Re: New integer multiplication algorithm
From: Proginoskes (proginoskes_at_email.msn.com)
Date: 02/17/05
- Next message: Jason: "Re: John Gabriel's Theorem Revisited."
- Previous message: mensanator_at_aol.compost: "Re: Collatz conjecture"
- In reply to: fiziwig: "Re: New integer multiplication algorithm"
- Next in thread: Oscar Lanzi III: "Re: New integer multiplication algorithm"
- Reply: Oscar Lanzi III: "Re: New integer multiplication algorithm"
- Messages sorted by: [ date ] [ thread ]
Date: 17 Feb 2005 13:16:00 -0800
fiziwig wrote:
>
> What about even numbers? [...]
For the ultimate application -- factoring -- we won't have to worry
about even numbers. It's easy to tell whether a number is even, no
matter what base it's in. (Either look at the last digit, if the base
is even; or add up the digits, if the base is odd.)
> On another tangent, what about this:
>
> (x,y) = (9,13)
>
> MIN(9,13) = 9 = 1001
> ABS(x-y)/2 = 2 = 0010
>
> 10, 00, 01, 10 -> {2012}
>
> {abcd} # {stuv}
> = 8{a}#(stuv) + 4{b}#(stuv) + 2{c}#(stuv) + {d}#(stuv)
> = 64{a}#{s} + 32{a}#{t} + 16{a}#{u} + 8{a}#{v} +
> 32{b}#{s} + 16{b}#{t} + 8{b}#{u} + 4{b}#{v} +
> 16{c}#{s} + 8{c}#{t} + 4{c}#{u} + 2{c}#{v} +
> 8{d}#{s} + 4{d}#{t} + 2{d}#{u} + {d}#{v}
>
> for {2012} # {2012}
>
> a = s = 2
> b = t = 0
> c = u = 1
> d = v = 2
>
> {2012} # {2012} = 64 + 16 + 8 + 16 + 2 + 8 + 2 + 1 = 117
>
> and, indeed 9 * 13 = 117
>
> However, suppose we express 117 as (1,117)
>
> MIN(1,117) = 1 = 0000001
> ABS(x-y)/2 = 116 = 1110100
>
> 01,01,01,00,01,00,10 -> {1110102}
>
> OR, suppose we express 9 * 13 as 3 * 39 = 117
>
> MIN(3,39) = 3 = 00011
> ABS(x,-y)/2 = 18 = 10010
>
> 01,00,00,11,10 -> {10032}
>
> Now the real challenge is to find out how to get from one to
> the other and show that
>
> {2012} = {10032} = {1110102}
This is NOT what Risto is proposing; Risto is proposing that
{2012} # {2012} = {10032} # {10032} = {1110102} # {1110102},
which is plausible.
I think Risto's algorithm to multiply works, and that it can be proven
by writing out x and (y - x)/2 (where x <= y) in binary and following
the process through. (This is what the MIN and ABS are doing; setting
up x and (y - x)/2.) I gave out the original post to my problem solving
class, and we may be working through it as a group.
> Because if you can get from {1110102} to any other form, then
> you have the factors.
>
> Another example, 11*31 = 341
>
> (11 ,31) -> {3032}
> (1,341) -> {10101012}
>
> And {3032} = {10101012}
>
> But the trick is to get from the string {10101012}
> to the string {3032}.
... and that's the part which will probably prove tricky; finding all
solutions to the equation x # x = {10101012} # {10101012} probably
can't be done easily, say in polynomial time. (We need to determine
whether there are at least one solution, because one is already present
in the equation.
On yet another tangent, try solving the equation x # {3032} = 341,
which should be "easier" than solving x # x = N.
--- Christopher Heckman
- Next message: Jason: "Re: John Gabriel's Theorem Revisited."
- Previous message: mensanator_at_aol.compost: "Re: Collatz conjecture"
- In reply to: fiziwig: "Re: New integer multiplication algorithm"
- Next in thread: Oscar Lanzi III: "Re: New integer multiplication algorithm"
- Reply: Oscar Lanzi III: "Re: New integer multiplication algorithm"
- Messages sorted by: [ date ] [ thread ]
Relevant Pages
|