m |
(New section: One Comparison Between SVM and Neural Network - Paper) |
||
| (One intermediate revision by one other user not shown) | |||
| Line 1: | Line 1: | ||
| − | |||
You can find an example for classifying a data using Matlab in the following link | You can find an example for classifying a data using Matlab in the following link | ||
| − | + | ||
| − | + | http://www.igi.tugraz.at/lehre/EW/tutorials/nnt_intro/index.html | |
In the file 'nnt_intro_classification.m' it creates NN with 5 hidden units, 3 output units, logsig activation function for each layer. You need to change the configuration to improve the performance and refer to the pdf file in the downloaded zipped file. | In the file 'nnt_intro_classification.m' it creates NN with 5 hidden units, 3 output units, logsig activation function for each layer. You need to change the configuration to improve the performance and refer to the pdf file in the downloaded zipped file. | ||
| − | + | ||
You can download SVM library in the following link | You can download SVM library in the following link | ||
| − | + | http://www.csie.ntu.edu.tw/~cjlin/libsvm/ | |
You need to download the following file from the link | You need to download the following file from the link | ||
| − | + | MATLAB A simple MATLAB interface LIBSVM authors at National Taiwan University. 2.85 Zip | |
When you run 'make_win.m' in the Matlab command prompt it will generate dll files which are callable in Matlab. | When you run 'make_win.m' in the Matlab command prompt it will generate dll files which are callable in Matlab. | ||
Before running 'make_win.m' you need to setup the default compiler used by mex by running 'mex -setup' | Before running 'make_win.m' you need to setup the default compiler used by mex by running 'mex -setup' | ||
| + | Datasets at UCI Irwin Machine Learning laboratory .. | ||
| + | http://archive.ics.uci.edu/ml/datasets.html | ||
| + | |||
| + | [[Category:Matlab Neural Network Toolbox for Classification]] | ||
[[Category:Downloadable SVM libraries in C and how to run in Matlab]] | [[Category:Downloadable SVM libraries in C and how to run in Matlab]] | ||
| + | |||
| + | == One Comparison Between SVM and Neural Network - Paper == | ||
| + | |||
| + | |||
| + | There is one paper which compare the performance of SVM and Neural Network for Drug/Non-Drug classificaiton. The below is the link to the paper. | ||
| + | |||
| + | http://pubs.acs.org/cgi-bin/article.cgi/jcisd8/2003/43/i06/pdf/ci0341161.pdf | ||
| + | |||
| + | In the paper,the author applies two method as binary decision problems to classify drug and non-drug materials. The result shows that SVM performs better than Neural Network since SVM has higher accuracy in predicting for new data set. | ||
Latest revision as of 19:30, 30 March 2008
You can find an example for classifying a data using Matlab in the following link
http://www.igi.tugraz.at/lehre/EW/tutorials/nnt_intro/index.html
In the file 'nnt_intro_classification.m' it creates NN with 5 hidden units, 3 output units, logsig activation function for each layer. You need to change the configuration to improve the performance and refer to the pdf file in the downloaded zipped file.
You can download SVM library in the following link
http://www.csie.ntu.edu.tw/~cjlin/libsvm/
You need to download the following file from the link MATLAB A simple MATLAB interface LIBSVM authors at National Taiwan University. 2.85 Zip
When you run 'make_win.m' in the Matlab command prompt it will generate dll files which are callable in Matlab. Before running 'make_win.m' you need to setup the default compiler used by mex by running 'mex -setup'
Datasets at UCI Irwin Machine Learning laboratory .. http://archive.ics.uci.edu/ml/datasets.html
One Comparison Between SVM and Neural Network - Paper
There is one paper which compare the performance of SVM and Neural Network for Drug/Non-Drug classificaiton. The below is the link to the paper.
http://pubs.acs.org/cgi-bin/article.cgi/jcisd8/2003/43/i06/pdf/ci0341161.pdf
In the paper,the author applies two method as binary decision problems to classify drug and non-drug materials. The result shows that SVM performs better than Neural Network since SVM has higher accuracy in predicting for new data set.
