Tuesday, September 14, 2010

Knots as processes now on arXiv

A draft of our knots-as-processes paper is now available on arXiv. Here's the translation to English: to each knot we assign (calculate, really) a certain kind of computer program -- a program made of concurrently executing smaller programs. These programs are "invariant" -- meaning that transformations of the knot that leave the knot essentially unchanged (like untwisting a loop caused by a twist) correspond to transformations of the program (assigned to the knot) that leave the program essentially unchanged.

Typically, knot invariants have been algebraic structures (like polynomials -- but also more exotic structures -- like quandles) that are "static". They don't squirm and wiggle on their own the way a computer program might. Our invariants are a new class of invariant because they incorporate this dynamical aspect -- they squirm and wiggle. Moreover, their squirming and wiggling correspond to knot structure in a specific and detailed way.

Further, our invariants come with a proof method, known in the literature as bisimulation. The method has been known in science for donkey's years -- two things must be considered the same if you can't find an experiment that distinguishes them. But, it took several hundred years to find a formulation of this method that evinces some of it's real power and flexibility. We illustrate with this application that the proof method of bisimulation is just as applicable to topology as it is to concurrency. This is perhaps the most important contribution of the paper.

Further, the programs we use come with their own special family of logics. By "factoring through" our translation we arrive at a family of "knot logics". This allows us to provide a language for picking out classes of knots. This is a unique capability, to the best of our understanding -- though Conway's knotation offers more than a little hint of this capability. One of the most exciting potentials in this is the class of applications to reasoning about protein folding.

Finally, we show that the invariant is exceptionally fine-grained in the sense that two knots are "equal" if and only if their invariants are "equal". That's about as high a quality detection as is possible. To compare it with other invariants: in the case of the Jones polynomial, for which Vaughn Jones got the Fields medal, it was unknown whether it detected knottedness for decades. This invariant comes with a proof method that decides this question at the outset.

For those interested in where such ideas come from, in this case the basic idea flew by my mind's eye in 1998 while i was reflecting on the knotwork on the Discipline album. It took me about 3 years from the basic insight to have the skills and direct understanding to write down the idea. It took me about 3 more years to find a collaborator, in the person of David Snyder, patient enough and diligent enough to work with me. His probing questions helped us improve the results -- especially the work on the R3 move. It took us almost 6 years (with lots of stuff going on in between) to pull this draft together.

No comments: