Couldn't sleep. Turns out it soesn't take too long to brute force it by
pencil and paper.
Given z is 28..1, y is 27..1 and x is 26..1, you start with z at 28 and y
at 27 and x at 2. It took me 15 minutes to enumerate the list of 61
solutions. Certainly not near J's speed, but I would have thought longer
than 15 minutes.

Good night.

Rob


On Mon, May 27, 2013 at 2:54 AM, Robert Herman <[email protected]> wrote:

> My reference to Perl's output was of the solutions printed on the count
> lines such as this:
>
> 1: 2 27 28
> 2: 3 26 28
> 3: 4 25 28
> .....
>
> 60: 17 19 21
> 61: 18 19 20
> Total 61
>
> Raul, thank you so much for your other pointers on inspecting the arrays.
> I will have two more days after work going over all of that. It's amazing
> how working through this seemingly simple problem with your help has taught
> me more about J than the hours I spent just doing single bits from the J
> dictionary. What would be the best J book to go through the vocabulary in
> problem sets such as this? Rosetta code? J for C programmers?
>
> My son said that one of the students arrived at the correct answer of 61
> without a computer. I can't believe this given the time it would require to
> go through the answer space via pencil and paper (28^3). Is there a closed
> form solution, or algebraic solution?
>
> Good night, it's almost 3 am here in Macau.
>
> Thanks,
>
> Rob
>
>
>
>
> On Mon, May 27, 2013 at 2:19 AM, Raul Miller <[email protected]>wrote:
>
>> On Sun, May 26, 2013 at 2:01 PM, Robert Herman <[email protected]>
>> wrote:
>> > I have worked through the solution arrived at by a few of you to better
>> > learn J, namely this one:
>> >
>> > +/,(57=x+/y+/z)*x</y*y</z
>> >
>> > In order to fully understand the way the matrix operations are carried
>> out,
>> > I minimized the answer space by making P = 13, and worked through the
>> > output (not shown for brevity's sake) of the following input lines I
>> placed
>> > in a project file (learning the IDE with this problem too):
>> >
>> > P =: 13
>> > x =: >:i.4
>> > y =: >:x
>> > z =: >:y
>> >
>> > 3 4 $ x,y,z
>> >
>> > (13=x+/y+/z)*x</y*y</z
>> > +/,(13=x+/y+/z)*x</y*y</z
>> >
>> > +/,((13=[+/+/)*[</]*</)~i.7
>>
>> Since this is a rank 3 bit array, we can ask J to display it more
>> compactly.
>>
>>    <"2 '.*'{~((13=[+/+/)*[</]*</)~i.7
>>
>> We can also use this prefix to inspect some of the intermediate results:
>>    <"2 '.*'{~([</]*</)~i.7
>>    <"2 '.*'{~(</)~i.7
>>
>> That said, some interemdiate results are not bit arrays and are
>> perhaps best inspected directly:
>>    (]*</)~i.7
>>
>>
>> > Anyway, I was tempted to try and make the J program look like the Perl
>> > oneby using
>> > for. but I realized I was lapsing into old habits. My minimal
>> programming
>> > experience is all with imperative languages or some Lisp too. My next
>> > question is how do you get the J program to print the solutions as in
>> the
>> > Perl program?
>>
>> That's an open-ended question, since "as in" has a contextual and case
>> specific meaning.
>>
>> Personally, I usually try and understand intermediate results (usually
>> there is only one and I can just extract an expression which computes
>> it and see what it does with sample data.
>>
>> > In my 3 x 4 matrices of the reduced problem set, I could see
>> > how the 1's in the solution selection matrix set related to the
>> original 3
>> > x 4 matrix of x, y and z rows. Without re-writing it, is there a way to
>> > index the solution?
>>
>> I am not sure what you mean by this.
>>
>> I mean, of course you can index the solution, but the solution has
>> only a single number.  You can also rewrite the expression which
>> creates that solution and index from those results. But I feel like I
>> am not understanding your question.
>>
>> > One question that is on my mind is the tacit vs. explicit style. I am in
>> > love with the way you all distill the solution down to a tacit
>> one-liner,
>> > just as a maths formula is simplified, however, I am starting to think
>> that
>> > if and when I revisit this problem in the future, it will take a bit of
>> > jarring to read it, translate it if you will. I am guessing for. loops
>> are
>> > very expensive cycle-wise and it would be just as easy to make words for
>> > phrases if I wish to preserve future readability. I am getting addicted
>> to
>> > the fast way of working in the J environment, and I am grateful to all
>> on this
>> > list for sharing their wisdom.
>>
>> I think that any time we delve into some area of computation that
>> leaving and coming back after an extended absence can leave us in a
>> jarring situation where we need to reacquaint ourselves with the
>> original topic. I do not know of any general cure for this, finding
>> good names for things, and good examples, can be fruitful.
>>
>> Thanks,
>>
>> --
>> Raul
>> ----------------------------------------------------------------------
>> For information about J forums see http://www.jsoftware.com/forums.htm
>>
>
>
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm

Reply via email to