++ed by:
Lei Sun (孙 磊)

# NAME

AI::MaxEntropy::Util - Utilities for doing experiments with ME learners

# SYNOPSIS

``````  use AI::MaxEntropy;
use AI::MaxEntropy::Util qw/:all/;

my \$me = AI::MaxEntropy->new;

my \$samples = [
[['a', 'b', 'c'] => 'x'],
[['e', 'f'] => 'y' => 1.5],
...
];

my (\$result, \$model) = train_and_test(\$me, \$samples, 'xxxo');

print precision(\$result)."\n";
print recall(\$result, 'x')."\n";``````

# DESCRIPTION

This module makes doing experiments with Maximum Entropy learner easier.

Generally, an experiment involves a training set and a testing set (sometimes also a parameter adjusting set). The learner is trained with samples in the training set and tested with samples in the testing set. Usually, 2 measures of performance are concerned. One is precision, indicating the percentage of samples which are correctly predicted in the testing set. The other one is recall, indicating the precision of samples with a certain label.

# FUNCTIONS

## train_and_test

This function automated the process of training and testing.

``````    my \$me = AI::MaxEntropy->new;
my \$sample = [
[ ['a', 'b'] => 'x' => 1.5 ],
...
];
my (\$result, \$model) = train_and_test(\$me, \$sample, 'xxxo');``````

First, the whole samples set will be divided into a training set and a testing set according to the specified pattern. A pattern is a string, in which each character stands for a part of the samples set. If the character is `'x'`, the corresponding part is used for training. If the character is `'o'`, the corresponding part is used for testing. Otherwise, the corresponding part is simply ignored.

For example, the pattern 'xxxo' means the first three forth of the samples set are used for training while the last one forth is used for testing.

The function returns two values. The first one is an array ref describe the result of testing, in which each element follows a structure like `[sample => result]`. The second one is the model learnt from the training set, which is an AI::MaxEntropy::Model object.

## traverse_partially

This function is the core implementation of "train_and_test". It traverse through some of the elements in an array according to a pattern, and does some specified actions with each of these elements.

``````  my \$arr = [1, 2, 3, 4, 5];

# print out the first two firth of the array
traverse_partially { print } \$arr, 'xx---';

# do the same thing, using custom significant character 'o'
traverse_partially { print } \$arr, 'oo---' => 'o';

my \$samples = [
[['a', 'b'] => 'x'],
[['c', 'd'] => 'y' => 1.5],
...
];
my \$me = AI::MaxEntropy->new;

# see the first one third and the last one third samples
traverse_partially { \$me->see(@\$_) } \$samples, 'x-x';``````

## map_partially

This function is similar to "traverse_partially". However, it returns an array ref in which all elements in the original array is mapped according to the code snippet's return value.

``````  my \$arr = [1, 2, 3, 4, 5];

# increase the last one third of the elements by 1
\$arr = map_partially { \$_ + 1 } \$arr, '--x';``````

## precision

Calculates the precision based on the result returned by "train_and_test".

``````  ...
my (\$result, \$model) = train_and_test(...);
print precision(\$result)."\n";``````

Note that the weights of samples are taken into consideration.

## recall

Calculates the recall of a certain label based on the result returned by "train_and_test".

``````  ...
my (\$result, \$model) = train_and_test(...);
print recall(\$result, 'label')."\n";``````

Note that the weights of samples are taken into consideration.

# AUTHOR

Laye Suen, <laye@cpan.org>