news

Lin-Kernighan heuristic for the TSP and minimum weight perfect matching

* src/infill.w: new file
* src/clustcalc.w: new file
* src/clusternoise.w: new file
* src/clusterdiscount.w: new file

* tspps.w: Added option to generate EPS files directly.
* Forked prolog.ps into prolog.ps and prolog.eps to simplify tspps.w's EPS work

Major changes in version 0.4.17 since 0.4:
* Milestone logging in matching code; improved in TSP code.
* Now works with DSJ_RANDOM, GEO, and ATT cost functions,.
* kdtree now computes neighbours in bulk, and is smart about not looking
  in saturated quadrants.  This is much faster now than before.
* It actually works on length_t == int, and for non-Euclidean cases.
  Before I was pretty narrowly focused on 2-d Euc and Euc-like.
* Many more scripts to generate data, and to extract answers from them.
* Improved expt.pl experiment-running script.
* Held-Karp bounding is now tuned to give pretty good answers, and it saves
  time by using sparse subgraphs most times.
* tspgen now takes a real MST as input instead of just MST edge lengths.

Major changes in version 0.4:
* First major release.

Major changes in version 0.1.1:

* Added module decluster.  It passes its own self-test.

Major changes in version 0.1:

* Use Autoconf, Automake, and conform to many GNU coding standards.