update

Author: zhedongzheng

classic-models/bagging_clf.py

@@ -1,5 +1,5 @@
 import numpy as np
-import tensorflow as tf
+from utils import one_hot
 from sklearn.tree import DecisionTreeClassifier
 
 
@@ -19,7 +19,7 @@ def fit(self, X, y):
 
     def predict(self, X):
         ys = [model.predict(X) for model in self.models]
-        ys_one_hot = [tf.contrib.keras.utils.to_categorical(y) for y in ys]
+        ys_one_hot = [one_hot(y) for y in ys]
         return np.argmax(np.sum(ys_one_hot, axis=0), axis=1)
 
 

classic-models/random_forest_clf.py

@@ -1,7 +1,7 @@
 from sklearn.tree import DecisionTreeClassifier
 import math
 import numpy as np
-import tensorflow as tf
+from utils import one_hot
 
 
 class RandomForestClassifier:
@@ -23,7 +23,7 @@ def fit(self, X, y):
 
     def predict(self, X):
         ys = [tree.predict(X[:, self._features[i]]) for i, tree in enumerate(self._forest)]
-        ys_one_hot = [tf.contrib.keras.utils.to_categorical(y) for y in ys]
+        ys_one_hot = [one_hot(y) for y in ys]
         return np.argmax(np.sum(ys_one_hot, axis=0), axis=1)
 
 

classic-models/utils.py

@@ -33,3 +33,13 @@ def plot_decision_boundary(X, model):
     # Put the result into a color plot
     Z = Z.reshape(xx.shape)
     plt.contour(xx, yy, Z, cmap=plt.cm.Paired)
+
+
+def one_hot(y, num_classes=None):
+    y = np.array(y, dtype='int').ravel()
+    if not num_classes:
+        num_classes = np.max(y) + 1
+    n = y.shape[0]
+    categorical = np.zeros((n, num_classes))
+    categorical[np.arange(n), y] = 1
+    return categorical

nlp-models/tensorflow/word2vec_skipgram.py

@@ -144,7 +144,7 @@ def fit(self, n_epoch=10, batch_size=1000, top_k=5, eval_step=1000):
                 y = np.array(y)[:, np.newaxis]
                 _, loss = self.sess.run([self.train_op, self.loss], {self.x: x, self.y: y})
                 if local_step % 50 == 0:
-                    print ('Epoch %d/%d | Batch %d/%d | train loss: %.4f |' %
+                    print ('Epoch %d/%d | Batch %d/%d | train loss: %.4f' %
                            (epoch+1, n_epoch, local_step, n_batch, loss))
                 global_step += 1
                 if local_step % eval_step == 0: