Data mining in Mathematics: A human example

Computers beat humans at Computation, Enumeration and even (though most people don’t realise this) Deduction. However, a central feature of most mathematics is relating to the literature, which in practise is gathering together results (and concepts) to be used for deduction. However, computers at present do not search the (human)  literature. In the hope of eventually teaching them, here are the steps by which I proved a lemma.

The  lemma is a statement of the following nature.

Lemma: For words a, b, A and B in the fundamental group of a surface satisfying some conditions, if ab=AB then a=A and b=B.

Here are the steps in my discovery of a proof:

  1. The starting point was a result of Chas-Krongold, which was a special case in a more restricted context, plus knowing from Moira Chas that she had proved (but not published) a result similar to my lemma but in a more restricted context.
  2. The Chas-Krongold work used what is called small cancellation theory. It is well known that a method that proves the same results in more generality (at the cost of weaker estimates) is $\latex\delta$-hyperbolicity, which is applicable in our context.
  3. To use word-hyperbolicity, I tried to construct quasi-geodesics associated to ab and AB. This is what we can try to associate to elements.
  4. The construction of the quasi-geodesics was the most obvious one. The issue was to prove it was a quasi-geodesic. Here I used a non-trivial theorem – local quasi-geodesics are quasi-geodesics.
  5. To verify that what we obtained was a local quasi-geodesic, one used some geometry and another lemma – a lower bound on angles.
  6. To get the lower bound on angles involved using another known trick from the literature – consider the commutator and show that it is small.
  7. Now that we have quasi-geodesics, we use one of the main theorems  concerning $\latex\delta$-hyperbolicity, that quasi-geodesics are close to geodesics.
  8. Now we use some geometry parallel to the combinatorics of Chas-Krongold and the trick of considering commutators to finish the proof.

The deductions are not difficult at any point. The main obstacle to automation (say of a computer as a collaborator) is to be able to make the right analogies and dig up the literature. Indeed the analogy is also not that obscure – early papers on $\latex\delta$-hyperbolicity do show that the results include those of small-cancellation theory.

The main barrier is understanding natural language. In the sequel I speculate on a scheme to try this.


About sgadgil

I am on the faculty of the Department of Mathematics, Indian Institute of Science.
This entry was posted in Uncategorized. Bookmark the permalink.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s