Transparallel
pencil selection
Usually,
items are processed serially
(i.e. one after the
other by one processor) or in parallel
(i.e.,
simultaneously by many
processors). Transparallel
processing means that items are processed simultaneously by
one processor (i.e., as
if only one item were concerned). This may look like
science
fiction but the next example shows it is not.
Suppose that, for some odd reason, you want to select the longest
pencil from a number of pencils:
One way to do this is that you measure the lengths of all
pencils serially (i.e., one pencil after the other):
You could of course also call a number of friends to do the
measurements in parallel, that is, the friends work simultaneously and
each friend measures only one pencil or only a few pencils. This would
reduce the amount of time needed to
find
the longest pencil, but it would
not reduce the total amount
of work.
If you are smart, however, you follow a much
simpler
procedure, which reduces both time and work needed to
find
the longest pencil. You gather all pencils in one bundle, you put the
bundle upright on
the
table,
and you select the longest pencil in one glance:
This is not a procedure by which the pencils are measured one after
the other by one processor or simultaneously by many processors. It is
a
procedure by
which the pencils are processed simultaneously by one processor. This
is
what I call transparallel processing which, in this example, is
possible by collapsing the separate pencils into one bundle.
For a further positioning with respect to other forms of processing, see
Smart processing
For a formal account of transparallel processing, see
Proceedings
of the National
Academy of Sciences USA 2004
For its potential relevance in cognitive (neuro)science, see
Cognitive architecture and
Cognitive Processing 2012