10.12.2012 Views

Prime Numbers

Prime Numbers

Prime Numbers

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

6.6 Research problems 317<br />

be determined,<br />

z = γ ×− γ,<br />

where ×− denotes negacyclic convolution, and z is the signal consisting<br />

of the zj coefficients. But we know how to do negacyclic convolution via<br />

fast transform methods. Writing<br />

<br />

d−1<br />

Γk = γjα j α −2kj ,<br />

j=0<br />

one can establish the weighted-convolution identity<br />

<br />

d−1<br />

−n 1<br />

zn = α Γ<br />

d<br />

k=0<br />

2 kα +2nk .<br />

The deconvolution idea, then, is simple: Given the signal z to be squarerooted,<br />

transform this last equation above to obtain the Γ2 k , then assign<br />

one of 2d−1 distinct choices of sign for the respective ± Γ2 k , k ∈ [1,d−1],<br />

then solve for γj via another transform. This negacyclic deconvolution<br />

procedure will result in a correct square root γ of γ 2 . The research question<br />

is this: Since we know that number fields based on f(x) =x d + 1 are<br />

easily handled in many other ways, can this deconvolution approach be<br />

generalized? How about f(x) =x d + c, orevenmuchmoregeneralf?<br />

It is also an interesting question whether the transforms above need to<br />

be floating-point ones (which does, in fact, do the job at the expense of<br />

the high precision), or whether errorless, pure-integer number-theoretical<br />

transforms can be introduced.<br />

(3) For any of these various ideas, a paramount issue is how to avoid the rapid<br />

growth of coefficient sizes. Therefore one needs to be aware that a squareroot<br />

procedure, even if it is numerically sound, has to somehow keep<br />

coefficients under control. One general suggestion is to combine whatever<br />

square-rooting algorithm with a CRT; that is, work somehow modulo<br />

many small primes simultaneously. In this way, machine parallelism may<br />

be possible as well. As we intimated in text, ideas of Couveignes and<br />

Montgomery have brought the square-root obstacle down to a reasonably<br />

efficient phase in the best prevailing NFS implementations. Still, it would<br />

be good to have a simple, clear, and highly efficient scheme that generalizes<br />

not just to cases of parity on the degree d, but also manages somehow to<br />

control coefficients and still avoid CRT reconstruction.

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!