fixed typos in examples/graph &examples.keras_recipes. (#1776)

examples/graph/gat_node_classification.py

@@ -11,7 +11,7 @@
 ## Introduction
 
 [Graph neural networks](https://en.wikipedia.org/wiki/Graph_neural_network)
-is the prefered neural network architecture for processing data structured as
+is the preferred neural network architecture for processing data structured as
 graphs (for example, social networks or molecule structures), yielding
 better results than fully-connected networks or convolutional networks.
 
@@ -131,7 +131,7 @@
 aggregated information of *N*-hops (where *N* is decided by the number of layers of the
 GAT). Importantly, in contrast to the
 [graph convolutional network](https://arxiv.org/abs/1609.02907) (GCN)
-the GAT makes use of attention machanisms
+the GAT makes use of attention mechanisms
 to aggregate information from neighboring nodes (or *source nodes*). In other words, instead of simply
 averaging/summing node states from source nodes (*source papers*) to the target node (*target papers*),
 GAT first applies normalized attention scores to each source node state and then sums.

examples/graph/gnn_citations.py

@@ -521,7 +521,7 @@ def call(self, inputs):
 1. Apply preprocessing using FFN to the node features to generate initial node representations.
 2. Apply one or more graph convolutional layer, with skip connections,  to the node representation
 to produce node embeddings.
-3. Apply post-processing using FFN to the node embeddings to generat the final node embeddings.
+3. Apply post-processing using FFN to the node embeddings to generate  the final node embeddings.
 4. Feed the node embeddings in a Softmax layer to predict the node class.
 
 Each graph convolutional layer added captures information from a further level of neighbours.

examples/graph/ipynb/gat_node_classification.ipynb

@@ -23,7 +23,7 @@
     "## Introduction\n",
     "\n",
     "[Graph neural networks](https://en.wikipedia.org/wiki/Graph_neural_network)\n",
-    "is the prefered neural network architecture for processing data structured as\n",
+    "is the preferred neural network architecture for processing data structured as\n",
     "graphs (for example, social networks or molecule structures), yielding\n",
     "better results than fully-connected networks or convolutional networks.\n",
     "\n",
@@ -51,7 +51,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 0,
+   "execution_count": null,
    "metadata": {
     "colab_type": "code"
    },
@@ -89,7 +89,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 0,
+   "execution_count": null,
    "metadata": {
     "colab_type": "code"
    },
@@ -142,7 +142,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 0,
+   "execution_count": null,
    "metadata": {
     "colab_type": "code"
    },
@@ -167,7 +167,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 0,
+   "execution_count": null,
    "metadata": {
     "colab_type": "code"
    },
@@ -204,7 +204,7 @@
     "aggregated information of *N*-hops (where *N* is decided by the number of layers of the\n",
     "GAT). Importantly, in contrast to the\n",
     "[graph convolutional network](https://arxiv.org/abs/1609.02907) (GCN)\n",
-    "the GAT makes use of attention machanisms\n",
+    "the GAT makes use of attention mechanisms\n",
     "to aggregate information from neighboring nodes (or *source nodes*). In other words, instead of simply\n",
     "averaging/summing node states from source nodes (*source papers*) to the target node (*target papers*),\n",
     "GAT first applies normalized attention scores to each source node state and then sums."
@@ -239,7 +239,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 0,
+   "execution_count": null,
    "metadata": {
     "colab_type": "code"
    },
@@ -336,8 +336,7 @@
     "        else:\n",
     "            outputs = tf.reduce_mean(tf.stack(outputs, axis=-1), axis=-1)\n",
     "        # Activate and return node states\n",
-    "        return tf.nn.relu(outputs)\n",
-    ""
+    "        return tf.nn.relu(outputs)\n"
    ]
   },
   {
@@ -357,7 +356,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 0,
+   "execution_count": null,
    "metadata": {
     "colab_type": "code"
    },
@@ -425,8 +424,7 @@
     "        # Update metric(s)\n",
     "        self.compiled_metrics.update_state(labels, tf.gather(outputs, indices))\n",
     "\n",
-    "        return {m.name: m.result() for m in self.metrics}\n",
-    ""
+    "        return {m.name: m.result() for m in self.metrics}\n"
    ]
   },
   {
@@ -440,7 +438,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 0,
+   "execution_count": null,
    "metadata": {
     "colab_type": "code"
    },
@@ -499,7 +497,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 0,
+   "execution_count": null,
    "metadata": {
     "colab_type": "code"
    },
@@ -562,4 +560,4 @@
  },
  "nbformat": 4,
  "nbformat_minor": 0
-}
+}