Updates the NSL IMDB tutorial to use the new LSH support when building the graph.

PiperOrigin-RevId: 326516612

Author: Neural-Link Team

g3doc/tutorials/graph_keras_lstm_imdb.ipynb

@@ -514,11 +514,18 @@
         "graph will correspond to similarity between pairs of nodes.\n",
         "\n",
         "Neural Structured Learning provides a graph building library to build a graph\n",
-        "based on sample embeddings. It uses **cosine similarity** as the similarity\n",
-        "measure to compare embeddings and build edges between them. It also allows us to\n",
-        "specify a similarity threshold, which can be used to discard dissimilar edges\n",
-        "from the final graph. In this example, using 0.99 as the similarity threshold,\n",
-        "we end up with a graph that has 445,327 bi-directional edges."
+        "based on sample embeddings. It uses\n",
+        "[**cosine similarity**](https://en.wikipedia.org/wiki/Cosine_similarity) as the\n",
+        "similarity measure to compare embeddings and build edges between them. It also\n",
+        "allows us to specify a similarity threshold, which can be used to discard\n",
+        "dissimilar edges from the final graph. In this example, using 0.99 as the\n",
+        "similarity threshold and 12345 as the random seed, we end up with a graph that\n",
+        "has 429,415 bi-directional edges. Here we're using the graph builder's support\n",
+        "for [locality-sensitive hashing](https://en.wikipedia.org/wiki/Locality-sensitive_hashing)\n",
+        "(LSH) to speed up graph building. For details on using the graph builder's LSH\n",
+        "support, see the\n",
+        "[`build_graph_from_config`](https://www.tensorflow.org/neural_structured_learning/api_docs/python/nsl/tools/build_graph_from_config)\n",
+        "API documentation."
       ]
     },
     {
@@ -531,9 +538,35 @@
       },
       "outputs": [],
       "source": [
-        "nsl.tools.build_graph(['/tmp/imdb/embeddings.tfr'],\n",
-        "                      '/tmp/imdb/graph_99.tsv',\n",
-        "                      similarity_threshold=0.99)"
+        "graph_builder_config = nsl.configs.GraphBuilderConfig(\n",
+        "    similarity_threshold=0.99, lsh_splits=32, lsh_rounds=15, random_seed=12345)\n",
+        "nsl.tools.build_graph_from_config(['/tmp/imdb/embeddings.tfr'],\n",
+        "                                  '/tmp/imdb/graph_99.tsv',\n",
+        "                                  graph_builder_config)"
+      ]
+    },
+    {
+      "cell_type": "markdown",
+      "metadata": {
+        "colab_type": "text",
+        "id": "4dk9xfQcK553"
+      },
+      "source": [
+        "Each bi-directional edge is represented by two directed edges in the output TSV\n",
+        "file, so that file contains 429,415 * 2 = 858,830 total lines:"
+      ]
+    },
+    {
+      "cell_type": "code",
+      "execution_count": null,
+      "metadata": {
+        "colab": {},
+        "colab_type": "code",
+        "id": "dDPwTpZcJ3zF"
+      },
+      "outputs": [],
+      "source": [
+        "!wc -l /tmp/imdb/graph_99.tsv"
       ]
     },
     {
@@ -1532,6 +1565,10 @@
       "collapsed_sections": [
         "24gYiJcWNlpA"
       ],
+      "last_runtime": {
+        "build_target": "//learning/deepmind/public/tools/ml_python:ml_notebook",
+        "kind": "private"
+      },
       "name": "Graph regularization for sentiment classification using synthesized graphs",
       "private_outputs": true,
       "provenance": [],