The proof refers back to identity A1, and there it says +/5⍴6 ←→ 6×5 and
+/⍳5 ←→ (6×5)÷2, and A1 is +/⍳n ←→ ((n+1)×n)÷2.  A close reading of the
text (including the figures) surrounding A1 indicates that the RHS could
easily (and more "naturally") be 5×6 and (5×6)÷2, respectively, and A1 is
more naturally +/⍳n ←→ (n×(n+1))÷2.



On Mon, Jul 27, 2015 at 7:40 AM, Roger Hui <[email protected]>
wrote:

> I had in fact marked the Lemma line and the immediately following line in
> the Errata section of the webpage.
>
> On Mon, Jul 27, 2015 at 7:36 AM, Roger Hui <[email protected]>
> wrote:
>
>> +/⍳n
>> +/⌽⍳n                      + is associative and commutative
>> ((+/⍳n)+(+/⌽⍳n))÷2         (x+x)÷2←→x
>> (+/(⍳n)+(⌽⍳n))÷2           + is associative and commutative
>> (+/((n+1)⍴n))÷2            Lemma
>> ((n+1)×n)÷2                Definition of ×
>>
>> If a typo is any deviation in transcription from the printed paper to the
>> web page, I can tell you that there was one:
>>
>> (+/(⍳n)+(⌽⍳n))÷2           + is associative and commutative  (current web
>> page)
>> (+/((⍳n)+(⌽⍳n)))÷2         + is associative and commutative  (original
>> text)
>>
>> I will correct this in the web page shortly.
>>
>> (The extra parens are unnecessary and make the expression harder to read,
>> but they were there in the original.)
>>
>> Regarding the "Lemma" line:  I agree that (⍳n)+(⌽⍳n) is more naturally
>> n⍴n+1, since both are vectors of length n and each vector element is 1+n.
>> That is, they are identical vectors.  (n+1)⍴n has the same sum but is a
>> different vector, and it is unnecessary in the proof to change the vector.
>> Nevertheless, ((n+1)⍴n) is what was in the original paper,
>>
>> (If the Lemma line is (+/n⍴n+1)÷2, that would lead to (n×(n+1))÷2 as the
>> last line.)
>>
>> p.s.  This is one of the areas in the paper where index-origin 0 would
>> have improved things.  (There are no areas I know of there 0-origin
>> would make things more complicated.)  In 0-origin the sum is equivalent
>> to 2!n.  As it is, in 1-origin, the sum is equivalent to 2!n+1.
>>
>> Another thing that would have improved things, is if # (tally) were
>> available and used.
>>
>>
>>
>>
>> On Mon, Jul 27, 2015 at 6:20 AM, June Kim (김창준) <[email protected]>
>> wrote:
>>
>>> Hello
>>>
>>> I am still translating Iverson's paper. It really takes time. One more
>>> question:
>>>
>>> <quote>
>>> +/⍳n+/⌽⍳n+ is associative and commutative((+/⍳n)+(+/⌽⍳n))÷2(x+x)÷2←→x
>>> (+/(⍳n)+(⌽⍳n))÷2+ is associative and commutative(+/((n+1)⍴n))÷2Lemma
>>> ((n+1)×n)÷2Definition of ×
>>>
>>> The fourth annotation above concerns an identity which, after observation
>>> of the pattern in the special case (⍳5)+(⌽⍳5) , might be considered
>>> obvious
>>> or might be considered worthy of formal proof in a separate lemma.
>>>
>>> http://www.jsoftware.com/papers/tot.htm
>>>
>>> </quote>
>>>
>>> In the above, I think more natural way of thinking is, the fifth line
>>> should've been instead (+/(n⍴n+1))÷2,since (⍳n)+(⌽⍳n) is identical to
>>> n⍴n+1.
>>>
>>> Do you also think this was a typo? Or is there any other thing that I am
>>> missing?
>>>
>>> June
>>> ----------------------------------------------------------------------
>>> 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