1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
|
package cardinality;
import shared.Parser;
import shared.Tools;
import structures.LongList;
/**
* @author Brian Bushnell
* @date Mar 6, 2020
*
*/
public final class LogLog16 extends CardinalityTracker {
/*--------------------------------------------------------------*/
/*---------------- Initialization ----------------*/
/*--------------------------------------------------------------*/
/** Create a LogLog with default parameters */
LogLog16(){
this(2048, 31, -1, 0);
}
/** Create a LogLog with parsed parameters */
LogLog16(Parser p){
super(p);
maxArray=new char[buckets];
}
/**
* Create a LogLog with specified parameters
* @param buckets_ Number of buckets (counters)
* @param k_ Kmer length
* @param seed Random number generator seed; -1 for a random seed
* @param minProb_ Ignore kmers with under this probability of being correct
*/
LogLog16(int buckets_, int k_, long seed, float minProb_){
super(buckets_, k_, seed, minProb_);
maxArray=new char[buckets];
}
/*--------------------------------------------------------------*/
/*---------------- Methods ----------------*/
/*--------------------------------------------------------------*/
//Restores floating point to integer
private long restore(int score){
long lowbits=(~score)&mask;
int leading=(int)(score>>>mantissabits);
long mantissa=(1L<<mantissabits)|lowbits;
int shift=wordlen-leading-mantissabits-1;
long original=mantissa<<shift;
return original;
}
@Override
public final long cardinality(){
double difSum=0;
double hSum=0;
double gSum=0;
double rSum=0;
double estLogSum=0;
int count=0;
LongList list=new LongList(buckets);
for(int i=0; i<maxArray.length; i++){
int max=maxArray[i];
long val=restore(max);
if(max>0 && val>0){
long dif=val;
difSum+=dif;
hSum+=1.0/Tools.max(1, dif);
gSum+=Math.log(Tools.max(1, dif));
rSum+=Math.sqrt(dif);
count++;
double est=2*(Long.MAX_VALUE/(double)dif)*SKIPMOD;
estLogSum+=Math.log(est);
list.add(dif);
}
}
final int div=Tools.max(count, 1);//Could be count or buckets but one causes problems
final double mean=difSum/div;
double hmean=hSum/div;
double gmean=gSum/div;
double rmean=rSum/div;
hmean=1.0/hmean;
gmean=Math.exp(gmean);
rmean=rmean*rmean;
list.sort();
final long median=list.median();
final double mwa=list.medianWeightedAverage();
//What to use as the value from the counters
final double proxy=(USE_MEAN ? mean : USE_MEDIAN ? median : USE_MWA ? mwa : USE_HMEAN ? hmean : USE_GMEAN ? gmean : mean);
final double estimatePerSet=2*(Long.MAX_VALUE/proxy)*SKIPMOD;
final double total=estimatePerSet*div*((count+buckets)/(float)(buckets+buckets));
final double estSum=div*Math.exp(estLogSum/(Tools.max(div, 1)));
double medianEst=2*(Long.MAX_VALUE/(double)median)*SKIPMOD*div;
// new Exception().printStackTrace();
// System.err.println(maxArray);
//// Overall, it looks like "total" is the best, then "estSum", then "medianEst" is the worst, in terms of variance.
// System.err.println("difSum="+difSum+", count="+count+", mean="+mean+", est="+estimatePerSet+", total="+(long)total);
// System.err.println("estSum="+(long)estSum+", median="+median+", medianEst="+(long)medianEst);
long cardinality=(long)(total);
lastCardinality=cardinality;
return cardinality;
}
@Override
public final void add(CardinalityTracker log){
assert(log.getClass()==this.getClass());
add((LogLog16)log);
}
public void add(LogLog16 log){
if(maxArray!=log.maxArray){
for(int i=0; i<buckets; i++){
maxArray[i]=Tools.max(maxArray[i], log.maxArray[i]);
}
}
}
@Override
public void hashAndStore(final long number){
// if(number%SKIPMOD!=0){return;} //Slows down moderately
long key=number;
// key=hash(key, tables[((int)number)&numTablesMask]);
key=Tools.hash64shift(key);
// if(key<0 || key>maxHashedValue){return;}//Slows things down by 50% lot, mysteriously
int leading=Long.numberOfLeadingZeros(key)&63;//mask is used to keep number in 6 bits
// counts[leading]++;
// if(leading<3){return;}//Speeds up by 20%, even more at 4. Slows at 2.
int shift=wordlen-leading-mantissabits-1;
int score=(leading<<mantissabits)+(int)((~(key>>>shift))&mask);
// assert(false) : "\n"+Long.toBinaryString(key)+", leading="+leading+", shift="+shift+"\n"+Long.toBinaryString(score);
//+"\n"+score+"\n"+restore(score);
// final int bucket=(int)((number&Integer.MAX_VALUE)%buckets);
final int bucket=(int)(key&bucketMask);
int newValue=Tools.max(score, maxArray[bucket]);
assert(newValue>=0 && newValue<=Character.MAX_VALUE) : newValue;
maxArray[bucket]=(char)newValue;
}
@Override
public final float[] compensationFactorLogBucketsArray(){
return null;
}
/*--------------------------------------------------------------*/
/*---------------- Fields ----------------*/
/*--------------------------------------------------------------*/
private final char[] maxArray;
/*--------------------------------------------------------------*/
/*---------------- Statics ----------------*/
/*--------------------------------------------------------------*/
public static void setMantissaBits(int x){
assert(x>=0 && x<25);
assert(x+6<32);
mantissabits=x;
mask=(1<<mantissabits)-1;
}
private static final int wordlen=64;
/** Precision or mantissa bits.
* This should not be changed. As long as it is >10 the result will be accurate.
* At low values like 2 the cardinality estimate becomes too high due to a loss of precision,
* and would need a fixed multiplier.
*/
private static int mantissabits=10;//10 is the max possible
private static int mask=(1<<mantissabits)-1;
}
|
