From Trent Adams at matchmine
"While playing around with similarity metrics, I’ve been struck by how many different approaches there are for comparing one thing to another. For example, I’ve got a movie and want to find out how similar it is to another movie. There are probably a zillion different ways to code up a solution culled from between the covers of a standard computer science textbook. A matrix can then be cooked up for a bake-off between them.
What I find particularly interesting, though, is when you deviate from the CompSci path and try walking toward the solution in the shoes of a different discipline. Looking through the lens of my astronomer’s telescope, for example, I see a similarity dominated by Euclidean geometry and something similar to inverse-squared law of Newton’s gravitational effects. Digging into the more advanced tools of the astrophysicist, we can glean similarities based on more esoteric concepts like string theory to get from one side of n-space to another (not to mention quantum entanglement).
Jumping a bit further afield, we’ve been dabbling with some dynamic programming techniques (don’t get me started on that term that’s rife with misinterpretation). With a nod to our genetic engineering pals who like to “Think Globally, and Align Locally,” it appears to be a promising approach to identifying an optimal comparative alignment.
The question, with all of these interesting inputs, is which world view dominates the discussion? When driving toward a common, workable solution, it’s unlikely we’ll magically stumble upon the grand unified theory of the mediaverse. Only time will tell who wins the debate (if a winner is ever crowned), so for now we’re playing the field(s)."
Comments