public static Dictionary <string, Tuple <T1, T1, double> > getActualPairsDictionary <T1>(Dictionary <T1, List <int> > wordList, double threshold)
{
Dictionary <string, Tuple <T1, T1, double> > pairsDictionary = new Dictionary <string, Tuple <T1, T1, double> >();
List <T1> docList = wordList.Keys.ToList();
int i, j;
string sum;
double jaccard;
for (i = 0; i < docList.Count; i++)
{
for (j = i + 1; j < docList.Count; j++)
{
//sum = docList[i] + "#" + docList[j];
sum = getKeyFromPair(docList[i], docList[j]);
if (!pairsDictionary.ContainsKey(sum))
{
jaccard = Jaccard.Calc(wordList[docList[i]], wordList[docList[j]]);
if (jaccard >= threshold)
{
pairsDictionary.Add(sum, new Tuple <T1, T1, double>(docList[i], docList[j], jaccard));
}
}
}
}
return(pairsDictionary);
}
public static double calculateMinHashFunctionsAccuracy <T1>(Dictionary <T1, List <int> > wordListActual, Dictionary <T1, List <uint> > wordListMinHash)
{
List <T1> docList = wordListActual.Keys.ToList();
int i, j;
double jaccard_actual, jaccard_minhash;
double total_diff_perc = 0;
double diff_perc;
int pair_count = 0;
for (i = 0; i < docList.Count; i++)
{
for (j = i + 1; j < docList.Count; j++)
{
jaccard_actual = Jaccard.Calc(wordListActual[docList[i]], wordListActual[docList[j]]);
if (jaccard_actual > 0)
{
jaccard_minhash = Jaccard.Calc(wordListMinHash[docList[i]],
wordListMinHash[docList[j]]);
diff_perc = (Math.Abs(jaccard_minhash - jaccard_actual) / jaccard_actual) * 100;
total_diff_perc += diff_perc;
pair_count++;
}
}
}
double avg_diff_perc = total_diff_perc / pair_count;
Console.WriteLine("Average diff from Actual and MinHash Jaccard is: " + avg_diff_perc + " %");
return(avg_diff_perc);
}
public static void JaccardTest1()
{
Dictionary <int, string> wordDict = new Dictionary <int, string>();
wordDict.Add(1, "Word1");
wordDict.Add(2, "Word2");
wordDict.Add(3, "Word3");
wordDict.Add(4, "Word4");
List <int> doc1 = new List <int>();
doc1.Add(2);
doc1.Add(3);
doc1.Add(4);
doc1.Add(2);
List <int> doc2 = new List <int>();
doc2.Add(1);
doc2.Add(5);
doc2.Add(4);
doc2.Add(2);
List <int> doc3 = new List <int>();
doc3.Add(1);
Console.WriteLine("Jaccard: " + Jaccard.Calc(doc1, doc2));
Console.WriteLine("Jaccard: " + Jaccard.Calc(doc1, doc1));
Console.WriteLine("Jaccard: " + Jaccard.Calc(doc1, doc3));
}
static void processNumbersTest3(bool sample = false, int sample_size = 1000)
{
int numHashFunctions = 128;
int universeSize = 1000;
double simThreshold = 0.65;
double atn = 0.05;
MinHasher2 mh2 = new MinHasher2(numHashFunctions, simThreshold);
NumberDocumentCreator numDocCreator2 = new NumberDocumentCreator(10, universeSize);
int[] a1 = numDocCreator2.createDocument(universeSize);
int[] a2 = numDocCreator2.createDocument(universeSize);
Console.WriteLine("Actual jaccaard: " + MinHasher2.calculateJaccard(a1, a2));
Console.WriteLine("MinHash jaccaard: " + MinHasher2.calculateJaccard(mh2.getMinHashSignature(a1), mh2.getMinHashSignature(a2)));
return;
MinHasher3 mh = new MinHasher3(universeSize, numHashFunctions);
MinHasher_Buckets3 mhb = new MinHasher_Buckets3(mh, simThreshold, atn);
NumberDocumentCreator numDocCreator = new NumberDocumentCreator(10, universeSize);
List <int> s1 = numDocCreator.createDocument(universeSize).ToList();
List <int> s2 = numDocCreator.createDocument(universeSize).ToList();
Console.WriteLine("Actual jaccaard: " + Jaccard.Calc(s1, s2));
Console.WriteLine("MinHash jaccaard: " + Jaccard.Calc(mh.GetMinHash(s1), mh.GetMinHash(s2)));
return;
Dictionary <int, List <int> > wordList = numDocCreator.documentCollectionList;
//Now create a MinHasher object to minhash each of the documents created above
//using 300 unique hashing functions.
