add attribute support_ #138

Author: QinbinLi

include/thundersvm/model/svmmodel.h

@@ -111,6 +111,9 @@ class SvmModel {
 
     //return sv_max_index
     int get_sv_max_index() const;
+
+    //return sv_indices
+    const vector<int> &get_sv_ind() const;
 protected:
 
     /**
@@ -146,6 +149,8 @@ class SvmModel {
      */
 
     DataSet::node2d sv;
+    ///the indices of support vectors
+    vector<int> sv_indices;
     ///the number of support vectors for each class
     SyncArray<int> n_sv;
 

python/README.md

@@ -100,6 +100,9 @@ The usage of thundersvm scikit interface is similar to sklearn.svm.
     The seed of the pseudo random number generator to use when shuffling the data. If int, random_state is the seed used by the random number generator; If RandomState instance, random_state is the random number generator; If None, the random number generator is the RandomState instance used by np.random.
 
 ### Attributes
+*support_*: array-like, shape = [n_SV]\
+    indices of support vectors.
+
 *support_vectors_*: array-like, shape = [n_SV, n_features]\
     support vectors.
 

python/thundersvm/thundersvm.py

@@ -129,14 +129,16 @@ def fit(self, X, y):
         csr_col = (c_int * (self.n_sv * self.n_features))()
         csr_data = (c_float * (self.n_sv * self.n_features))()
         data_size = (c_int * 1)()
-        thundersvm.get_sv(csr_row, csr_col, csr_data, data_size, c_void_p(self.model))
+        sv_indices = (c_int * self.n_sv)()
+        thundersvm.get_sv(csr_row, csr_col, csr_data, data_size, sv_indices, c_void_p(self.model))
         self.row  = np.array([csr_row[index] for index in range(0, self.n_sv + 1)])
         self.col  = np.array([csr_col[index] for index in range(0, data_size[0])])
         self.data = np.array([csr_data[index] for index in range(0, data_size[0])])
 
         self.support_vectors_ = sp.csr_matrix((self.data, self.col, self.row))
         if self._sparse == False:
             self.support_vectors_ = self.support_vectors_.toarray(order = 'C')
+        self.support_ = np.array([sv_indices[index] for index in range(0, self.n_sv)]).astype(int)
 
         dual_coef = (c_float * ((self.n_classes - 1) * self.n_sv))()
         thundersvm.get_coef(dual_coef, self.n_classes, self.n_sv, c_void_p(self.model))

src/thundersvm/model/oneclass_svc.cpp

@@ -40,6 +40,7 @@ void OneClassSVC::train(const DataSet &dataset, SvmParam param) {
     for (int i = 0; i < n_instances; ++i) {
         if (alpha_data[i] != 0) {
             sv.push_back(dataset.instances()[i]);
+            sv_indices.push_back(i);
             coef_vec.push_back(alpha_data[i]);
         }
     }

src/thundersvm/model/svc.cpp

@@ -68,6 +68,7 @@ void SVC::train(const DataSet &dataset, SvmParam param) {
             if (is_sv[original_index[j]]) {
                 n_sv_data[i]++;
                 sv.push_back(i_instances[j]);
+                sv_indices.push_back(original_index[j]);
             }
         }
     }

src/thundersvm/model/svmmodel.cpp

@@ -324,6 +324,10 @@ int SvmModel::get_n_classes() const {
     return n_classes;
 }
 
+const vector<int> &SvmModel::get_sv_ind() const {
+    return sv_indices;
+}
+
 void SvmModel::set_max_iter(int iter) {
     max_iter = iter;
 }

src/thundersvm/model/svr.cpp

@@ -49,6 +49,7 @@ void SVR::save_svr_coef(const SyncArray<float_type> &alpha_2, const DataSet::nod
         float_type alpha_i = alpha_2_data[i] - alpha_2_data[i + n_instances];
         if (alpha_i != 0) {
             sv.push_back(instances[i]);
+            sv_indices.push_back(i);
             coef_vec.push_back(alpha_i);
         }
     }

src/thundersvm/thundersvm-scikit.cpp

@@ -260,7 +260,7 @@ extern "C" {
         return;
     }
 
-    void get_sv(int* row, int* col, float* data, int* data_size, SvmModel* model){
+    void get_sv(int* row, int* col, float* data, int* data_size, int* sv_indices, SvmModel* model){
         DataSet::node2d svs = model->svs();
         row[0] = 0;
         int data_ind = 0;
@@ -277,6 +277,11 @@ extern "C" {
             }
         }
         data_size[0] = data_ind;
+
+        vector<int> sv_index = model->get_sv_ind();
+        for(int i = 0; i < sv_index.size(); i++){
+            sv_indices[i] = sv_index[i];
+        }
         return ;
     }