1
  2
  3
  4
  5
  6
  7
  8
  9
 10
 11
 12
 13
 14
 15
 16
 17
 18
 19
 20
 21
 22
 23
 24
 25
 26
 27
 28
 29
 30
 31
 32
 33
 34
 35
 36
 37
 38
 39
 40
 41
 42
 43
 44
 45
 46
 47
 48
 49
 50
 51
 52
 53
 54
 55
 56
 57
 58
 59
 60
 61
 62
 63
 64
 65
 66
 67
 68
 69
 70
 71
 72
 73
 74
 75
 76
 77
 78
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176

//! The hyperparameter initializers.

use crate::{
    consts::{DEFAULT_COEF0, DEFAULT_COST, DEFAULT_DEGREE, DEFAULT_MODEL_EPSILON, DEFAULT_NU},
    error::Error,
    model::{KernelParams, ModelParams, Svm, SvmParams},
    state::Untrained,
};
use std::{
    collections::HashMap,
    convert::{TryFrom, TryInto},
};

/// The model type initializer.
#[derive(Clone, Debug)]
pub enum ModelInit {
    CSvc { cost: Option<f64> },
    NuSvc { nu: Option<f64> },
    NuSvr { nu: Option<f64> },
    OneClass { nu: Option<f64> },
    EpsilonSvr { epsilon: Option<f64> },
}

/// The kernel initializer.
#[derive(Clone, Debug)]
pub enum KernelInit {
    Linear,
    Polynomial {
        gamma: Option<f64>,
        coef0: Option<f64>,
        degree: Option<usize>,
    },
    Rbf {
        gamma: Option<f64>,
    },
    Sigmoid {
        gamma: Option<f64>,
        coef0: Option<f64>,
    },
    // Precomputed,
}

/// The SVM model initializer.
#[derive(Debug, Clone)]
pub struct SvmInit {
    pub model: Option<ModelInit>,
    pub kernel: Option<KernelInit>,
    pub cache_size: Option<usize>,
    pub probability_estimates: Option<bool>,
    pub shrinking: Option<bool>,
    pub termination_eps: Option<f64>,
    pub label_weights: Option<HashMap<usize, f64>>,
}

impl Default for SvmInit {
    fn default() -> Self {
        Self {
            model: None,
            kernel: None,
            cache_size: None,
            probability_estimates: None,
            shrinking: None,
            termination_eps: None,
            label_weights: None,
        }
    }
}

impl SvmInit {
    /// Builds SVM model from the initializer.
    pub fn build(&self) -> Result<Svm<Untrained>, Error> {
        let SvmParams {
            model,
            kernel,
            cache_size,
            probability_estimates,
            shrinking,
            termination_eps,
            label_weights,
        } = TryFrom::try_from(self)?;

        let (svm_type, cost, nu, model_epsilon) = match model {
            ModelParams::CSvc { cost } => {
                (libsvm_sys::C_SVC, cost, DEFAULT_NU, DEFAULT_MODEL_EPSILON)
            }
            ModelParams::NuSvc { nu } => {
                (libsvm_sys::NU_SVC, DEFAULT_COST, nu, DEFAULT_MODEL_EPSILON)
            }
            ModelParams::OneClass { nu } => (
                libsvm_sys::ONE_CLASS,
                DEFAULT_COST,
                nu,
                DEFAULT_MODEL_EPSILON,
            ),
            ModelParams::EpsilonSvr { epsilon } => {
                (libsvm_sys::EPSILON_SVR, DEFAULT_COST, DEFAULT_NU, epsilon)
            }
            ModelParams::NuSvr { nu } => {
                (libsvm_sys::NU_SVR, DEFAULT_COST, nu, DEFAULT_MODEL_EPSILON)
            }
        };

        let (kernel_type, gamma_opt, coef0, degree) = match kernel {
            KernelParams::Linear => (libsvm_sys::LINEAR, None, DEFAULT_COEF0, DEFAULT_DEGREE),
            KernelParams::Polynomial {
                gamma,
                coef0,
                degree,
            } => (libsvm_sys::POLY, gamma, coef0, degree),
            KernelParams::Rbf { gamma } => (libsvm_sys::RBF, gamma, DEFAULT_COEF0, DEFAULT_DEGREE),
            KernelParams::Sigmoid { gamma, coef0 } => {
                (libsvm_sys::SIGMOID, gamma, coef0, DEFAULT_DEGREE)
            }
        };

        let (nr_weight, weight_labels, weights) = {
            let nr_weight: i32 =
                label_weights
                    .len()
                    .try_into()
                    .map_err(|err| Error::InvalidHyperparameter {
                        reason: format!("invalid number of weights: {}", err),
                    })?;

            let (labels, weights) = label_weights.into_iter().fold(Ok((vec![], vec![])), |result, (index, weight)| {
                let (mut labels, mut weights) = result?;

                if weight < 0.0 || weight > 1.0 {
                    return Err(Error::InvalidHyperparameter {
                        reason: format!("the label weights in label_weights must be in range of [0, 1], but found {}", weight)
                    });
                }

                labels.push(index as i32);
                weights.push(weight);

                Ok((labels, weights))
            })?;

            (nr_weight, labels, weights)
        };

        let params = libsvm_sys::svm_parameter {
            svm_type: svm_type as i32,
            kernel_type: kernel_type as i32,
            degree: degree
                .try_into()
                .map_err(|err| Error::InvalidHyperparameter {
                    reason: format!("invalid degree parameter: {}", err),
                })?,
            gamma: gamma_opt.unwrap_or(0.0),
            coef0,
            cache_size: cache_size as f64,
            eps: termination_eps,
            C: cost,
            nu,
            p: model_epsilon,
            shrinking: shrinking as i32,
            probability: probability_estimates as i32,
            nr_weight,
            weight_label: std::ptr::null_mut(),
            weight: std::ptr::null_mut(),
        };

        let svm = Svm {
            state: Untrained {
                gamma_opt,
                params,
                weight_labels,
                weights,
            },
        };

        Ok(svm)
    }
}