//MinHasher minHasher = new MinHasher(500, 5);
Console.WriteLine("\r\nGenerating MinHash signatures ... ");
Dictionary <int, List <uint> > docMinhashes = mhb.createMinhashCollection(wordList); //minHasher.createMinhashCollection(wordList);
double avg_diff_perc_from_actual_and_minhash_jaccard = Util.calculateMinHashFunctionsAccuracy(wordList, docMinhashes);
/*StringDocumentCreator strDocCreator = new StringDocumentCreator(100, 10000);
*
* Dictionary<int, string[]> wordList2 = strDocCreator.documentCollection;
*
* //Now create a MinHasher object to minhash each of the documents created above
* //using 300 unique hashing functions.
* //MinHasher minHasher = new MinHasher(500, 5);
* Console.WriteLine("\r\nGenerating MinHash signatures ... ");
* Dictionary<int, int[]> docMinhashes2 = minHasher.createMinhashCollection(wordList2);
* double avg_diff_perc_from_actual_and_minhash_jaccard2 = Util.calculateMinHashFunctionsAccuracy(wordList2, docMinhashes2);
*/
Console.ReadKey();
}
/*
* Graph::generateCommonPairs
* Generate the list of Vertex pairs that share common properties. The pairs in this list will be input to the OurSim calculations
* We are doing this to recover from n square complexity
*/
public Dictionary <string, Tuple <T1, T1, double> > generateVertexPairs <T1, T>(Dictionary <string, HashSet <T1> > m_lshBuckets, Dictionary <T1, List <uint> > docMinhashes, Dictionary <T1, List <int> > wordList, bool exclude_sim_under_threshold, string output_file_name)
{
//Dictionary<string, HashSet<int>> m_lshBuckets = new Dictionary<string, HashSet<int>>();
Dictionary <string, Tuple <T1, T1, double> > pairsDictionary = new Dictionary <string, Tuple <T1, T1, double> >();
List <T1> docList;
string sum;
int loopCount = 0;
double jaccard;
int bucketIndex = 0;
StreamWriter wr = null;
string sep = " #-# ";
string temp_file_name = output_file_name + ".temp";
if (output_file_name != null)
{
wr = new StreamWriter(temp_file_name); //write the pairs to a file
wr.WriteLine("-common_pairs-");
wr.WriteLine(string.Format("vid1{0}vid2{0}minhash_sim", sep));
}
foreach (var bucket in m_lshBuckets)
{
bucketIndex++;
if (bucket.Value.Count <= 1)
{
continue;
}
docList = bucket.Value.ToList();
int i = 0;
int j = i + 1;
for (i = 0; i < docList.Count; i++)
{
for (j = i + 1; j < docList.Count; j++)
{
//sum = docList[i] + "#" + docList[j];
sum = Util.getKeyFromPair(docList[i], docList[j]);
if (!pairsDictionary.ContainsKey(sum))
{
//jaccard = calculateJaccard(docMinhashes[docList[i]], docMinhashes[docList[j]]);
jaccard = Jaccard.Calc(wordList[docList[i]], wordList[docList[j]]);
if (!exclude_sim_under_threshold || jaccard >= sim_threshold)
{
pairsDictionary.Add(sum, new Tuple <T1, T1, double>(docList[i], docList[j], jaccard));
if (output_file_name != null)
{
wr.WriteLine(docList[i] + sep + docList[j] + sep + jaccard);
}
}
}
loopCount++;
}
}
}
Console.WriteLine("\r\nBucket generating candidate pairs complexity: " + loopCount);
if (wr != null)
{
wr.Close();
wr = new StreamWriter(output_file_name);
