MCMC Methods for MLP and GP and Stuff (for Matlab 6.*)
Maintainer: Aki Vehtari <Aki.Vehtari@hut.fi>
This software is distributed under the GNU General Public Licence
(version 2 or later); please refer to the file Licence.txt,
included with the software, for details.
Table of contents:
1. INTRODUCTION
2. DOCUMENTATION AND DEMOS
3. INSTALLING THE TOOLBOX
------------------------------------------
1. INTRODUCTION
Basic design of this toolbox is based on Netlab
<http://www.ncrg.aston.ac.uk/netlab/>. However, this
toolbox is not compatible with Netlab, because the option handling
has been changed to use structures similar to current default in
Mathworks' toolboxes. Furthermore, the code in this toolbox has been
streamlined and optimized for faster computation, and it has been
extended to include some of the features present in FBM
<http://www.cs.toronto.edu/~radford/fbm.software.html> and some
other features. Some of the most computationally critical parts
have been coded in C. For easier introduction to MLP's and GP's,
Netlab is better suited, especially because of the accompanying text
book (Nabney I.T. (2001) Netlab: Algorithms for Pattern
recognition). Purpose of this toolbox was to port some of the features
in the FBM to Matlab environment for easier development for Matlab
users.
Most of the code has been written by Aki Vehtari in the Laboratory
of Computational Engineering <http://www.lce.hut.fi> at Helsinki
University of Technology (HUT) <http://www.hut.fi>. Currently
there is also code written by (in alphabetical order) Toni Auranen,
Christopher M Bishop, James P. LeSage, Ian T Nabney, Radford Neal,
Carl Edward Rasmussen, and Simo S�rkk�. During the summer of 2004 the
code was fully documented and pulled together as a toolbox by Jarno
Vanhatalo, an undergraduate student at HUT. Speciall thanks are
directed to Prof. Jouko Lampinen who helped in converting the C
sources into mex-files for Windows.
2. DOCUMENTATION AND DEMOS:
In the toolbox there are two demonstration programs, that solve a two
dimensional regression problem with MLP network and Gaussian
process. These are demo_2input and demo_2ingp. The demos are discussed
in more detail in the home page of the toolbox
<http://www.lce.hut.fi/research/compinf/mcmcstuff/demos> and in
documentation of the software that is available from the home page
of the toolbox <http://www.lce.hut.fi/research/compinf/mcmcstuff/>.
Features treated in the documentation are:
- Baysian learning for MLP in a regression problem
- a Gaussian Hierarchical prior structure with Automatic Relevance
Determination (ARD)
- noise model with normal distribution
- Markov Chain Monte Carlo method
- hybrid Monte Carlo and Gibbs sampling
- a Gaussian process in a regression problem
- Bayesian learning for MLP in classification problems with two or
many classes
- a Gaussian process in classification problem with two classes
- noise model with Students t-distribution
- input variable selection for MLP and GP (various methods)
Features not treated in the documentation are:
- covariate dependent groupped noise model
3. INSTALLING THE TOOLBOX
-All files needed to run the toolbox in linux, Alfa and Windows (*.m,
*.mexaxp, *.mexglx, *.dll) are located in the same directory with
this Readme.txt (mcmcstuff/). The toolbox should work in the above
architechtures after the mcmcstuff/ is added to Matlab path.
-In the directory mcmcstuff/ there are also two subdirectories
linuxCsource/ and winCsource/ containing the functions written in C
for Linux, Alfa (linuxCsource/) and Windows (winCsource/). The reason
for two different folders is that the 'math.h' of C library is
different in Linux and Windows and that the function names in BLAS
libraries are different in Windows than in Linux and Alfa. For
example in Linux 'dpotrf_' is in Windows 'dpotrf'.
-The toolbox can be tested by running demonstration program
demo_2ingp (NB! the demo takes quit a long time, approximately 30
minutes on a 2400MHz Intel Pentium workstation, it is sufficient
enough to run the demo for few minutes to be sure that it works). If
there is no right mex-files for the architechture of your computer,
you get an error message "error, No mex-file for this
archtitecture. See Matlab help and convert.m in ./linuxCsource or
./winCsource for help."
