However, just one tolerant comparison is not enough.
There are cases when magnitudes depend on scale, such as
quantities of length, distance, etc., where error is proportional
to magnitude which can be thought of as relative.
In other cases magnitudes are purely coincidental, such
as positions in space or time, where error is estimated
regardless of point's proximity to origin, so we can
call it absolute.
In your case of time samples, to relatively express,
say 4 seconds you would have to use an unwieldy
relative tolerance of
4000 % 1138000
0.00351494
So to treat each case separately, we will use separate
absolute and relative tolerance comparison. While
relative tolerance can be expressed in terms of absolute.
acmp=: 1 : ' x (<&| * [EMAIL PROTECTED]) -'
rcmp=: 1 : '(m * >.) acmp'
Here's how relative tolerance impacts comparison
(1134000 1142000+/i:16) (4000 acmp) 1138000
_1 _1 _1 _1 _1 _1 _1 _1 _1 _1 _1 _1 _1 _1 _1 _1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
1
(1134000 1142000+/i:16) (0.00351494 rcmp) 1138000
_1 _1 _1 _1 _1 _1 _1 _1 _1 _1 _1 _1 _1 _1 _1 _1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1
1
'<=>'{~(i.~~.@,)(1134000 1142000+/i:40) (4000 acmp)"_ 0/ 1138000 1146000
<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<=========================================
=========================================>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<
<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<=========================================
'<=>'{~(i.~~.@,)(1134000 1142000+/i:40) (0.00351494 rcmp)"_ 0/ 1138000
1146000
<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<=========================================
=======================================================>>>>>>>>>>>>>>>>>>>>>>>>>>
<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<
<<<<<<<<<<<<=====================================================================
--- Oleg Kobchenko <[EMAIL PROTECTED]> wrote:
> Sorry, wrong scale direction, with increase
> of radius is should become more tolerant (returning 0).
>
> Correct is
>
> tcmp=: 1 : '- ([EMAIL PROTECTED] * >&|) m*>.'
>
> 1134000 (0.001 tcmp) 1138000
> _1
> 1134000 (0.001 tcmp)~ 1138000
> 1
> 1134000 (0.01 tcmp) 1138000
> 0
>
> 2 2.01 2.02 2.03 (0.01 tcmp) 2
> 0 0 0 1
> 2 2.01 2.02 2.03 (0.01 tcmp)&- 2
> 0 0 0 _1
> _1 1 (0.01 tcmp) 1 _1
> _1 1
>
>
> 2 2.001 2.01 2.1 2.5 (0.0001 tcmp) 2
> 0 1 1 1 1
> 2 2.001 2.01 2.1 2.5 (0.001 tcmp) 2
> 0 0 1 1 1
> 2 2.001 2.01 2.1 2.5 (0.01 tcmp) 2
> 0 0 0 1 1
> 2 2.001 2.01 2.1 2.5 (0.1 tcmp) 2
> 0 0 0 0 1
> 2 2.001 2.01 2.1 2.5 (0.5 tcmp) 2
> 0 0 0 0 0
>
>
> --- Bj�rn Helgason <[EMAIL PROTECTED]> wrote:
>
> > Right to the point as usual!
> >
> > 2007/3/29, Oleg Kobchenko <[EMAIL PROTECTED]>:
> > >
> > > Maybe you need to convert the timestamps to
> > > integers and use a scaled tolerant comparison:
> > >
> > > tcmp=: 1 : '(m*>.) (>&| * [EMAIL PROTECTED]) -'
> > > 1134000 (0.01 tcmp) 1138000
> > > _1
> > > 1134000 (0.001 tcmp) 1138000
> > > 0
> > > 1134000 (0.01 tcmp)~ 1138000
> > > 1
> > >
> > > This is by extended definition of ''tolerance''.
> > >
> > > http://www.google.com/search?q=tolerance&sitesearch=jsoftware.com
> > >
> > >
> > >
> > > --- Bj�rn Helgason <[EMAIL PROTECTED]> wrote:
> > >
> > > > Lets say that the problem is to do a fuzzy match
> > > > Not exact match.
