English
资源说明：The original LibSVM Python version does not have the interface to computethe value of |w|^2, which can be used for computing the distance of a given sampletowards the hyperplane. We modified the C++ files and corresponding python functions, also re-built the dll file. The modified svm-model object contains the following additional attributes: w_2(|w|^2)+),rho, nSV(no of support vectors for each class), probA, probB. Please refer to readme.wdi file for detailed information.
THIS FILE INCLUDES USAGE INSTRUCTIONS
We provide a freely available Segmentation Toolbox that
can readily be used for image segmentation tasks.
Please acknowledge the authors if you are using this
ToolBox.
Readme-file
===========
Author: Di, Wei
Email : vanessa.wdi@gmail.com
Date : April 3rd, 2011
General Info
=============
The original LibSVM Python version does not have the interface to compute
the value of |w|^2, which can be used for computing the distance of a given sample
towards the hyperplane.
We modified the C++ files and corresponding python functions, also re-built the dll file.
Now the svm-modle object contans the following additional attrs:
+) w_2 (|w|^2)
+) rho
+) nSV
+) probA
+) probB
For detailed information about the above variables, please check svm.h
Note:
1. These files are run under window paltform
2. Modification is currently only for C-SVM
3. The re-built dll file has been already put under folder ./python
4. Test run-file: wdi_libsvm_extend_testing.py
5. Testing data are in folder ./data/letters
Modification in C++ files
===========================
In total, 3 structs are modified (w_2 used to denote |w|^2)
-1- add member of w_2 to svm_model (svm.h)
-2- add member of w_2 to SolutionInfo (svm.cpp)
-3- add member of w_2 to decision_function (svm.cpp)
w_2 is computed in function: 'solve_c_svc' by:
si->w_2 = 2*(si->obj + sum_alpha);
and then recored in SolutionInfo si.
The path follows:
si(SolutionInfo) --> f (decision_function) --> w_2 (svm_model)
Multiple places are updated, please use search 'w_2' to find each
changed line. We use the same common name 'w_2' in each struct, thus,
you will find all changes corresponding to all the 3 different sturcts.
5 additional functions are defined in svm.cpp, which
can return the values in svm-model, including:
w_2 /* in this model, we added this variable w_2[k*(k-1)/2]*/
rho /* constants in decision functions (rho[k*(k-1)/2]) */
nSV /* number of SVs for each class (nSV[k]) */
probA /* pariwise probability information */
robB
please check the svm.h file for the definition of each variable.
These are functions will serve as the interface to Python.
To add more function, you need to also update the svm.h and svm.def
Then, build the files to obtain new libsvm.dll file.
For building dll file under windows, check the Q-A page:
http://www.csie.ntu.edu.tw/~cjlin/libsvm/faq.html#f211
With these additional functions, you also need to update svm.py by
adding new def-functions, and the interface to load c-functions in python
Modification in python files
============================
[+] In svm.py added functions to 'class svm_model':
- get_w2_vector
- get_nSV
- get_rho_vector
- get_probA_vector
- get_probB_vector
And functions interacting with c-files:
# Functions to return w_2, rho, nSV, probA and probB, repectively
fillprototype(libsvm.svm_get_model_w2_vector, None, [POINTER(svm_model), POINTER(c_double)])
fillprototype(libsvm.svm_get_model_rho_vector, None, [POINTER(svm_model), POINTER(c_double)])
fillprototype(libsvm.svm_get_model_nSV_vector, None, [POINTER(svm_model), POINTER(c_int)])
fillprototype(libsvm.svm_get_model_probA_vector, None, [POINTER(svm_model), POINTER(c_double)])
fillprototype(libsvm.svm_get_model_probB_vector, None, [POINTER(svm_model), POINTER(c_double)])
[+]In svmutil.py add helper function to compute the distance to hyperplane:
distance_to_hyperplane
Additional Notes
============================
For copyright information please check "libsvm_COPYRIGHT" file.
* Redistributions of source code must retain the above copyright
notice, this list of conditions and the following disclaimer.
* Redistributions in binary form must reproduce the above copyright
notice, this list of conditions and the following disclaimer in
the documentation and/or other materials provided with the distribution
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
POSSIBILITY OF SUCH DAMAGE.
本源码包内暂不包含可直接显示的源代码文件，请下载源码包。
