Dear Sir, thank you for your help. The reason I asked about the magnitude of
C+ was because for my dataset, the optimal value seems to be around 1000. I
was just a little worried I was doing something terribly wrong.
Thanks.
----- Original Message -----
From: "Chew, Hong Gunn" <[log in to unmask]>
To: <[log in to unmask]>
Sent: Friday, August 23, 2002 8:41 AM
Subject: Re: How to handle unbalanced data
> Hi Adal,
>
> > I have one question though, regarding the magnitude of the C+ and C-
> > reported in table 1 of "Target detection in radar imagery using SVM".
You
> > yourself, it is typical to use a value of 1 to 100. But your values in
the
> I don't think I have indicated the use of C = 1 to 100. Similar to what
> Goekhan has mentioned, I use a binary search method to determine the
> optimal C to use.
> > table are from 10^-6 to 1. You did not consider any value of C+/C-
greater
> > than 1, say 1000. Why was this ? Is this particular to the dataset you
were
> >From the table, it can be seen that the performance peaks at around
> C-=100x10^-6, so it is unlikely that the performance would improve for
> C- > 1.
>
> The C to use for each dataset varies, and would need to be determined
> individually.
>
> Cheers,
> Hong-Gunn
>
|