> > > > As you see the time difference is sometimes 18 56, 18 55, 18 57 so a
> > > good
> > > > match is something like 18 56 +/- 1 or 2 seconds
> > > > It would not be any problem if the difference is always exactly 18 56
> > > > The problem is also finding what the time difference is and in what
> > > > direction.
> > > > Time difference by some measurements outside the few second span would
> > > not
> > > > be good.
> > > >
> > > > So getting a 0/1 answer would rather be something like an interval of a
> > > few
> > > > seconds for the timedifferences in each case
> > > >
> > > > This problem is fairly trivial to do manually.
> > > >
> > > > Finding the probe by looking at the image was also trivial manually.
> > > > Seeing if the probe was there and if it was in the center or in the
> > > > periphery.
> > > > Doing it by a program seemed to me like an impossibility a month ago.
> > > > That part is the easy part now.
> > > >
> > > > Doing this fuzzy matching is probably easy as well even if it is not
> > > crystal
> > > > clear to me right now.
> > > > The differences in time are probably triggered by that the probe is
> > > > triggered by receiving exposure and it seems to vary a second or two.
> > > > Also the clock in each equipment is not in sync.
> > > >
> > > > The correct match could be within quite a span if the number is coded
> > > >
> > > > 24 60 60 1000 #.0 18 54 0
> > > > 1134000
> > > > 24 60 60 1000 #.0 18 58 0
> > > > 1138000
> > > >
> > > > 2007/3/29, Devon McCormick <[EMAIL PROTECTED]>:
> > > > >
> > > > > So, taking your guess here of the time offset of 18 minutes, 56
> > > seconds= 0
> > > > > 18 56 0, I can
> > > > > adjust c:
> > > > >
> > > > > 3{.adjc=. 24 60 60 1000 #: (24 60 60 1000 #.c)-24 60 60 1000 #. 0
> > > 18 56
> > > > > 0
> > > > > 10 57 51 450
> > > > > 10 57 59 150
> > > > > 10 58 6 950
> > > > >
> > > > > and come up with a boolean to interpolate it into a,b:
> > > > >
> > > > > (#adjc)>gv=. /: adjc,a,b
> > > > > 1 1 1 1 1 0 1 1 0 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1 0 1
> > > 1 0
> > > > > 0
> > > > > 1 0 1 1 0 0 1 0 1 0 1 0 1 1 0 0 1 0 1 0 1 1 0 1 0 0 1
> > > > >
> > > > > where the 1s mark successive elements of c and the 0s mark successive
> > > > > elements of a,b. Is this any help?
> > > > >
> > > > >
> > > > > On 3/29/07, Bj�rn Helgason <[EMAIL PROTECTED]> wrote:
> > > > > >
> > > > > > All of c come from a CT or similar equipment.
> > > > > > a and b come from a instrument that are put into the CT.
> > > > > >
> > > > > > The CT gives exposure and in c come the times - and lots of other
> > > > > > information when the exposure is made.
> > > > > >
> > > > > > a and b come from the instrument - a probe in a phantom - measure
> > > doses
> > > > > > and
> > > > > > such and also register time.
> > > > > >
> > > > > > I have been making program to find the place where the probe was on
> > > the
> > > > > > image.
> > > > > >
> > > > > > I have gathered the information from different files I get.
> > > > > >
> > > > > > I may not be getting the correct files, files may be missing.
> > > > > >
> > > > > > I know a,b and c came from the same session.
> > > > > >
> > > > > > The person did some exposures before the probe was connected so some
> > > > > > values
> > > > > > are missing.
> > > > > >
> > > > > > It is quite obvious that the clock in the probe (a and b) and in the
> > > CT
> > > > > > (c)
> > > > > > are not in sync.
> > > > > > The time difference is ca 18 minutes and 56 seconds as you can see
> > > doing
> > > > > > the
> > > > > > calculations.
> > > > > >
> > > > > > I have a gap at the beginning introducing d to fill the gap.
> > > > > > The gap could have been in the middle.
> > > > > > I have no way of knowing in advance.
> > > > > >
> > > > > > I am trying to make a reasonable guess which times fit so I can
> > > match
> > > > > the
> > > > > > exposure doses gathered from the probe (a and b) and not having to
> > > do it
> > > > > > manually.
