
Ublas : 
Subject: Re: [ublas] Question/request for matrix powers and other stuff
From: Jens Seidel (jensseidel_at_[hidden])
Date: 20081022 16:59:26
On Wed, Oct 22, 2008 at 03:38:49PM 0400, Daryle Walker wrote:
> On Oct 22, 2008, at 4:02 AM, Jens Seidel wrote:
>> On Tue, Oct 21, 2008 at 04:18:52PM 0400, Daryle Walker wrote:
> [SNIP]
>>> The code I'm adapting sometimes enters a LOT of zerovalues in a row.
>>> The original author realized that, since inserting a zero can be
>>> expressed as a linear transformation, using the matrix form and
>>> raising
>>
>> Heh?
>
> or in rowvector * matrix form:
>
> [0 1 0 ... 0 ]
> [0 0 1 ... 0 ]
> [NewL0 ... NewL7] = [L0 ... L7] * ...
> [0 0 0 ... 1 ]
> [G0 G1 G2 ... G7]
OK.
>>> it to a power can _save_ time over inserting each zero manually.
>>> (The
>>> original author used a bitpacking scheme for the GF(2) matrices
>>> involved, which I'll add later.) The manual method is by definition
>>> linear on the number of zeros added. Using matrices and powers,
>>> especially with squareandmultiply, should represent a savings when
>>> the
>>> length gets long enough.
>>
>> I don't understand this :)
>
>
> You do know that a matrix multiplication can be used to carry out a
> linear transformation, right? Multiple transformations in a row can be
Yep. I realized now that my problem understanding you is my poor
English. I considered "in a row" as "in the matrix row" but it stands
for "behind each other" (or similar).
Read your mail with this in mind and you will be confused as well :)
Will think about your problem tomorrow ...
Jens