Lines Matching +full:38 +full:- +full:bloom

9 --------
30 <img src="data-flow.png" alt="Diagram of RAPPOR Data Flow" />
35   2. Obscure each value with the RAPPOR privacy-preserving reporting mechanism.
39   1. Aggregate the reports by summing bits (i.e. make a counting Bloom filter)
49 -----------------------------
53 <!-- TODO: a realistic data set would be nice? How could we generate one?  -->
88 -------------------
111     2,      38,     0010001111001010
120 We also get a one-row CSV file that contains the RAPPOR parameters:
133 - `exp_hist.csv`: The true histogram of input values.  This is used only for
135 - `exp_true_inputs.txt`: A list of the unique values reported, e.g. `v1` ..
140 ---------------------
152 values expected.  <!-- TODO: more detail -->
154 There are 17 columns.  The left-most column is the total number of reports in
156 Bloom filter.  Each column contains the number of reports with that bit set
162 --------------------
165 the candidate strings.  This is done in the `more-candidates` /
166 `print-candidates` functions in `demo.sh`.
169 editing the invocation of `print-candidates` in `run-dist`:
174     print-candidates $dist 'v1|v2'  > _tmp/${dist}_candidates.txt
176 In general, coming up with candidates is an application- or metric-specific
192 columns: for `m = 64` cohorts times `k = 2` hash functions in the Bloom filter.
194 <!-- TODO: more detail about setting params?  Examples of coming up with
195 candidate strings? -->
198 ------------------
216 <!-- TODO: 
217      - how to change flags
218      - more detail on what the various parameters do
219      - association analysis
220      - basic RAPPOR
221      - longitudinal privacy
222 -->
225 ----------
233 [rappor-[email protected]](https://groups.google.com/forum/#!forum/rappor-discuss).