Re: compression of gaussian distribution?
- From: "Nils" <bla@xxxxxxx>
- Date: Fri, 11 Nov 2005 00:18:48 +0100
Some counter-questions:
- Does the compression have to be exact, or is some (very small) deviation
between exact and compressed value allowed?
If it doesn't have to be 100% exact, you could do some things:
1. Transform the distribution so it is no longer gaussian, but approximately
linear
2. Quantise the values
- Is the order of values important?
If the order is not important, you can sort the values, and just code the
difference of previous vs current value (delta). Combined with quantisation
(see pt 2 above) this can bring considerable compression.
Nils
<budgetanime@xxxxxxxxxxxxxx> schreef in bericht
news:1131660187.425238.156640@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
> Hi,
>
> I have a set of data points stored as 32 bit integers. The data points
> form a gaussian distribution between 0 and some value. There is no
> correlation between consequtive data points. The higest value depends
> on parameters to the algorithm which generates the data points. The
> number of data points depends on input to the algorithm.
>
> I am able to find the mean and standard deviation.
>
>>>From this information is it possible to create a compression algorithm
> will compress these data points?
>
> A typical situation is that there is 6000 data points which range from
> 0 to 2^18.
>
.
- Follow-Ups:
- Re: compression of gaussian distribution?
- From: budgetanime
- Re: compression of gaussian distribution?
- References:
- compression of gaussian distribution?
- From: budgetanime
- compression of gaussian distribution?
- Prev by Date: Re: A simple question...
- Next by Date: Re: compression of gaussian distribution?
- Previous by thread: compression of gaussian distribution?
- Next by thread: Re: compression of gaussian distribution?
- Index(es):
Relevant Pages
|