> > > > > >
> > > > > > The time difference is not exactly 18 minutes and 56 seconds between
> > > all
> > > > > > corresponding lines in c and the lines in a and b. I do not know in
> > > > > > advance
> > > > > > if I do have a time difference and I do not know which way the
> > > > > difference
> > > > > > is
> > > > > > going to be.
> > > > > >
> > > > > > It is fair to assume all the times in c to be sorted.
> > > > > > It is also fair to assume the the individual files a and b are
> > > sorted.
> > > > > >
> > > > > > I have to find where lines in a matches lines in c
> > > > > > I have to find where lines in b matches lines in c
> > > > > >
> > > > > > The time difference in all cases should be roughly the same.
> > > > > >
> > > > > > I guess if I get a file - lets call it e - I would have to try to be
> > > > > able
> > > > > > to
> > > > > > make an educated guess that I could not find any match and it would
> > > be
> > > > > > from
> > > > > > some other test.
> > > > > >
> > > > > > Hope that explains the task at hand fairly well.
> > > > > >
> > > > > > Up to now I have done this part manually but would like to be able
> > > to
> > > > > see
> > > > > > if
> > > > > > I could get this to work.
> > > > > > I have never tried to match something like this with a program
> > > before.
> > > > > > Most of the time there is an exact match or there is not.
> > > > > > Not such a fuzzy match of matching roughly.
> > > > > >
> > > > > > Obviously I have to match the dates as well and both the formats of
> > > the
> > > > > > date
> > > > > > and the time were different.
> > > > > > I also added 000 in both a and b
> > > > > > If I am not lucky the test person could be sending in files from
> > > > > different
> > > > > > tests.
> > > > > > I am working on what requirements the test persons are supposed to
> > > have.
> > > > > > Try to exclude complete morons somehow politely.
> > > > > >
> > > > > > 2007/3/29, Devon McCormick <[EMAIL PROTECTED]>:
> > > > > > >
> > > > > > > Gosi - I don't understand what you're trying to do here.
> > > > > > > Are these timestamps that need to be matched? If so, it looks
> > > like
> > > > > > > all of "c" comes after everything in "a" and "b".
> > > > > > >
> > > > > > > gv=. /:c,a,b
> > > > > > > gv<#c
> > > > > > > 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1
> > > 1 1
> > > > > 1
> > > > > > 1
> > > > > > > 1
> > > > > > > 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
> > > > > > >
> > > > > > > Also, is there any reason to complicate the problem by having a
> > > > > separate
> > > > > > > "a"
> > > > > > > and "b"?
> > > > > > > Aren't you dealing with ab=. a,b versus c?
> > > > > > >
> > > > > > > Maybe I need either more context or a simpler version of this.
> > > > > > >
> > > > > > > On 3/29/07, Bj�rn Helgason <[EMAIL PROTECTED]> wrote:
> > > > > > > >
> > > > > > > > Does anyone have a good way of matching times?
> > > > > > > >
> > > > > > > > Here I have an example they happen to be somewhat in order.
> > > > > > > > c contains 35 lines that need to be matched with a and b.
> > > > > > > > There is an excess of 5 lines in c compared to a+b so d contains
> > > 5
> > > > > > > > dummies.
> > > > > > > >
> > > > > > > > The times need not be sorted like this and they may be missing
> > > in
> > > > > > either
> > > > > > > > direction.
> > > > > > > >
> > > > > > > > The task is to get a best match which gives the least time
> > > > > difference
> > > > > > > and
> > > > > > > > get the indexes for c that give the best match for a+b
> > > > > > > >
> > > > > > > > There are a lot of fields in each line but only connection
> > > between
> > > > > > them
> > > > > > > is
> > > > > > > > the time field.
