classify

Description: Classify a set of unlabeled vectors using a trained support vector machine.

Usage: classify [options] -train <filename> -weights <filename> -test <filename> -label <filename>

Input:

• -train <filename> - an RDB file of training examples. The first column contains labels, and the remaining columns contain real-valued features.
• -weights <filename> - a two-column RDB file of weights, as output from compute-weights. The kernel parameters are read from the header of this file.
• -test <filename> - an RDB file of test examples to be classified.
• -label <name> - the label that appears in the first row, first column of the data file. (This parameter is admittedly useless and will be removed from future versions of the program.)

Note that, if the given weights file was trained on an externally defined kernel (using the '-matrix' option to compute-weights), then the file specified by the '-test' option should contain a kernel matrix, rather than test set examples. The kernel matrix is an n+1 (rows) by m+1 (columns) RDB matrix, where n is the number of test examples and m is the number of training examples. The first row and column contain data labels. The matrix entry for row x, column y, contains the kernel value K(x,y). If a kernel matrix is supplied in place of the test set, then no training set need be specified. However, if the kernel is normalized or is radial basis, then the '-selftrain' and '-selftest' files must be supplied (see below).

Output: A three-column RDB file containing data labels, classifications (-1 or 1) and discriminants for the test set.

Options:

• -selftrain <filename> - Read from the given file a series of n values of the form K(x,x), where K is the base kernel function and x is an element in the training set. This option is only necessary if the base kernel function is supplied from a file (using the '-matrix' option) and the kernel is normalized or radial basis. The input file should be in RDB format, with data labels in the first column and values in the second column.
• -selftest <filename> - Similar to '-selftrain', but for the test set.
• -noclasses - Compute and print the kernel matrix to stdout. Do not compute the classifications.
• -notime - Do not include timing information in the output header.
• -verbose 1|2|3|4|5 - Set the verbosity level of the output to stderr. The default level is 2.

Calls: cat, column, headchg, awk