In corrollary, do the processes that a graph undergoes change over time? Such that how it grows at the beginning is different than how it grows once it gains a certain amount of structure. Can this be embodied inside a single algorithm? In competing algorithms?
A way to grow this network would be to have no Erdos at the beginning. Let nodes join the network and through some process, coauthor with one another. After a certain amount of time, pick an Erdos and calculate everyone's Erdos number.
A fun addition would be the conference event where two distant nodes (who met at a conference) coauthor.
Last Edit: Mon, 4 Nov 2002 02:25:27 -0800 Revisions: 3