asvm_source

所属分类:数据挖掘/数据仓库
开发工具:matlab
文件大小:106KB
下载次数:69
上传日期:2006-01-23 22:02:30
上 传 者jweinus
说明:  active support vector machine

文件列表:
active.dsp (4721, 2000-05-20)
active.dsw (2498, 2000-05-18)
active.ncb (156672, 2000-06-08)
asvm.m (3023, 2000-06-08)
asvminfo.m (1223, 2000-06-07)
asvmmex.cpp (4779, 2000-06-08)
AsvmTest.cpp (3264, 2000-06-07)
AsvmTest\AsvmTest.dsp (5027, 2000-05-20)
AsvmTrain.cpp (25400, 2000-06-08)
BLAS\dgemm.c (9721, 2000-05-17)
BLAS\dgemv.c (7208, 2000-05-17)
BLAS\dger.c (4527, 2000-05-17)
BLAS\dscal.c (1637, 2000-05-17)
BLAS\dswap.c (1905, 2000-05-17)
BLAS\dtrmm.c (10514, 2000-05-17)
BLAS\dtrmv.c (7888, 2000-05-17)
BLAS\dtrsm.c (11073, 2000-05-17)
BLAS\f2c.h (4485, 2000-05-17)
BLAS\idamax.c (1558, 2000-05-17)
BLAS\lsame.c (2642, 2000-05-17)
BLAS\Makefile (1614, 2000-05-17)
BLAS\xerbla.c (1212, 2000-05-17)
CLAPACK\dgetf2.c (4298, 2000-05-17)
CLAPACK\dgetrf.c (5779, 2000-05-17)
CLAPACK\dgetri.c (6566, 2000-05-17)
CLAPACK\dlaswp.c (3089, 2000-05-17)
CLAPACK\dtrti2.c (4364, 2000-05-17)
CLAPACK\dtrtri.c (6334, 2000-05-17)
CLAPACK\f2c.h (4485, 2000-05-17)
CLAPACK\ilaenv.c (15977, 2000-05-17)
CLAPACK\lsame.c (2490, 2000-05-17)
CLAPACK\Makefile (1480, 2000-05-17)
CLAPACK\xerbla.c (1212, 2000-05-17)
DataSource.cpp (38818, 2000-06-07)
DataSource.H (7132, 2000-06-07)
f2c.h (4485, 2000-05-16)
libF77\f2c.h (4485, 2000-05-17)
libF77\makefile (874, 2000-05-17)
... ...

Active Set Support Vector Machine Classification Software Copyright (C) 2000 David R. Musicant and Olvi L. Mangasarian. This software is free for academic use only. For commercial use, contact musicant@cs.wisc.edu. -------------------------------------------------------------- How to use ----------- The executable archive contains five files: asvm (or asvm.exe): This is the main program used for training the support vector machine. Syntax: asvm [options] training_file solution_file Type asvm with no arguments at the command prompt for a list of options. Input: A comma delimited file of numeric attributes, where the last column is either a 1 or -1 to indicate to which class the point belongs. For example: 94, 91, 27, 20, 15, 0.5, 1 87, 86, 28, 23, 21, 4.0, -1 89, 82, 23, 14, 35, 0.5, 1 87, 52, 21, 19, 30, 0.5, -1 93, 43, 11, 16, 54, 6.0, 1 Output: A flat file indicating the separating surface found, as well as any necessary normalization coefficients. Sample usage: asvm -converge -nu 500 trainset out.txt Iteration description: ASVM provides a description for each iteration. Here is an explanation of what the descriptions mean: Normal iteration: ASVM has successfully decreased the objective value with a standard iteration. Global grad projection: ASVM has reached the minimum on its face of active constraints, and therefore performs a gradient projection step on the set of _all_ constraints to obtain a new basic set with a decreased objective. Project active set min: ASVM has not reached the minimum on its face of active constraints, and the normal ASVM iteration does not improve the objective value. Therefore, ASVM moves in the direction of the unconstrained minimum on the face, and stops when it reaches an active constraint. This is the "active set strategy" part of the algorithm. Face grad: This is a last step effort when all else fails, which should only happen due to numerical problems such as bad conditioning. ASVM performs a projected gradient step on the face of active constraints, and continues to do so until optimality on the face is acheived or the set of active constraints changes. asvm_test (or asvm_test.exe): This program is used to test the generalization capability of the separating surface which was found with the asvm program. Syntax: asvm_test [options] training_file solution_file Type asvm_test with no arguments at the command prompt for a list of options. Input: A file of the same format as input to asvm. The solution file from asvm. Output: Generalization accuracy. Sample usage: asvm_test test_set out.txt asvm.m: This is an m-file which handles the link between MATLAB and the ASVM mex file. In MATLAB, do "help asvm" for more information. asvminfo.m: This is an m-file which contains additional information about ASVM. asvmmex.mexsol (or asvmmex.dll): This is the binary executable for the MATLAB mexfile. Compiling the source code ------------------------- UNIX ---- On a UNIX platform, type "make" at the command prompt. Windows ------- The Windows platform is significantly more difficult. Open the Visual C++ project "active.dsw". First, you'll need to choose whether you want a "Debug" build or a "Release" build. The Release build runs faster and is recommended - but does not function correctly with MATLAB. Hence you'll need to do a Debug build for purposes of generating the mex file, if this is what you want. Do this by selecting "Build, Set Active Configuration", and choose "active - Win32 Release" (or Debug). Next, choose "Build, Batch Build" from the menu. You should do this a few times until it is clear that all the executables have been built. The executables active.exe and AsvmTest.exe are scattered in different subdirectories scattered throughout the project. You'll need to find them and rename them to asvm.exe and asvm_test.exe. Finally, do create the mexfile, you should do all of the above. Then, go to the command line, and execute the batch file "vcvars32.bat" (this should be somewhere on your system." Finally, run the batch file "makemex.bat" from the ASVM project directory. This should compile and link the mexfile.

近期下载者

相关文件


收藏者