> > > > > > > > In this example the time difference between the equipment is
> > > roughly
> > > > > > 19
> > > > > > > > minutes
> > > > > > > >
> > > > > > > >
> > > > > > > >
> > > > > > > > NB. -----------------head
> > > > > > > >
> > > > > > > > a=:".;._2 (0 : 0)
> > > > > > > > 11 00 29 000
> > > > > > > > 11 00 38 000
> > > > > > > > 11 00 44 000
> > > > > > > > 11 00 52 000
> > > > > > > > 11 01 00 000
> > > > > > > > 11 03 20 000
> > > > > > > > 11 03 28 000
> > > > > > > > 11 03 35 000
> > > > > > > > 11 03 43 000
> > > > > > > > 11 03 51 000
> > > > > > > > 11 05 09 000
> > > > > > > > 11 05 17 000
> > > > > > > > 11 05 25 000
> > > > > > > > 11 05 32 000
> > > > > > > > 11 05 40 000
> > > > > > > > )
> > > > > > > >
> > > > > > > > NB. -------------------- abdomen
> > > > > > > >
> > > > > > > > b=:".;._2 (0 : 0)
> > > > > > > > 11 11 04 000
> > > > > > > > 11 11 11 000
> > > > > > > > 11 11 19 000
> > > > > > > > 11 11 27 000
> > > > > > > > 11 11 34 000
> > > > > > > > 11 12 46 000
> > > > > > > > 11 12 54 000
> > > > > > > > 11 13 02 000
> > > > > > > > 11 13 10 000
> > > > > > > > 11 13 17 000
> > > > > > > > 11 14 36 000
> > > > > > > > 11 14 44 000
> > > > > > > > 11 14 52 000
> > > > > > > > 11 15 00 000
> > > > > > > > 11 15 07 000
> > > > > > > > )
> > > > > > > >
> > > > > > > > NB. ----------------- dcm
> > > > > > > >
> > > > > > > > c=:".;._2 (0 : 0)
> > > > > > > > 11 16 47 450
> > > > > > > > 11 16 55 150
> > > > > > > > 11 17 02 950
> > > > > > > > 11 17 10 600
> > > > > > > > 11 17 18 300
> > > > > > > > 11 19 25 350
> > > > > > > > 11 19 33 200
> > > > > > > > 11 19 41 000
> > > > > > > > 11 19 48 700
> > > > > > > > 11 19 56 450
> > > > > > > > 11 22 16 250
> > > > > > > > 11 22 24 050
> > > > > > > > 11 22 31 750
> > > > > > > > 11 22 39 500
> > > > > > > > 11 22 47 200
> > > > > > > > 11 24 05 650
> > > > > > > > 11 24 13 450
> > > > > > > > 11 24 21 150
> > > > > > > > 11 24 28 850
> > > > > > > > 11 24 36 600
> > > > > > > > 11 29 59 550
> > > > > > > > 11 30 07 350
> > > > > > > > 11 30 15 050
> > > > > > > > 11 30 22 750
> > > > > > > > 11 30 30 450
> > > > > > > > 11 31 42 600
> > > > > > > > 11 31 50 400
> > > > > > > > 11 31 58 100
> > > > > > > > 11 32 05 850
> > > > > > > > 11 32 13 550
> > > > > > > > 11 33 32 400
> > > > > > > > 11 33 40 200
> > > > > > > > 11 33 47 900
> > > > > > > > 11 33 55 600
> > > > > > > > 11 34 03 300
> > > > > > > > )
> > > > > > > >
> > > > > > > > NB. ------------------ auka
> > > > > > > > d=:".;._2 (0 : 0)
> > > > > > > > 0 0 0 0
> > > > > > > > 0 0 0 0
> > > > > > > > 0 0 0 0
> > > > > > > > 0 0 0 0
> > > > > > > > 0 0 0 0
> > > > > > > > )
> > > > > > > >
> > > > > > > > NB. ------------------- example calculations - the last one
> > > gives
> > > > > the
> > > > > > > > correct lines hooked together
> > > > > > > > c-a,b,d
> > > > > > > > 24 60 60 1000 #:((24 60 60 1000 #.c)-24 60 60 1000 #.a,b,d)
> > > > > > > >
> > > > > > > > c-~a,b,d
> > > > > > > > 24 60 60 1000 #:((24 60 60 1000 #.c)-~24 60 60 1000 #.a,b,d)
> > > > > > > >
> > > > > > > > c-d,a,b
> > > > > > > > 24 60 60 1000 #:((24 60 60 1000 #.c)-24 60 60 1000 #.d,a,b)
____________________________________________________________________________________
8:00? 8:25? 8:40? Find a flick in no time
with the Yahoo! Search movie showtime shortcut.
http://tools.search.yahoo.com/shortcuts/#news
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm
