MATLAB: How to define a kernel for the support vector machine using the BioInformatics Toolbox in MATLAB 7.8 (R2009a)

Bioinformatics Toolboxkernelmachinesupportvector

I am trying to define a kernel for my support vector machine, but I do not understand the documentation. For example, the kernel function must have inputs u and v, but I do not know what these inputs are.

Best Answer

In the kernel function, u and v can be thought of as vectors from the vector space of the data. The kernel function defines an inner product on that vector space. When the kernel function is called during SVMTRAIN, the entire set of training data is passed as an MXN matrix. Here, M is the number of data points, and N is the length of a single data vector. The returned matrix is an MXM matrix for which the (i,jth) element of that matrix gives the inner product between the ith and jth training vector. So for example, to let the kernel be the L2-norm, one could use:

function y = kfun(u,v)
y = u*v';

Related Solutions

MATLAB: Does the Statistics Toolbox support hybrid kernels to train Support Vector Machines in MATLAB

The Statistics Toolbox R2014a offers the following kernels with the "svmtrain" function:

- Linear

- Quadratic

- Polynomial

- Gaussian Radial Basis Function

- Multilayer Perceptron

Hybrid kernels are not readily available. However, there is an option to provide a custom kernel function using a function handle.

To train an SVM using a hybrid kernel, you would have to write the code for the hybrid kernel function. The kernel function must be of the form:

       function K = kfun(U,V)

where the returned value, "K", is a matrix of size M-by-N, and "U" and "V" have "M" and "N" rows respectively.

You could use the custom kernel "kfun" by specifying the "kernel_function" argument as follows:

       load fisheriris
       xdata = meas(51:end,3:4);
       group = species(51:end);
       svmStruct = svmtrain(xdata,group,'ShowPlot',true,'kernel_function',@kfun);

Here is an example of a hyperbolic tangent kernel that could be used with "svmtrain":

       function K = kfun(U,V)
       K = tanh(U*V');

MATLAB: Do I get an error when using BLASTFORMAT in the BioInformatics Toolbox 3.3 (R2009a)

This most likely indicates that the system Path variable was not set up correctly. There is a ../bin directory that was extracted when you ran blast-2.2.20-ia32-win32.exe after downloading it. This directory must be included in your system Path and the executables in that directory must be working.

In Windows, to ensure that the executables are working and that the file is on your system path please do the following:

1) In Windows Explorer, navigate to the ../bin directory that was extracted when you ran blast-2.2.20-ia32-win32.exe.

2) To verify that 'formatdb.exe' runs, please double click on the executable 'formatdb.exe' in the ../bin directory. This should run the executable. Please verify that when this executable is run a log file, 'formatdb.log,' is created. Please let me know if the log file is not created.

3) Next, go to Start Menu->Settings->Control Panel->System->Advanced->Environment Variables

4) Under 'System Variables', select the 'Path' variable and click 'edit'.

5) Verify that the ../bin directory is included in the system path. If the full name of the ../bin directory is "C:\Program Files\Blast\bin", then the "Variable Value" field should contain something like this:

      <other_program_directories>;C:\Program Files\Blast\bin

Best Answer

Related Solutions

MATLAB: Does the Statistics Toolbox support hybrid kernels to train Support Vector Machines in MATLAB

MATLAB: Do I get an error when using BLASTFORMAT in the BioInformatics Toolbox 3.3 (R2009a)

Related Question