> -----Oorspronkelijk bericht-----
> Van: [email protected] [mailto:programming-
> [email protected]] Namens Raul Miller
> Verzonden: dinsdag 18 augustus 2009 0:37
> Aan: Programming forum
> Onderwerp: Re: [Jprogramming] minmax differences
>
(...)
>
> (R.E. Boss's approach also seems to be relevant, but I
> need to study his some more.)
>
Let me try to help you.
mmd=: 4 : 0
z0=. (i.x) * y <....@%&<: x
z1=. <:y - {:z0
z2=. (z1 combnu >:x) (+/@:{)"1 _ |.|:(>:/&i.>:)x
(<,z0) +"1&.> (([:>./ 2 -~/\ ])"1 </. ])z2
)
Looking for a maximum minimum distance between the numbers of an item of x
comb y , it is obvious that this is achieved in z0 , since that
max-min-distance is equal to y <....@%&<: x .
All the other items with this max-min-distance differ at most z1 , which
is essentially x |&<: y . All these differences are z2 .
Now z2 is boxed according to its maximum distance and than z0 is added
to it.
Each box contains the items with the required minimum distance and the
maximum distance 0,1,2,.... more.
This is shown by
([:~. ([:(<./,>./) 2 -~/\ ])"1)&.> 11 mmd 55
+---+---+---+---+---+
|5 5|5 6|5 7|5 8|5 9|
+---+---+---+---+---+
The advantage of the method is that x comb y is not needed.
Compare
11 ! 55
119653565850
or even worse
50 ! 102
3.9192407e29
with
$&.> 50 mmd 102
+----+--------+-------+-----+
|4 50|20923 50|2450 50|49 50|
+----+--------+-------+-----+
ts' 50 mmd 102'
0.11643995 43856640
R.E. Boss
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm
