web proxy访问我的blogspot

博客被封了，一直没有来维护，也一直挂着，最近找了一个还不错的web代理网站，居然可以访问被封已久的blogger，速度还是可以的。令人兴奋。the address is showed as follow:http://www.sneakme.net/

Multiple kernel learning's People

1.Gert Lanckriet----- Homepage

My research interests are on the interplay between machine learning, applied statistics and convex optimization techniques. I am interested in developing methods for pattern discovery from extremely large-scale data sets, taking uncertainty into account and respecting computational constraints. More precisely, my research focuses on the integration of multiple, heterogeneous data types for a variety of pattern discovery tasks where the amount of data is extremely large and the solutions desired to be sparse. An important challenge in the field of machine learning is to deal with the increasing amount of data that is available for learning and to leverage the (also increasing) diversity of information sources, describing these data. Beyond classical vectorial data formats, data in the format of graphs, trees, strings and beyond have become widely available for data mining, e.g., the linked structure of communication networks, amino acid sequences describing proteins, etc. Moreover, for interpretability, stability and economical reasons, decision rules that rely on a small subset of the information sources and/or a small subset of the features describing the data are highly desired: sparse learning algorithms are a must. My research is inspired by practical applications in computational genomics, financial engineering and computer music.

publication:

Lanckriet, G.R.G., Cristianini, N., Bartlett, P., El Ghaoui, L., Jordan, M.I. (2004).Learning the Kernel Matrix with Semidefinite Programming . Journal of Machine Learning Research, 5, 27-72, 2004.

2.Francis Bach-----homepage

I am a researcher at INRIA, working in the Willow project, which is located at Ecole Normale Superieure. I completed my Ph.D. in Computer Science at U.C. Berkeley, working with Professor Michael Jordan, and spent two years in the Mathematical Morphology group at Ecole des Mines de Paris. I am interested in statistical machine learning, and especially in graphical models, sparse methods, kernel-based learning, vision and signal processing. [CV (English)] [CV (French)]

publication:

F. Bach, G. R. G. Lanckriet, M. I. Jordan. Multiple Kernel Learning, Conic Duality, and the SMO Algorithm. Proceedings of the Twenty-first International Conference on Machine Learning, 2004 [pdf] [tech-report]

3.Alain Rakotomamonjy----Homepage

Research Interests
Kernel Methods and Support Vector Machines Algorithm
Regularization paths Multiple Kernel
Kernel Design Sparsity and variable selection Wavelet and Time-Frequency Signal AnalysisSignal Classification Brain-Computer Interfaces  Object Recognition

publication:

A. Rakotomamonjy, F. Bach, Y. Grandvalet, S. Canu, SimpleMKL,  Journal of Machine Learning Research, Vol. 9, pp 2491-2521, 2008. [JMLR page][PDF] [code]

4.M. Gönen----Homepage

Research Interests
Support Vector MachinesKernel MethodsSimulation and Real-time Control of Flexible Manufacturing Systems

publication:

M. Gönen and E. Alpaydın (2008) ”Localized Multiple Kernel Learning”, In Proceedings of the 25th International Conference on Machine Learning, 352-359.

5.S. Sonnenburg----Homepage

I am currently a postdoc at the Machine Learning in Biology Group at the Friedrich Miescher Laboratory of the Max Planck Society in Tübingen. I have been working in the IDA group at the Fraunhofer Institute FIRST.

I am intrigued by sequence based machine learning methods involving large data sets and have developed several machine learning methods for bioinformatics applications such as splice site recognition, promoter detection and gene finding. I also worked on microarray analysis and motif discovery.

publication:

S. Sonnenburg, G. R¨atsch, C. Sch¨afer, and B. Sch¨olkopf. Large scale multiplekernel learning. Journal of Machine Learning Research, 7, 2006.

how to use CRF++

in recently ,i learned the CRF for information extraction ,studyed its theory and  collected the CRF tools.first of all,i get the information of CRF from wiki websit where have so many CRF resource,including theory ,people who are studying ,offering a lot of valid papers links,

References

• Lafferty, J., McCallum, A., Pereira, F.: Conditional random fields: Probabilistic models for segmenting and labeling sequence data. In: Proc. 18th International Conf. on Machine Learning, Morgan Kaufmann, San Francisco, CA (2001) 282–289
• McCallum, A.: Efficiently inducing features of conditional random fields. In: Proc. 19th Conference on Uncertainty in Artificial Intelligence. (2003)
• Sha, F., Pereira, F.: Shallow parsing with conditional random fields. Technical Report MS-CIS-02-35, University of Pennsylvania (2003)
• Wallach, H.M.: Conditional random fields: An introduction. Technical Report MS-CIS-04-21, University of Pennsylvania (2004)
• Sutton, C., McCallum, A.: An Introduction to Conditional Random Fields for Relational Learning. In "Introduction to Statistical Relational Learning". Edited by Lise Getoor and Ben Taskar. MIT Press. (2006)
• Klinger, R., Tomanek, K.: Classical Probabilistic Models and Conditional Random Fields. Algorithm Engineering Report TR07-2-013, Department of Computer Science, Dortmund University of Technology, December 2007. ISSN 1864-4503. Online PDF

what the most importance thing  is the there  are a good many CRF tools.The listed tools is the most popular tools for using.   

Software

This is a partial list of software that implement CRF related tools.

• MALLET (Java)
• ABNER (Java)
• MinorThird (Java)
• Kevin Murphy's MATLAB CRF code (Matlab)
• Sunita Sarawagi's CRF package (Java)
• HCRF library (including CRF and LDCRF) (C++, Matlab)
• CRFSuite Fast CRF implementation (C++)
• Xcrf for Xml data (Java)
• CRF++ (C++)
• sgd: An LGPL C++ library implementing Stochastic Gradient Descent with application to learning CRF andSupport Vector Machine
• FlexCRFs (including a parallel implementation) (C++)

I choose the CRF++ (C++) for my project.so I take attenttion in how to using this software.

I have been basically understood its usage through several days learning.if you are a beginner ,you can get quick start from its online document .when you using this tool,some note you should attention:

      1.the column of Token must be seperated by space or tab.It imply that the partition word must not include the space or tab.for instance "hello world    NB    1","hello world" is not allowed for including space.

      2.if you want to use the API for your project in MS operating system.you can look the fold:\sdk which tell you how to call the dll .

      3.i found that i can do when using java for my project on ms operating system.so i can but rewrite it, i do not know what is the reason.

数学符号和希腊字母中英文发音列表

序号 大写 小写 英文注音 国际音标注音 中文注音 
1    Α    α    alpha    a:lf       阿尔法 
2    Β    β    beta     bet        贝塔 
3    Γ    γ    gamma    ga:m       伽马 
4    Δ    δ    delta    delt       德尔塔 
5    Ε    ε    epsilon  ep`silon   伊普西龙 
6    Ζ    ζ    zeta     zat        截塔 
7    Η    η    eta      eit        艾塔 
8    Θ    θ    thet     θit        西塔 
9    Ι    ι    iot      aiot       约塔 
10   Κ    κ    kappa    kap        卡帕 
11   Λ    λ    lambda   lambd      兰布达 
12   Μ    μ    mu       mju        缪 
13   Ν    ν    nu       nju        纽 
14   Ξ    ξ    xi       ksi        克西 
15   Ο    ο    omicron  omik`ron   奥密克戎 
16   Π    π    pi       pai        派 
17   Ρ    ρ    rho      rou        肉 
18   Σ    σ    sigma   `sigma      西格马 
19   Τ    τ    tau      tau        套 
20   Υ    υ    upsilon  jup`silon  宇普西龙 
21   Φ    φ    phi      fai        佛爱 
22   Χ    χ    chi      phai       西 
23   Ψ    ψ    psi      psai       普西 
24   Ω    ω    omega    o`miga     欧米伽

1  Α  α  alpha  a:lf  
阿尔法  角度；系数 
2  Β  β  beta  bet  
贝塔  磁通系数；角度；系数 
3  Γ  γ  gamma  ga:m  
伽马  电导系数（小写） 
4  Δ  δ  delta  delt  
德尔塔  变动；密度；屈光度 
5  Ε  ε  epsilon  ep`silon  伊普西龙  对数之基数 
6  Ζ  ζ  zeta  zat  
截塔  系数；方位角；阻抗；相对粘度；原子序数 
7  Η  η  eta  eit  
艾塔  磁滞系数；效率（小写） 
8  Θ  θ  thet  
θit  
西塔  温度；相位角 
9  Ι  ι  iot  
aiot  
约塔  微小，一点儿 
10  Κ  κ  kappa  kap  
卡帕  介质常数 
11  ∧  λ  lambda  lambd  
兰布达  波长（小写）；体积 
12  Μ  μ  mu  mju  缪  磁导系数；微（千分之一）；放大因数（小写） 
13  Ν  ν  nu  nju  纽  磁阻系数 
14  Ξ  ξ  xi  ksi  克西  　 
15  Ο  ο  omicron  omik`ron  奥密克戎  　 
16  ∏  π  pi  pai  派  圆周率=圆周÷直径=3.14159 26535 89793 
17  Ρ  ρ  rho  rou  肉  电阻系数（小写） 
18  ∑  σ  sigma  `sigma  西格马  总和（大写），表面密度；跨导（小写） 
19  Τ  τ  tau  tau  套  时间常数 
20  Υ  υ  upsilon  jup`silon  宇普西龙  位移 
21  Φ  φ  phi  fai  佛爱  磁通；角 
22  Χ  χ  chi  phai  西    
23  Ψ  ψ  psi  psai  普西  角速；介质电通量（静电力线）；角 
24  Ω  ω  omega  o`miga  欧米伽  欧姆（大写）；角速（小写）；角 

How to write and add your own algorithm to Weka

Recently，I am studying the weka .My  purpose is that add my own algorithm to Weka.

  

I have collected some resource about these aspects from Weka website。 The links show as follow：

1. Frequently Asked Questions

        In this webpage,it list many frequently asked questions by two catalogue：uing weka and developing weka。

       2.How to write a new classifier to weka

Due to siginificant changes in the API beyond version 3.5.2, there are now two different HOWTOs available.

Please choose the version of Weka you want to develop for:

• <= 3.5.2 (includes 3.4.x)
• > 3.5.2

3.How do I add a new classifier, filter, kernel, etc.?

As of Weka 3.4.4, all the derived classes of superclasses that can be edited in the GenericObjectEditor, like subclasses ofweka.classifiers.Classifier for instance, can be determined dynamically at runtime. Read here for more information.
Note: Weka 3.5.8 turned the automatic discovery off by default.

The GenericObjectEditor is the core component in Weka for modifying schemes, like classifiers and filters in the GUI. It has to be configured correctly in order to show default and additional schemes. See the following articles for more details:

• GenericObjectEditor (book version)
• GenericObjectEditor (developer version)

My Experience：

1.netbean+weka for developing    english and chinese

2.if you want to use the weka3.5.8,due to Weka 3.5.8 turned the automatic discovery off by default.so you must make the adjust:

Automatic class discovery was turned off in 3.5.8. GUIs load faster   

without it turned on and it doesn't work when Weka is loaded as a   
plugin via a custom class loader. However, it is easily re-enabled. Just edit the   
GenericPropertiesCreator.props file (located in weka/gui) and change   
the "UseDynamic" property from "false" to "true". 

3.you can be easy to add a new algorithm to weka.if you are famliar with chinese .the follow website is a good introduction:http://blog.chinaunix.net/u1/34560/showart_269880.html

4.when adding a new algorithm to weka.you may take attention to the problem,namely ,that  your own algorithm can not display in the GUI(i.e  Exprolor) even you have modify the "UseDynamic".For this problem.you should adjust your new package's position,For Instance:you want to add a new package"weka.classifier.cxk".you must configure the GenericPropertiesCreator.props  again:

Original:

weka.classifiers.Classifier=\ 

weka.classifiers.bayes,\ 

weka.classifiers.functions,\

weka.classifiers.lazy,\

weka.classifiers.trees,\ 

weka.classifiers.meta,\

weka.classifiers.cxk,\

 weka.classifiers.rules

Modify

weka.classifiers.Classifier=\ 

weka.classifiers.cxk,\

weka.classifiers.bayes,\ 

weka.classifiers.functions,\

weka.classifiers.lazy,\

weka.classifiers.trees,\ 

weka.classifiers.meta,\

 weka.classifiers.rules

CRF 下载地址

CRF工具下载地址

CRF(Java):http://sourceforge.net/project/showfiles.php?group_id=105386

CRF++(C++):http://sourceforge.net/projects/crfpp

FlexCRF(C++):http://sourceforge.net/project/showfiles.php?group_id=184307

pocket CRF(C++): http://sourceforge.net/project/showfiles.php?group_id=201943

关于核矩阵的两种极端情况分析

核包括所有可能提供给学习机的关于输入在特征空间中的相对位置的信息。很自然的，如果要在数据集内发现几何结构，数据必须通过核矩阵展现这一结构。
如果这个核过于一般，对于具体类型的相似性没有给出足够的意义。用我们讨论先验知识的语言来说，这相当于为太多的分类赋予权重（可以看成是太多的坐标轴，即基太大）。所以，核把具有相等权值的任何一对输入看成相似（只有自己核自己相似）或者不相似（两两不同的向量是不相似的）。这样的核矩阵中除对角线以外的其他元素变得非常小，而同时对角线元素接近于1。因此核只能表示成单位矩阵的概念。这导致了过度拟合问题，因为我们把训练集正确分类很容易，但是核没有办法对新数据进行泛化处理。
另一种极端情况是，如果核矩阵是完全均一的，那么每一个输入都和每个其他的输入相似。这相当于每个输入都被映射到同一个特征向量，这导致数据欠拟，因为这是能够容易的表达出来的函数只有那些把所有的点都映射到同一个类的函数。
从几何上将，第一情形相当于把输入映射到特征空间中的标准正交点，而第二种情况，所有点都被合并到同一个映像。
这两种情况数据中都不存在非平凡的自然类，从而没 有真正的结构可以泛化。