SatPrem Reddy
SLT,as I think ,is based on following points .
1.the VC dimension thoery, discussing the capicity of the learn machine.The bigger
the VC dimesion of a learning machine, the more it can simulating any decision function
of any type.
2.the Generalization ability of a learning machine.It is inverse of the VC dimension of a learning machine.
3.So Structrual Risk Minimization (SRM) is proposed.To get a tradeoff between the training error and the structural risk.
When we use SVM, searching for the widest margin, it is the process we try to minimize the VC dimension of the learning machine.what's more, we could tuning the parameter C to make a SRM to make balance between the conflict of training error and Structurl error.
This is my own opnion, maybe some mistake included.
-------------------
>Hi all,
> What is the connection between statistical learning theory
>and SVM? How are SVMs based on SLT? I have read papers
>on SLT and SVM but none mention how they are related.
> TIA.
>
>-SatPrem
longbinchen
[log in to unmask]
------------
P.O.BOX 2728# ,NLPR,
BeiJing P.R.China,100080
TEL:+86-10-82613867
13661238601
Institute of Automation,
Chinese Academy of Sciences (CASIA)
P.R.China
|