Description: Compute a square kernel matrix from a given matrix.
Usage:
gist-kernel [options] -data <filename>Input:
-data <filename>- a labeled, tab-delimited file of data. Missing values are not allowed.Output: A square, labeled tab-delimited kernel matrix, with dimension N by N, where N is the number of rows in the data file.
Options:
-kernel <filename>- a file containing a list of kernel transformations, one per line. Valid operations are listed below.-iskernel- Specify that the data file contains a kernel, so do not perform the initial scalar product operation.-rdb- Allow the program to read and create RDB formatted files, which contain an additional format line after the first line of text.-precision <value>- Number of digits after the decimal place in the output file. The default value is 4.-verbose 1|2|3|4|5- Set the verbosity level of the output to stderr. The default level is 2.Kernel transformations:
The kernel transformations file specifies on transformation per line. The transformations are carried out on the kernel matrix in the order listed, and may be repeated. Some transformations include a required argument, which appears immediately after the transformation name.
constant <value>- Add a given constant to every element in the kernel matrix.coefficient <value>- Multiply every element in the kernel matrix by the given value.power <value>- Raise every kernel matrix element to the given power.radial <value>- Perform a radial basis transformation of the kernel using the given width.diagonal <value>- Add a given value to every diagonal element in the kernel.normalize- Project each vector onto the unit sphere in the feature space.center- Center each data point in feature space.diffusion <value>- Convert the given kernel to a matrix of Euclidean distances, and run that matrix through a diffusion kernel computation, with the given value as the diffusion constant.Warning messages:
None
Bugs:
- Does not yet implement the radial basis width selection heuristic, nor the asymmetric soft margin heuristic.
- Only creates square kernel matrices.