wr.WriteLine(pairsDictionary.Count); //prepending the size of the pairs. needed for c++ vector space allocation
StreamReader rd = new StreamReader(temp_file_name);
string buf;
while ((buf = rd.ReadLine()) != null)
{
wr.WriteLine(buf);
}
rd.Close();
wr.Close();
File.Delete(temp_file_name);
}
return(pairsDictionary);
}
// Calculates the similarity of two lists of min hash values. Approximately Numerically equivilant to Jaccard Similarity
public double Similarity(List <uint> l1, List <uint> l2)
{
Jaccard jac = new Jaccard();
return(Jaccard.Calc(l1, l2));
}
public static void MinHasher3TestFunc1()
{
List <int> inums1 = new List <int>();
inums1.Add(10);
inums1.Add(8);
inums1.Add(11);
inums1.Add(13);
inums1.Add(2);
inums1.Add(17);
inums1.Add(3);
inums1.Add(1);
inums1.Add(19);
inums1.Add(11);
inums1.Add(100);
inums1.Add(82);
inums1.Add(115);
inums1.Add(13);
inums1.Add(2);
inums1.Add(107);
inums1.Add(3);
inums1.Add(1);
inums1.Add(19);
inums1.Add(110);
inums1.Add(10);
inums1.Add(8);
inums1.Add(110);
inums1.Add(131);
inums1.Add(2);
inums1.Add(173);
inums1.Add(3);
inums1.Add(1);
inums1.Add(19);
inums1.Add(114);
inums1.Add(10);
inums1.Add(8);
inums1.Add(11);
inums1.Add(13);
inums1.Add(2);
inums1.Add(17);
inums1.Add(3);
inums1.Add(1);
inums1.Add(19);
inums1.Add(115);
inums1.Add(10);
inums1.Add(8);
inums1.Add(11);
inums1.Add(133);
inums1.Add(2);
inums1.Add(17);
inums1.Add(3);
inums1.Add(1);
inums1.Add(19);
inums1.Add(11);
inums1.Add(10);
inums1.Add(8);
inums1.Add(11);
inums1.Add(13);
inums1.Add(2);
inums1.Add(17);
inums1.Add(3);
inums1.Add(1);
inums1.Add(19);
inums1.Add(171);
List <int> inums2 = new List <int>();
inums2.Add(1);
inums2.Add(2);
inums2.Add(5);
inums2.Add(9);
inums2.Add(12);
inums2.Add(17);
inums2.Add(13);
inums2.Add(11);
inums2.Add(9);
inums2.Add(10);
inums2.Add(1);
inums2.Add(2);
inums2.Add(5);
inums2.Add(9);
inums2.Add(12);
inums2.Add(17);
inums2.Add(13);
inums2.Add(11);
inums2.Add(9);
inums2.Add(10);
inums2.Add(1);
inums2.Add(2);
inums2.Add(5);
inums2.Add(9);
inums2.Add(12);
inums2.Add(17);
inums2.Add(13);
inums2.Add(151);
inums2.Add(9);
inums2.Add(510);
inums2.Add(1);
inums2.Add(2);
inums2.Add(5);
inums2.Add(9);
inums2.Add(12);
inums2.Add(17);
inums2.Add(13);
inums2.Add(11);
inums2.Add(95);
inums2.Add(10);
inums2.Add(1);
inums2.Add(23);
inums2.Add(5);
inums2.Add(9);
inums2.Add(162);
inums2.Add(17);
inums2.Add(13);
inums2.Add(11);
inums2.Add(93);
inums2.Add(10);
inums2.Add(19);
inums2.Add(23);
inums2.Add(5);
inums2.Add(9);
inums2.Add(12);
inums2.Add(17);
inums2.Add(13);
inums2.Add(141);
inums2.Add(94);
inums2.Add(10);
int universeSize = Jaccard.unionSize(inums1, inums2);
MinHasher3 mh = new MinHasher3(universeSize, 135);
List <uint> hvs1 = mh.GetMinHash(inums1).ToList();
List <uint> hvs2 = mh.GetMinHash(inums2).ToList();
Console.WriteLine();
Console.WriteLine("Estimated similarity: " + mh.Similarity(hvs1, hvs2));
Console.WriteLine("Jaccard similarity: " + Jaccard.Calc(inums1, inums2));
Console.WriteLine("done");
}
