Code for the Location Heatmaps paper.

analytics/location_heatmaps/README.md

@@ -1,4 +1,24 @@
-# Location Heatmaps
+# Towards Sparse Federated Analytics: Location Heatmaps under Distributed Differential Privacy with Secure Aggregation
 
-This directory contains work-in-progress code to experiment with differential
-privacy on location heatmaps.
+This directory contains code to experiment with differential
+privacy on location heatmaps and based on [this paper](https://arxiv.org/abs/2111.02356).
+Here is the example of the algorithm running on NYC location map from 
+[NYTimes](https://www.nytimes.com/2018/12/14/reader-center/phone-data-location-investigation.html):
+
+![Alt Text](animation.gif)
+
+To experiment with the code there is a working [notebook](dp_location_heatmaps.ipynb) 
+with all the examples from the paper, please don't hesitate to contact the 
+[author](mailto:eugene@cs.cornell.edu) or raise an issue.
+
+For citation:
+```
+@misc{bagdasaryan2021sparse,
+      title={Towards Sparse Federated Analytics: Location Heatmaps under Distributed Differential Privacy with Secure Aggregation}, 
+      author={Eugene Bagdasaryan and Peter Kairouz and Stefan Mellem and Adrià Gascón and Kallista Bonawitz and Deborah Estrin and Marco Gruteser},
+      year={2021},
+      eprint={2111.02356},
+      archivePrefix={arXiv},
+      primaryClass={cs.CR}
+}
+```

analytics/location_heatmaps/config.py

@@ -0,0 +1,31 @@
+from dataclasses import dataclass
+from typing import List, Any
+import numpy as np
+TOPK = 1000
+TOTAL_SIZE = 1024
+
+
+@dataclass
+class Config:
+    dataset: List
+    image: Any
+    level_sample_size: int = 10000
+    secagg_round_size: int = 10000
+    split_threshold: float = 0
+    collapse_threshold: float = None
+    eps_func: Any = lambda x, y: 1
+    total_epsilon_budget: float = None
+    top_k: int = TOPK
+    partial:int = 100
+    max_levels: int = 10
+    split_threshold_func: Any = None
+    collapse_func: Any = None
+    total_size: int = TOTAL_SIZE
+    min_dp_size: int = None
+    dropout_rate: int = None
+    output_flag: int = True
+    quantize: bool = None
+    noise_class: Any = None
+    save_gif: bool = False
+    aux_data: bool = False
+    start_with_level: int = 0