Remove public_partitions from AggregateParams (#270)

OpenMined · May 10, 2022 · 3837118 · 3837118
1 parent bf5c1c4
commit 3837118
Show file tree

Hide file tree

Showing 11 changed files with 274 additions and 293 deletions.
diff --git a/examples/codelab/codelab_PipelineDP.ipynb b/examples/codelab/codelab_PipelineDP.ipynb
@@ -662,15 +662,14 @@
         "      noise_kind=pipeline_dp.NoiseKind.LAPLACE,\n",
         "      metrics=[pipeline_dp.Metrics.COUNT],\n",
         "      max_partitions_contributed=1,\n",
-        "      max_contributions_per_partition=1,\n",
-        "      public_partitions=public_partitions)\n",
+        "      max_contributions_per_partition=1)\n",
         "\n",
         "  data_extractors = pipeline_dp.DataExtractors(\n",
         "      privacy_id_extractor=lambda row: row.user_id,\n",
         "      partition_extractor=lambda row: row.product_view_0,\n",
         "      value_extractor=lambda row: row.has_conversion)\n",
         "\n",
-        "  dp_result = dp_engine.aggregate(data, params, data_extractors)\n",
+        "  dp_result = dp_engine.aggregate(data, params, data_extractors, public_partitions)\n",
         "\n",
         "  budget_accountant.compute_budgets()\n",
         "\n",
@@ -810,8 +809,7 @@
         "     noise_kind=pipeline_dp.NoiseKind.LAPLACE,\n",
         "     metrics=[pipeline_dp.Metrics.COUNT],\n",
         "     max_partitions_contributed=1,\n",
-        "     max_contributions_per_partition=1,\n",
-        "     public_partitions=public_partitions_products)\n",
+        "     max_contributions_per_partition=1)\n",
         "```\n",
         "\n",
         "  If you use public partitions and `LAPLACE` noise, it's possible to set the `total_delta` argument to a `0` value.\n",
@@ -970,7 +968,6 @@
         "     max_contributions_per_partition=1,\n",
         "     min_value=0,\n",
         "     max_value=100,\n",
-        "     public_partitions=public_partitions_products,\n",
         "     budget_weight=1/3)\n",
         "```\n",
         "\n",
@@ -989,7 +986,8 @@
         "\n",
         " dp_result_conversion_value_metrics = (\n",
         "     dp_engine.aggregate(data, params_conversion_value_metrics,\n",
-        "         data_extractors_conversion_value_metrics))\n",
+        "         data_extractors_conversion_value_metrics,\n",
+        "         public_partitions_products))\n",
         "```\n",
         "\n",
         "\n",
@@ -1003,7 +1001,6 @@
         "     max_contributions_per_partition=1,\n",
         "     min_value=0,\n",
         "     max_value=1,\n",
-        "     public_partitions=public_partitions_products,\n",
         "     budget_weight=2/3)\n",
         "     ```\n",
         "     ```\n",
@@ -1016,7 +1013,8 @@
         " dp_result_conversion_rate_metrics = (\n",
         "     dp_engine.aggregate(\n",
         "         data, params_conversion_rate_metrics,\n",
-        "         data_extractors_conversion_rate_metrics))\n",
+        "         data_extractors_conversion_rate_metrics,\n",
+        "         public_partitions_products))\n",
         "```\n",
         "\n",
         "    The only change is in the `pipeline_dp.AggregateParams` instance, in which you now define `mean` and `count` as aggregations, and assign two-thirds of your privacy budget to this calculation. Because you want to have the same contribution bounds for both statistics and calculate them on top of the same `has_conversion` variable, you can combine them in the same `pipeline_dp.AggregateParams` instance and calculate them at the same time.\n",
@@ -1074,7 +1072,6 @@
         "      max_contributions_per_partition=1,\n",
         "      min_value=0,\n",
         "      max_value=100,\n",
-        "      public_partitions=public_partitions,\n",
         "      budget_weight=1/3)\n",
         "\n",
         "  data_extractors_conversion_value_metrics = pipeline_dp.DataExtractors(\n",
@@ -1084,7 +1081,7 @@
         "\n",
         "  dp_result_conversion_value_metrics = (\n",
         "      dp_engine.aggregate(data, params_conversion_value_metrics,\n",
-        "          data_extractors_conversion_value_metrics))\n",
+        "          data_extractors_conversion_value_metrics, public_partitions))\n",
         "\n",
         "  params_conversion_rate_metrics = pipeline_dp.AggregateParams(\n",
         "      noise_kind=pipeline_dp.NoiseKind.LAPLACE,\n",
@@ -1093,7 +1090,6 @@
         "      max_contributions_per_partition=1,\n",
         "      min_value=0,\n",
         "      max_value=1,\n",
-        "      public_partitions=public_partitions,\n",
         "      budget_weight=2/3)\n",
         "\n",
         "  data_extractors_conversion_rate_metrics = pipeline_dp.DataExtractors(\n",
@@ -1103,7 +1099,7 @@
         "\n",
         "  dp_result_conversion_rate_metrics = (\n",
         "      dp_engine.aggregate(data, params_conversion_rate_metrics,\n",
-        "          data_extractors_conversion_rate_metrics))\n",
+        "          data_extractors_conversion_rate_metrics, public_partitions))\n",
         "\n",
         "  budget_accountant.compute_budgets()\n",
         "\n",

diff --git a/examples/movie_view_ratings/run_all_frameworks.py b/examples/movie_view_ratings/run_all_frameworks.py
@@ -77,8 +77,7 @@ def calc_dp_rating_metrics(movie_views, backend, public_partitions):
         max_partitions_contributed=2,
         max_contributions_per_partition=1,
         min_value=1,
-        max_value=5,
-        public_partitions=public_partitions)
+        max_value=5)
 
     # Specify how to extract privacy_id, partition_key and value from an
     # element of movie view collection.
@@ -88,7 +87,8 @@ def calc_dp_rating_metrics(movie_views, backend, public_partitions):
         value_extractor=lambda mv: mv.rating)
 
     # Run aggregation.
-    dp_result = dp_engine.aggregate(movie_views, params, data_extractors)
+    dp_result = dp_engine.aggregate(movie_views, params, data_extractors,
+                                    public_partitions)
 
     budget_accountant.compute_budgets()
     return dp_result