-There are two different compiler m-files, compile.m in
mcmcstuff/linuxCsource/ and compile_windows.m in
mcmcstuff/winCsource/. The *.c files can be converted to mex-files
by running these files in Matlab in the same directory where the
it is located. If the compilers do not work directly the relations
between C-functions can still be seen from them.
没有合适的资源?快使用搜索试试~ 我知道了~
mcmc代码蒙特卡洛马尔科夫链
共389个文件
m:169个
c:58个
mexw32:50个
4星 · 超过85%的资源 需积分: 29 107 下载量 161 浏览量
2011-12-31
11:13:01
上传
评论 3
收藏 11.29MB ZIP 举报
温馨提示
关于MCMC的一份不错的代码 很不错 可用于马尔可夫场图像分割
资源推荐
资源详情
资源评论
收起资源包目录
mcmc代码蒙特卡洛马尔科夫链 (389个子文件)
dat.1 11KB
fort.6 64B
gp2fwd.asv 7KB
compile_windows.asv 3KB
gp2fwd.c 12KB
gp2fwd.c 12KB
gp2fwds.c 9KB
gp2fwds.c 9KB
ars.c 8KB
ars.c 8KB
gpvalues.c 7KB
gpvalues.c 7KB
gpexptrcovinv.c 5KB
gpexptrcovinv.c 5KB
rand.c 5KB
rand.c 5KB
gpexpedata.c 5KB
gpexpedata.c 5KB
gpexptrcov.c 5KB
gpexptrcov.c 4KB
mlp2bkp.c 4KB
mlp2bkp.c 4KB
bbprctile.c 3KB
bbprctile.c 3KB
mlp2fwd.c 3KB
mlp2fwd.c 3KB
cond_invgam_invgam1.c 3KB
cond_invgam_invgam1.c 3KB
gpexpcov.c 3KB
gpexpcov.c 3KB
invgamrand.c 3KB
invgamrand.c 3KB
gamrand.c 3KB
gamrand.c 3KB
resampres.c 2KB
resampres.c 2KB
bbmean.c 2KB
bbmean.c 2KB
trand.c 2KB
trand.c 2KB
resampdet.c 2KB
resampdet.c 2KB
digamma1.c 2KB
digamma1.c 2KB
catrand.c 2KB
catrand.c 2KB
resampstr.c 2KB
resampstr.c 2KB
dirrand.c 2KB
dirrand.c 2KB
resampsim.c 2KB
resampsim.c 2KB
exprand.c 2KB
exprand.c 2KB
invgamrand1.c 1KB
invgamrand1.c 1KB
gamrand1.c 1KB
gamrand1.c 1KB
tanh_f.c 872B
tanh_f.c 872B
binsgeq.c 721B
binsgeq.c 721B
cdata 38KB
ChangeLog 8KB
ChangeLog~ 8KB
convert_for_win 188B
gp2fwds.c~ 9KB
gp2fwd.c~ 8KB
gpvalues.c~ 7KB
gpexptrcov.c~ 4KB
bbprctile.c~ 3KB
rand.h 1KB
rand.h 1KB
ars.h 910B
ars.h 910B
binsgeq.h 266B
binsgeq.h 266B
mlp2r_mc.m 26KB
mlp2c_mc.m 25KB
gp2b_mc.m 22KB
mlp2b_mc.m 22KB
gp2r_mc.m 21KB
sls.m 14KB
Contents.m 13KB
hmc2.m 10KB
demo_3class.m 9KB
demo_rjmcmc.m 8KB
demo_2class.m 7KB
demo_tmlp.m 7KB
demo_tgp.m 6KB
demo_rjmcmcgp.m 6KB
demo_2classgp.m 5KB
demo_2input.m 5KB
scges.m 5KB
metrop2.m 5KB
gp2r_g.m 4KB
scg2.m 4KB
sls_opt.m 4KB
gp2fwds.m 3KB
gp2r_e.m 3KB
共 389 条
- 1
- 2
- 3
- 4
xuxuxiaoxiaomingming
- 粉丝: 0
- 资源: 1
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
- 1
- 2
- 3
前往页