没有合适的资源?快使用搜索试试~ 我知道了~
资源推荐
资源详情
资源评论
SVM and Kernel Metho ds
Bernhard Scholkopf
Max-Planck-Institut f ur biologische Kybernetik
72076 T ubingen, Germany
Biowulf Technologies
305 Broadway, New York, NY 10007, USA
bernhard.schoelkopf@tuebingen.mpg.de
(these slides are available from www.kernel-machines.org)
B. Scholkopf, NIPS, 3 December 2001
Roadmap
1. ideas of statistical learning theory
2. kernels and feature spaces
3. Supp ort vector algorithms
B. Scholkopf, NIPS, 3 December 2001
Statistical Learning Theory
1. started by Vapnik and Chervonenkis in the Sixties
2. mo del: we observe data generated by an unknown sto chastic
regularity
3.
learning
= extraction of the regularity from the data
4. the analysis of the learning problem leads to notions of
capacity
of the function classes that a learning machine can implement.
5.
support vector machines
use a particular typ e of function class:
classiers with large \margins" in a feature space induced by a
kernel
.
[49, 50]
B. Scholkopf, NIPS, 3 December 2001
Example: Regression Estimation
y
x
Data:
input-output pairs (
x
i
; y
i
)
2
R
R
Regularity:
(
x
1
; y
1
)
; : : :
(
x
m
; y
m
) drawn from P(
x; y
)
Learning:
cho ose a function
f
:
R
!
R
such that the error,
averaged over P, is minimized.
Problem:
P is unknown, so the average cannot b e computed
| need an \
induction
principle"
B. Scholkopf, NIPS, 3 December 2001
Pattern Recognition
Learn
f
:
X ! f
1
g
from examples
(
x
1
; y
1
)
; : : : ;
(
x
m
; y
m
)
2 X f
1
g
;
generated i.i.d. from P(
x; y
)
;
such that the exp ected misclassication error on a test set, also
drawn from P(
x; y
),
R
[
f
] =
Z
1
2
j
f
(
x
)
y
)
j
d
P(
x; y
)
;
is minimal
(Risk Minimization (RM))
.
Problem
: P is unknown.
!
need an
induction principle
.
Empirical risk minimization (ERM):
replace the average over
P(
x; y
) by an average over the training sample, i.e.
minimize the
training error
R
emp
[
f
] =
1
m
X
m
i
=1
1
2
j
f
(
x
i
)
y
i
j
B. Scholkopf, NIPS, 3 December 2001
剩余61页未读,继续阅读
资源评论
xiaoY2011
- 粉丝: 0
- 资源: 3
上传资源 快速赚钱
- 我的内容管理 展开
- 我的资源 快来上传第一个资源
- 我的收益 登录查看自己的收益
- 我的积分 登录查看自己的积分
- 我的C币 登录后查看C币余额
- 我的收藏
- 我的下载
- 下载帮助
安全验证
文档复制为VIP权益,开通VIP直接复制
信息提交成功