This is the Java library of the core algorithms used to do fair search.
You can import the library with maven in your pom.xml
file:
<dependency>
<groupId>com.github.fair-search</groupId>
<artifactId>fairsearch-core</artifactId>
<version>1.0.2</version>
</dependency>
or, if you are using Gradle, in your build.gradle
file add this in the dependencies
block:
compile "com.github.fair-search:fairsearch-core:1.0.2"
And, that's it!
Add the JAR file to the build path of your project and you are set. The key methods are contained in the following classes:
com.github.fairsearch.Fair
com.github.fairsearch.Simulator
The library contains sufficient Java doc for each of the functions.
Creating and analyzing mtables:
package com.github.fairsearch.examples;
import com.github.fairsearch.Fair;
import com.github.fairsearch.Simulator;
public class HelloWorld {
public static void main(String[] args) {
// number of topK elements returned (value should be between 10 and 400)
int k = 20;
// proportion of protected candidates in the topK elements (value shuld be between 0.02 and 0.98)
double p = 0.25;
// significance level (value should be between 0.01 and 0.15)
double alpha = 0.1;
//create the Fair object
Fair fiar = new Fair(k, p, alpha);
//create an mtable using alpha unadjusted
int[] unadjustedMTable = fair.createUnadjustedMTable();
//unadjustedMTable -> [0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 2, 2, 2, 2, 2, 3]
//analytically calculate the fail probability
analytical = fair.computeFailureProbability(unadjustedMTable);
//analytical -> 0.14688718869911077
//create an mtable using alpha adjusted
int[] adjustedMTable = fair.createAdjustedMTable();
//adjustedMTable -> [0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 2, 2, 2, 2, 2]
//analytically calculate the fail probability
analytical = fair.computeFailureProbability(adjustedMTable);
//analytical -> 0.10515247355215218
}
}
Generate random rankings and analyze them:
//set number of rankings you want to generate
int M = 10000;
//generate rankings using the simulator
TopDocs[] rankings = Simulator.generateRankings(M, k, p);
//experimentally calculate the fail probability
double experimental = Simulator.computeFailureProbability(mtable, rankings);
//experimental -> 0.1054
Let's get the alpha adjusted (used to create an adjusted mtable)
//get alpha adjusted
double alphaAdjusted = fair.adjustAlpha();
//alphaAdjusted -> 0.07812500000000001
Apply a fair re-ranking to a given ranking:
//import the FairScoreDoc class at the top
import com.github.fairsearch.FairScoreDoc;
//also, import Lucene's TopDocs class
import org.apache.lucene.search.TopDocs;
//let's manually create an unfair ranking (False -> unprotected, True -> protected)
//in this example the first document (docid=20) has a score of 20, the last document (docid=1) a score of 1
FairScoreDoc[] docs = {new FairScoreDoc(20, 20, false), new FairScoreDoc(19, 19, false),
new FairScoreDoc(18, 18, false), new FairScoreDoc(17, 17, false),
new FairScoreDoc(16, 16, false), new FairScoreDoc(15, 15, false),
new FairScoreDoc(14, 14, false), new FairScoreDoc(13, 13, false),
new FairScoreDoc(12, 12, false), new FairScoreDoc(11, 11, false),
new FairScoreDoc(10, 10, false), new FairScoreDoc(9, 9, false),
new FairScoreDoc(8, 8, false), new FairScoreDoc(7, 7, false),
new FairScoreDoc(6, 6, false), new FairScoreDoc(5, 5, true),
new FairScoreDoc(4, 4, true), new FairScoreDoc(3, 3, true),
new FairScoreDoc(2, 2, true), new FairScoreDoc(1, 1, true)};
TopDocs unfairRanking = new TopDocs(docs.length, docs1, Float.NaN);
// let's check if the ranking is fair
boolean isFair = fair.isFair(unfairRanking);
// isFair -> false
//now re-rank the unfair ranking
TopDocs reRanked = fair.reRank(unfairRanking);
// now let's see if the ranking is fair
boolean isFair = fair.isFair(reRanked);
// isFair -> true
Note: The numbers shown here may differ slightly from run to run as there is randomness factor involved.
- Clone this repository
git clone https://github.com/fair-search/fairsearch-core.git
- Change directory to the directory where you cloned the repository
cd WHERE_ITS_DOWNLOADED/fairsearch-core/java
- Use any IDE to work with the code
If you want to make your own builds you can do that with the Gradle wrapper:
- To make a JAR without the external dependencies:
./gradlew clean jar
- To make a JAR with all external dependencies included:
./gradlew clean farJar
The output will go under build/libs
.
Just run:
./gradlew clean check
Note: The simulator tests take a bit longer time to execute. Also, because there is a randomness factor involved in the tests, it can happen that (rarely) they fail sometimes.
The FA*IR algorithm is described on this paper:
- Meike Zehlike, Francesco Bonchi, Carlos Castillo, Sara Hajian, Mohamed Megahed, Ricardo Baeza-Yates: "FA*IR: A Fair Top-k Ranking Algorithm". Proc. of the 2017 ACM on Conference on Information and Knowledge Management (CIKM).
This code was developed by Ivan Kitanovski and Tom Sühr based on the paper. See the license file for more information.
For any questions contact Meike Zehlike
You can also see the FA*IR plug-in for ElasticSearch and FA*IR search core Python library.