This TPAMI paper has the same content as our 2013 tech report of the same
name. We released the tech report the same day that we submitted this paper to
TPAMI, and the paper spent nearly two years in review. Not because it was
rejected --- we just didn't hear anything from the publisher for months, and
only after repeated prodding did we finally get an "accept with no changes"
review. So why did this paper spend nearly two years in review? I wish I could
say, kind reader; one cannot hope to fully understand the mysterious ways of
"Ye Olde Transactions Regarding Patterns and the Wits of Machines". Maybe the
letterpress operators were stricken with dysentery, or perhaps the carrier
pigeon relaying the reviews were gunned down by one of the Kaiser's
sharpshooters. This is one of the great mysteries of life that we must accept,
like "why do computer scientists still publish journal papers instead of just
posting things to Arxiv", or "why would talented researchers freely volunteer
their time to review papers for journals that put articles behind paywalls and
charge open-access fees". It's best to just accept the simple fact that all
good papers must be kept secret for at least eighteen months, so that all of
the theorems and statistics contained therein have time to ferment and mature.