In-built packages are available in MATLAB for support vector machine and neural networks implementation. How can I analyze and test data sets? Can I use these packages or do I have to write altogether my own code to do so, either in MATLAB or C or any other programming language? Which is the correct way to do so?
There are packages for SVM and ANN are available in R which are open source, platform independent so you may modify it according to your requirement otherwise use as it is.
As Ram answered there are some powerful packages in different programming environments. PRtools is another useful free available matlab toolbox that have extra facilities in pattern recognition.
If you want to use neural network or SVM in traditional ways these toolboxes are sufficient But if you want to change some part of the algorithms its better to write the program by yourself. If you plan to Write the programs get help from open source codes.
Dear Manas,
Writing a code always takes a time to implement. I think it's better to ask other researchers in order to share their experience about the package. For example, I work with matlab NN package, and I found that my own version works better than this package.
best regards,
Hamid
Generally, it is always best to use existing packages. This is particularly the case for machine learning applications, since the algorithms can be quite complex to implement. There are a number of open source and commercial packages out there -- some of these have already been mentioned in previous responses. Using such packages has the advantage that they are often very well maintained and have grown to be very mature. I would like to add some additional ones, which are program libraries (in case you would like to work outside some proprietary environment):
That said, many open source packages have poor documentation and can be tricky to get started with. Sometimes they also don't fit the specific needs of your application. For instance, for a project where I use Bayesian Learning, I ended up writing my own library since the existing libraries did not fit the specific needs of my application.
For the case of python I would add:
scikit-Learn: http://scikit-learn.org/stable/
If you intend to use SVM, I raccomand to use this:
http://www.csie.ntu.edu.tw/~cjlin/libsvm/
In one place, the Support Vector Machine librarys already written in lot of langauges.
http://www.csie.ntu.edu.tw/~cjlin/libsvm/
- Badges
- Science topic
- Similar topics
- Computer Science and Engineering
- Programming Languages