The temperature reduction factor #tex2html_wrap_inline1846# and the the number of moves at
each temperature #tex2html_wrap_inline1848# are related, since we expect that if #tex2html_wrap_inline1850# is
replaced with its square root (so the number of temperatures in the
cooling schedule is doubled), and #tex2html_wrap_inline1852# is halved, the solution quality is
expected to be about the same, as long as #tex2html_wrap_inline1854# is large enough.
Usually #tex2html_wrap_inline1856# is chosen to be in the region #tex2html_wrap_inline1858#.
We found only a small variation in the results for this parameter range,
and decided to arbitrarily fix #tex2html_wrap_inline1860# to be 0.95, and use #tex2html_wrap_inline1864# as
a variable with which to compare solution quality.
In some of the runs we set this variable to be at least proportional to
the neighborhood size, or the size of our data, and in general we observed
that this yielded slightly better results.