Everything is essentially a first-order Markov chain. The soft links are associations between nodes, but in a first-order Markov chain the associations are pretty incoherent.

Given a higher link density, a second-order Markov chain for Everything would be very cool.

For this, the soft links you would see would be those to nodes which people have gone to after visiting the same previous two nodes, rather than the previous one node, which you have.

The link density would have to be much higher than the current density for this to work well.

I'm going to use parentheses to group logical points, since this next sentence is impossibly convoluted but would require several K if expanded properly into multiple sentences.

You would have to, of course, modify the softlink process to (expire softlinks (that have not been visited since their creation) (after x (amount of time between visits)/(number of visits) to one end of the softlink), thus proving that the softlink was a random and unneeded one.

The purpose of the modification would be to define criteria that would eliminate softlinks that were not utilized within a reasonable amount of (visits/time) from creation. The elimination process would be totally statistical, thus following in the spirit of the E2-as-Markov-chain idea, rather than by personal opinion, which conscious and subjective process would destroy the purely statistical nature of the chain.

See, I told you it was complicated.

Log in or register to write something here or to contact authors.