new branch for transfer multilabel of decision tree to cpp

examples/estimator/classifier/DecisionTreeClassifier/c/basics.pct.multilabel.py

@@ -0,0 +1,75 @@
+# %% [markdown]
+# # sklearn-porter
+#
+# Repository: [https://github.com/nok/sklearn-porter](https://github.com/nok/sklearn-porter)
+#
+# ## DecisionTreeClassifier
+#
+# Documentation: [sklearn.tree.DecisionTreeClassifier](http://scikit-learn.org/stable/modules/generated/sklearn.tree.DecisionTreeClassifier.html)
+
+# %%
+import sys
+sys.path.append('../../../../..')
+
+# %% [markdown]
+# ### Load data
+
+# %%
+from sklearn.datasets import load_iris
+
+iris_data = load_iris()
+
+X = iris_data.data
+y = iris_data.target
+
+print(X.shape, y.shape)
+
+# %% [markdown]
+# ### Train classifier
+
+# %%
+from sklearn.tree import tree
+import numpy as np
+
+# transfer single-output into multi-labels
+y_multi_label = []
+for x in iris_data.target:
+    if x == 0:
+        y_multi_label.append([1,1,0])
+    elif x == 1:
+        y_multi_label.append([0,1,1])
+    else:
+        y_multi_label.append([1,0,1])
+y = np.array(y_multi_label)
+
+from sklearn.model_selection import train_test_split
+X_train, X_test, y_train, y_test = train_test_split(
+    iris_data.data, y, test_size=0.33, random_state=42)
+
+clf = tree.DecisionTreeClassifier(random_state=0)
+clf.fit(X_train, y_train)
+
+# %% [markdown]
+# ### Transpile classifier
+
+# %%
+from sklearn_porter import Porter
+
+porter = Porter(clf, language='c')
+output = porter.export()
+
+print(output)
+
+# %% [markdown]
+# ### Run classification in C
+
+# %%
+# Save model:
+# with open('tree.c', 'w') as f:
+#     f.write(output)
+
+# Compile model:
+# $ gcc tree.c -std=c99 -lm -o tree
+
+# Run classification:
+# $ ./tree 1 2 3 4

sklearn_porter/estimator/classifier/DecisionTreeClassifier/__init__.py

@@ -188,6 +188,19 @@ def export(self, class_name, method_name, export_data=False,
         classes = ', '.join([temp_arr_scope.format(v) for v in classes])
         classes = temp_arr__.format(type='int', name='classes', values=classes,
                                     n=n, m=m)
+
+        # transfer dt to c language and is multilabel task
+        if self.target_language in ['c'] and self.estimator.tree_.value.ndim == 3:
+            import numpy as np
+            classes = np.argmax(self.estimator.tree_.value, axis=2).tolist()
+            n = len(classes)
+            m = len(classes[0])
+            classes = [', '.join([str(int(x)) for x in e]) for e in classes]
+            classes = ', '.join([temp_arr_scope.format(v) for v in classes])
+            classes = temp_arr__.format(type='int', name='classes', values=classes,
+                                    n=n, m=m)
+            self.n_outputs_ = self.estimator.n_outputs_
+
         self.classes = classes
 
         if self.target_method == 'predict':
@@ -254,6 +267,9 @@ def predict(self, temp_type='separated'):
 
         if temp_type == 'separated':
             separated_temp = self.temp('separated.class')
+            # transfer dt to c language and is multilabel task
+            if self.target_language in ['c'] and self.estimator.tree_.value.ndim == 3:
+                separated_temp = self.temp('separated.multilabels.class')
             return separated_temp.format(**self.__dict__)
 
         if temp_type == 'embedded':

sklearn_porter/estimator/classifier/DecisionTreeClassifier/templates/c/separated.multilabels.class.txt

@@ -0,0 +1,41 @@
+#include <stdlib.h>
+#include <stdio.h>
+#include <math.h>
+
+#define N_FEATURES {n_features}
+#define N_OUTPUTS {n_outputs_}
+
+{left_childs}
+{right_childs}
+{thresholds}
+{indices}
+{classes}
+
+
+int* {method_name}(double features[N_FEATURES]) {{
+    int node = 0; //root node id is 0
+    while (thresholds[node] != -2) {{
+        if (features[indices[node]] <= thresholds[node]) {{
+            node = lChilds[node];
+        }} else {{
+            node = rChilds[node];
+        }}
+    }}
+    return classes[node];
+}}
+
+int main(int argc, const char * argv[]) {{
+
+    /* Features: */
+    double features[argc-1];
+    int i;
+    for (i = 1; i < argc; i++) {{
+        features[i-1] = atof(argv[i]);
+    }}
+
+    /* Prediction: */
+    int* output = {method_name}(features);
+    for(int i=0;i<N_OUTPUTS;i++)
+        printf("%d\n", output[i]);
+    return 0;
+}}