Yes, this is one way to get high terms without having to evaluate all
their predecessors.
Bear in mind that it's calculating 2^n twice, and the zero-th term
should surely be the
starting value. Also, it assumes that starting value is 4; just change
"4" to "x" in the
13 : ... expression to make it apply for any starting value.
So here's a lovely compact tacit version which addresses those issues,
at the expense
of returning the wrong sign for a negative starting value:
fa=: (* ([: | -) 3 %~ [: <: ]) _2&^ NB. with a [: to keep Linda
happy
So,
4 fa 200x
5892106162282964343653861005250929542581410977203573729438379
cf
4 (13 :'x p.~(([:(2&^)>:),.~ _1r3 *([:(2&^)>:) + _1&^)y' ) 199x
5892106162282964343653861005250929542581410977203573729438379
If you want a table, the rank needs to be imposed:
4 5 fl"0/ 3 4 5
29 59 117
37 75 149
A possibly easier way to understand why we need ((_2^n) - 1) % 3 is perhaps
to inspect the general term.
We have
a1 = 2a0 - 1 NB. using "Mathematical" notationrather than APL/J
a2 = 2a1 + 1 = 4a0 - 2 + 1
a3 = 2a2 - 1 = 8a0 - 4 + 2 - 1
...
an = (2^n).a0 - (2^[n-1]) + (2^[n-2]) ... + (-1)^n
The absolute sum of all terms except the first is | sum {i=0,n-1} _2^i | ,
which we know from High School maths is | [(_2^n) - 1]%[_2 -1] | ,
ie | [(_2^n) - 1]% 3 |
If you want the correct sign for negative starting values, use
(fa * * [) NB. ... which isn't quite so neat.
Cheers,
Mike
NB - previous correspondence not trimmed
On 28/11/2018 09:49, Linda Alvord wrote:
Hi Cliff,
It took a while but I got the explicit verb to work.
f=: 13 :'4 p.~(([:(2&^)>:),.~ _1r3 *([:(2&^)>:) + _1&^)y'
f i.10
7 15 29 59 117 235 469 939 1877 3755
f
4 p.~ ([: 2&^ >:) ,.~ _1r3 * ([: 2&^ >:) + _1&^
Linda
-----Original Message-----
From: Programming<[email protected]> On Behalf Of Cliff
Reiter
Sent: Tuesday, November 27, 2018 12:38 PM
To:[email protected]
Subject: Re: [Jprogramming] Recursive verbs
I like that solution. Here is another approach
4 p.~(2&^@>: ,.~ _1r3 * 2&^@>: + _1&^)i. 10
7 15 29 59 117 235 469 939 1877 3755
I found this by composing the linear poly's and noting the constant terms
satisfy a two step recursion with eigenvalues 2 and _1.
lc=.({.@[+{:@[*{.@]),*&{:
lc/\.(7$_1 2,:1 2)
_43 128
_21 64
_11 32
_5 16
_3 8
_1 4
_1 2
On 11/27/2018 11:06 AM, Raul Miller wrote:
Sure, ... that's a bit bulkier than I prefer, but it works.
Here's another alternative:
evenodd=: , (_1 2,:_1 4)&p.@{:
evenodd evenodd 4
4 7 15 29 59
Thanks,
----------------------------------------------------------------------
For information about J forums
seehttps://eur04.safelinks.protection.outlook.com/?url=http%3A%2F%2Fwww.jsoftware.com%2Fforums.htm&data=02%7C01%7C%7C677fb1d024fe4e3d937308d6548f1c93%7C84df9e7fe9f640afb435aaaaaaaaaaaa%7C1%7C0%7C636789370963003978&sdata=iWGjGzFd70epkTr2UBP7DapC8imlNBoaBgTzKw3eTE8%3D&reserved=0
----------------------------------------------------------------------
For information about J forums seehttp://www.jsoftware.com/forums.htm
---
This email has been checked for viruses by Avast antivirus software.
https://www.avast.com/antivirus
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm
