Re: A problem in fitting curve
- From: makc.the.great@xxxxxxxxx
- Date: 30 Jun 2006 04:29:42 -0700
damn, I missed that x(i+1)-x(i) is not constant. well, then divide d(i)
by it.
makc.the.gr...@xxxxxxxxx wrote:
matt271829-news@xxxxxxxxxxx wrote:
Yu Bo wrote:
There are n samples: (x1,y1),...,(xn,yn), which are supposed to
fit a curve y=f(x,a). a is a parameter to calculate.
However, there are several noise samples which deviate from other
samples.
How can I reduce the negative impact of noise samples during the
fitting curve process and obtain an optimal value of parameter a?
Erm... ignore the noise samples? (How do you know which are noise?)
How do you know which are noise:
calculate d(i) := |2*|y(i)| - |y(i-1)| - |y(i+1)| |, sort samples by d,
and throw out top N, where N is determined by max |d(j) - d(j+1)| over
sorted list.
.
- References:
- A problem in fitting curve
- From: Yu Bo
- Re: A problem in fitting curve
- From: matt271829-news
- Re: A problem in fitting curve
- From: makc . the . great
- A problem in fitting curve
- Prev by Date: Re: A problem in fitting curve
- Next by Date: From the free iPod people...
- Previous by thread: Re: A problem in fitting curve
- Next by thread: Re: A problem in fitting curve
- Index(es):
Relevant Pages
|
|
