exampleSetMode - sets or returns the example set's example selection mode


    exampleSetMode <example-set-list> [<mode>]


The example set mode determines how it chooses the next example during training. There are six modes: ORDERED, PERMUTED, RANDOMIZED, PROBABILISTIC, PIPE, and CUSTOM. When no mode is specified, the names of the sets in the list along with their current modes will be returned. Given an example selection mode, it will change the modes and reset the example sets (see resetExampleSets).

In ORDERED mode, which is the default, examples will be presented in the order in which they were found in the example file.

In RANDOMIZED mode, examples will be selected at random with replacement, each having the same probability of selection. Note that this differs from PERMUTED because it uses replacement. It differs from PROBABILISTIC because it ignores the example frequency.

In PERMUTED mode, examples will be selected at random without replacement, each having the same probability of selection. A different order will be computed for each pass through the set.

In PROBABILISTIC mode, examples are selected based on their given frequency. Specified frequency values will be normalized over all examples and this distribution used for selection. If example sets are concatenated, the distribution will be recalculated based on the specified frequencies. An example with no frequency specified is given a value of 1.0.

PIPE mode is used for example sets that are reading from a pipe. The next example will be read from the pipe and stored temporarily in the example set's pipeExample field. This mode can only be used with example sets for which a pipe was opened with "loadExamples ... -m PIPE". If the pipe is exhausted and the example set's pipeLoop flag is set to TRUE, which is the default, the pipe will be re-opened automatically. If an example set contains both stored examples and an open pipe, you can switch between them by changing from PIPE mode to another mode.

CUSTOM mode allows you to write a procedure that generates the index of the next example. When it's time to choose the next example, the example set's chooseExample procedure will be called. This should return an integer between 0 and one less than the number of examples, inclusive.


To change the mode of set "set1" to RANDOMIZED:

    lens> exampleSetMode set1 RAND


resetExampleSets, loadExamples, doExample

Last modified: Tue Apr 11 22:31:24 EDT 2000