[DO NOT MERGE] Update for MXNet 1.6.0 (#548)

* [DO NOT MERGE] Update for MXNet 1.6.0

* Update house-price-kaggle

* Update random-variables.md

* Update dropout.md

* Update linear-algebra.md

* Update linear-algebra.md

Author: haojin2

chapter_appendix_math/integral-calculus.md

@@ -63,7 +63,7 @@ approx = np.sum(epsilon*f)
 true = np.log(2) / 2
 
 d2l.set_figsize()
-d2l.plt.bar(x, f, width = epsilon, align = 'edge')
+d2l.plt.bar(x.asnumpy(), f.asnumpy(), width = epsilon, align = 'edge')
 d2l.plt.plot(x, f, color='black')
 d2l.plt.ylim([0, 1])
 d2l.plt.show()

chapter_appendix_math/linear-algebra.md

@@ -312,8 +312,8 @@ In a fully machine learned solution, we would learn the threshold from the datas
 ```{.python .input}
 # Print test set accuracy with eyeballed threshold
 w = (ave_1 - ave_0).T
-predictions = 1*(X_test.reshape(2000, -1).dot(w.flatten()) > -1500000)
-np.mean(predictions==y_test)  # Accuracy
+predictions = X_test.reshape(2000, -1).dot(w.flatten()) > -1500000
+np.mean(predictions.astype(y_test.dtype)==y_test, dtype=np.float64)  # Accuracy
 ```
 
 ## Geometry of Linear Transformations

chapter_appendix_math/random-variables.md

@@ -499,9 +499,9 @@ d2l.plt.figure(figsize=(12, 3))
 for i in range(3) :
     X = np.random.normal(0, 1, 500)
     Y = covs[i]*X + np.random.normal(0, 1, 500)
-    
+
     d2l.plt.subplot(1, 4, i+1)
-    d2l.plt.scatter(X, Y)
+    d2l.plt.scatter(X.asnumpy(), Y.asnumpy())
     d2l.plt.xlabel('X')
     d2l.plt.ylabel('Y')
     d2l.plt.title("cov = {}".format(covs[i]))
@@ -572,9 +572,9 @@ d2l.plt.figure(figsize=(12, 3))
 for i in range(3) :
     X = np.random.normal(0, 1, 500)
     Y = cors[i] * X + np.sqrt(1 - cors[i]**2) * np.random.normal(0, 1, 500)
-    
+
     d2l.plt.subplot(1, 4, i + 1)
-    d2l.plt.scatter(X, Y)
+    d2l.plt.scatter(X.asnumpy(), Y.asnumpy())
     d2l.plt.xlabel('X')
     d2l.plt.ylabel('Y')
     d2l.plt.title("cor = {}".format(cors[i]))

chapter_multilayer-perceptrons/dropout.md

@@ -210,7 +210,7 @@ def dropout(X, drop_prob):
     if drop_prob == 1:
         return np.zeros_like(X)
     mask = np.random.uniform(0, 1, X.shape) > drop_prob
-    return mask * X / (1.0-drop_prob)
+    return mask.astype(np.float32) * X / (1.0-drop_prob)
 ```
 
 We can test out the `dropout` function on a few examples.

chapter_multilayer-perceptrons/kaggle-house-price.md

@@ -180,9 +180,9 @@ Finally, via the `values` attribute,
 
 ```{.python .input  n=9}
 n_train = train_data.shape[0]
-train_features = np.array(all_features[:n_train].values)
-test_features = np.array(all_features[n_train:].values)
-train_labels = np.array(train_data.SalePrice.values).reshape(-1, 1)
+train_features = np.array(all_features[:n_train].values, dtype=np.float32)
+test_features = np.array(all_features[n_train:].values, dtype=np.float32)
+train_labels = np.array(train_data.SalePrice.values, dtype=np.float32).reshape(-1, 1)
 ```
 
 ## Training