Shark machine learning library
About Shark
News!
Contribute
Credits and copyright
Downloads
Getting Started
Installation
Using the docs
Documentation
Tutorials
Quick references
Class list
Global functions
FAQ
Showroom
include
shark
Algorithms
Trainers
AbstractWeightedTrainer.h
Go to the documentation of this file.
1
//===========================================================================
2
/*!
3
*
4
*
5
* \brief Abstract Trainer Interface for trainers that support weighting
6
*
7
*
8
*
9
* \author O. Krause
10
* \date 2014
11
*
12
*
13
* \par Copyright 1995-2017 Shark Development Team
14
*
15
* <BR><HR>
16
* This file is part of Shark.
17
* <http://shark-ml.org/>
18
*
19
* Shark is free software: you can redistribute it and/or modify
20
* it under the terms of the GNU Lesser General Public License as published
21
* by the Free Software Foundation, either version 3 of the License, or
22
* (at your option) any later version.
23
*
24
* Shark is distributed in the hope that it will be useful,
25
* but WITHOUT ANY WARRANTY; without even the implied warranty of
26
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
27
* GNU Lesser General Public License for more details.
28
*
29
* You should have received a copy of the GNU Lesser General Public License
30
* along with Shark. If not, see <http://www.gnu.org/licenses/>.
31
*
32
*/
33
//===========================================================================
34
#ifndef SHARK_ALGORITHMS_TRAINERS_ABSTRACT_WEIGHTED_TRAINER_H
35
#define SHARK_ALGORITHMS_TRAINERS_ABSTRACT_WEIGHTED_TRAINER_H
36
37
#include <
shark/Data/WeightedDataset.h
>
38
#include <
shark/Algorithms/Trainers/AbstractTrainer.h
>
39
40
namespace
shark
{
41
42
43
/// \brief Superclass of weighted supervised learning algorithms
44
///
45
/// \par
46
/// AbstractWeightedTrainer is the super class of all trainers
47
/// that support weighted datasets. Weights are interpreted here
48
/// as the importance of a sample. unweighted training assumes
49
/// that all samples have the same importance, or weight.
50
/// The higher the weight, the more important a point. Weight
51
/// 0 is the same as if the point would not be part of the dataset.
52
/// Negative weights are not allowed.
53
///
54
/// When all weights are integral values there is a simple interpretation
55
/// of the weights as the multiplicity of a point. Thus training
56
/// with a dataset with duplicate points is the same as counting the duplicates
57
/// and run the algorithm with a weighted dataset where all points are unique and
58
/// have their weight is the multiplicity.
59
template
<
class
Model,
class
LabelTypeT =
typename
Model::OutputType>
60
class
AbstractWeightedTrainer
:
public
AbstractTrainer
<Model,LabelTypeT>
61
{
62
private
:
63
typedef
AbstractTrainer<Model,LabelTypeT>
base_type
;
64
public
:
65
typedef
typename
base_type::ModelType
ModelType
;
66
typedef
typename
base_type::InputType
InputType
;
67
typedef
typename
base_type::LabelType
LabelType
;
68
typedef
typename
base_type::DatasetType
DatasetType
;
69
typedef
WeightedLabeledData<InputType, LabelType>
WeightedDatasetType
;
70
71
/// \brief Executes the algorithm and trains a model on the given weighted data.
72
virtual
void
train
(ModelType& model, WeightedDatasetType
const
& dataset) = 0;
73
74
/// \brief Executes the algorithm and trains a model on the given unweighted data.
75
///
76
/// This method behaves as using train with a weighted dataset where all weights are equal.
77
/// The default implementation just creates such a dataset and executes the weighted
78
/// version of the algorithm.
79
virtual
void
train
(ModelType& model, DatasetType
const
& dataset){
80
train
(model,
WeightedDatasetType
(dataset, 1.0));
81
}
82
};
83
84
85
/// \brief Superclass of weighted unsupervised learning algorithms
86
///
87
/// \par
88
/// AbstractWeightedUnsupervisedTrainer is the super class of all trainers
89
/// that support weighted datasets. See AbstractWeightedTrainer for more information on
90
/// the weights.
91
/// \see AbstractWeightedTrainer
92
template
<
class
Model>
93
class
AbstractWeightedUnsupervisedTrainer
:
public
AbstractUnsupervisedTrainer
<Model>
94
{
95
private
:
96
typedef
AbstractUnsupervisedTrainer<Model>
base_type
;
97
public
:
98
typedef
typename
base_type::ModelType
ModelType
;
99
typedef
typename
base_type::InputType
InputType
;
100
typedef
typename
base_type::DatasetType
DatasetType
;
101
typedef
WeightedUnlabeledData<InputType>
WeightedDatasetType
;
102
103
/// \brief Excecutes the algorithm and trains a model on the given weighted data.
104
virtual
void
train
(ModelType& model, WeightedDatasetType
const
& dataset) = 0;
105
106
/// \brief Excecutes the algorithm and trains a model on the given undata.
107
///
108
/// This method behaves as using train with a weighted dataset where all weights are equal.
109
/// The default implementation just creates such a dataset and executes the weighted
110
/// version of the algorithm.
111
virtual
void
train
(ModelType& model, DatasetType
const
& dataset){
112
train
(model,
WeightedDatasetType
(dataset, 1.0));
113
}
114
};
115
116
117
}
118
#endif