Algorithm::SVM(3pm)

1Algorithm::SVM(3)     User Contributed Perl Documentation    Algorithm::SVM(3)
2
3
4

NAME

6       Algorithm::SVM - Perl bindings for the libsvm Support Vector Machine
7       library.
8

SYNOPSIS

10         use Algorithm::SVM;
11
12         # Load the model stored in the file 'sample.model'
13         $svm = new Algorithm::SVM(Model => 'sample.model');
14
15         # Classify a dataset.
16         $ds1 = new Algorithm::SVM::DataSet(Label => 1,
17                                            Data  => [0.12, 0.25, 0.33, 0.98]);
18         $res = $svm->predict($ds);
19
20         # Train a new SVM on some new datasets.
21         $svm->train(@tset);
22
23         # Change some of the SVM parameters.
24         $svm->gamma(64);
25         $svm->C(8);
26         # Retrain the SVM with the new parameters.
27         $svm->retrain();
28
29         # Perform cross validation on the training set.
30         $accuracy = $svm->validate(5);
31
32         # Save the model to a file.
33         $svm->save('new-sample.model');
34
35         # Load a saved model from a file.
36         $svm->load('new-sample.model');
37
38         # Retreive the number of classes.
39         $num = $svm->getNRClass();
40
41         # Retreive labels for dataset classes
42         (@labels) = $svm->getLabels();
43
44         # Probabilty for regression models, see below for details
45         $prob = $svm->getSVRProbability();
46

DESCRIPTION

48       Algorithm::SVM implements a Support Vector Machine for Perl.  Support
49       Vector Machines provide a method for creating classifcation functions
50       from a set of labeled training data, from which predictions can be made
51       for subsequent data sets.
52

CONSTRUCTOR

54         # Load an existing SVM.
55         $svm = new Algorithm::SVM(Model  => 'sample.model');
56
57         # Create a new SVM with the specified parameters.
58         $svm = new Algorithm::SVM(Type   => 'C-SVC',
59                                   Kernel => 'radial',
60                                   Gamma  => 64,
61                                   C      => 8);
62
63       An Algorithm::SVM object can be created in one of two ways - an
64       existing SVM can be loaded from a file, or a new SVM can be created an
65       trained on a dataset.
66
67       An existing SVM is loaded from a file using the Model named parameter.
68       The model file should be of the format produced by the svm-train
69       program (distributed with the libsvm library) or from the $svm->save()
70       method.
71
72       New SVM's can be created using the following parameters:
73
74         Type    - The type of SVM that should be created.  Possible values are:
75                   'C-SVC', 'nu-SVC', 'one-class', 'epsilon-SVR' and 'nu-SVR'.
76                   Default os 'C-SVC'.
77
78         Kernel  - The type of kernel to be used in the SVM.  Possible values
79                   are: 'linear', 'polynomial', 'radial' and 'sigmoid'.
80                   Default is 'radial'.
81
82         Degree  - Sets the degree in the kernel function.  Default is 3.
83
84         Gamma   - Sets the gamme in the kernel function.  Default is 1/k,
85                   where k is the number of training sets.
86
87         Coef0   - Sets the Coef0 in the kernel function.  Default is 0.
88
89         Nu      - Sets the nu parameter for nu-SVC SVM's, one-class SVM's
90                   and nu-SVR SVM's.  Default is 0.5.
91
92         Epsilon - Sets the epsilon in the loss function of epsilon-SVR's.
93                   Default is 0.1.
94
95       For a more detailed explanation of what the above parameters actually
96       do, refer to the documentation distributed with libsvm.
97

METHODS

99         $svm->degree($degree);
100         $svm->gamma($gamma);
101         $svm->coef0($coef0);
102         $svm->C($C);
103         $svm->nu($nu);
104         $svm->epsilon($epsilon);
105         $svm->kernel_type($ktype);
106         $svm->svm_type($svmtype);
107
108         $svm->retrain();
109
110       The Algorithm::SVM object provides accessor methods for the various SVM
111       parameters.  When a value is provided to the method, the object will
112       attempt to set the corresponding SVM parameter.  If no value is
113       provided, the current value will be returned.  See the constructor
114       documentation for a description of appropriate values.
115
116       The retrain method should be called if any of the parameters are
117       modified from their initial values so as to rebuild the model with the
118       new values.  Note that you can only retrain an SVM if you've previously
119       trained the SVM on a dataset.  (ie. You can't currently retrain a model
120       loaded with the load method.)  The method will return a true value if
121       the retraining was successful and a false value otherwise.
122
123         $res = $svm->predict($ds);
124
125       The predict method is used to classify a set of data according to the
126       loaded model.  The method accepts a single parameter, which should be
127       an Algorithm::SVM::DataSet object.  Returns a floating point number
128       corresponding to the predicted value.
129
130         $res = $svm->predict_value($ds);
131
132       The predict_value method works similar to predict, but returns a
133       floating point value corresponding to the output of the trained SVM.
134       For a linear kernel, this can be used to reconstruct the weights for
135       each attribute as follows: the bias of the linear function is returned
136       when calling predict_value on an empty dataset (all zeros), and by
137       setting each variable in turn to one and all others to zero, you get
138       one value per attribute which corresponds to bias + weight_i. By
139       subtracting the bias, the final linear model is obtained as sum of
140       (weight_i * attr_i) plus bias. The sign of this value corresponds to
141       the binary prediction.
142
143         $svm->save($filename);
144
145       Saves the currently loaded model to the specified filename.  Returns a
146       false value on failure, and truth value on success.
147
148         $svm->load($filename);
149
150       Loads a model from the specified filename.  Returns a false value on
151       failure, and truth value on success.
152
153         $svm->train(@tset);
154
155       Trains the SVM on a set of Algorithm::SVM::DataSet objects.  @tset
156       should be an array of Algorithm::SVM::DataSet objects.
157
158         $accuracy = $svm->validate(5);
159
160       Performs cross validation on the training set.  If an argument is
161       provided, the set is partioned into n subsets, and validated against
162       one another.  Returns a floating point number representing the accuracy
163       of the validation.
164
165         $num = $svm->getNRClass();
166
167       For a classification model, this function gives the number of classes.
168       For a regression or a one-class model, 2 is returned.
169
170         (@labels) = $svm->getLabels();
171
172       For a classification model, this function returns the name of the
173       labels in an array.  For regression and one-class models undef is
174       returned.
175
176         $prob = $svm->getSVRProbability();
177
178       For a regression model with probability information, this function
179       outputs a value sigma > 0.  For test data, we consider the probability
180       model: target value = predicted value + z, z: Laplace distribution
181       e^(-|z|/sigma)/2sigma)
182
183       If the model is not for svr or does not contain required information,
184       undef is returned.
185

MAINTAINER

187       Matthew Laird <matt@brinkman.mbb.sfu.ca> Alexander K. Seewald
188       <alex@seewald.at>
189

ACKNOWLEDGEMENTS

195       Thanks go out to Fiona Brinkman and the other members of the Simon
196       Fraser University Brinkman Laboratory for providing me the opportunity
197       to develop this module.  Additional thanks go to Chih-Jen Lin, one of
198       the libsvm authors, for being particularly helpful during the
199       development process.
200
201       As well to Dr. Alexander K. Seewald of Seewald Solutions for many bug
202       fixes, new test cases, and lowering the memory footprint by a factor of
203       20.  Thank you very much!
204
205
206
207perl v5.30.0                      2019-07-26                 Algorithm::SVM(3)