| line |
stmt |
bran |
cond |
sub |
pod |
time |
code |
|
1
|
|
|
|
|
|
|
package Math::Round::Fair; |
|
2
|
6
|
|
|
6
|
|
123381
|
use warnings; |
|
|
6
|
|
|
|
|
12
|
|
|
|
6
|
|
|
|
|
213
|
|
|
3
|
6
|
|
|
6
|
|
29
|
use strict; |
|
|
6
|
|
|
|
|
10
|
|
|
|
6
|
|
|
|
|
186
|
|
|
4
|
6
|
|
|
6
|
|
128
|
use 5.005000; |
|
|
6
|
|
|
|
|
25
|
|
|
|
6
|
|
|
|
|
320
|
|
|
5
|
6
|
|
|
6
|
|
4954
|
use Devel::Assert; |
|
|
6
|
|
|
|
|
69948
|
|
|
|
6
|
|
|
|
|
44
|
|
|
6
|
6
|
|
|
6
|
|
575
|
use Carp; |
|
|
6
|
|
|
|
|
12
|
|
|
|
6
|
|
|
|
|
513
|
|
|
7
|
6
|
|
|
6
|
|
31
|
use base qw/Exporter/; |
|
|
6
|
|
|
|
|
13
|
|
|
|
6
|
|
|
|
|
509
|
|
|
8
|
6
|
|
|
6
|
|
29
|
use List::Util qw/shuffle sum min/; |
|
|
6
|
|
|
|
|
10
|
|
|
|
6
|
|
|
|
|
661
|
|
|
9
|
6
|
|
|
6
|
|
5852
|
use POSIX qw/floor ceil DBL_EPSILON/; |
|
|
6
|
|
|
|
|
49074
|
|
|
|
6
|
|
|
|
|
42
|
|
|
10
|
|
|
|
|
|
|
|
|
11
|
|
|
|
|
|
|
our $VERSION = '0.03'; |
|
12
|
|
|
|
|
|
|
|
|
13
|
|
|
|
|
|
|
our @EXPORT_OK = qw/round_fair round_adjacent/; |
|
14
|
|
|
|
|
|
|
|
|
15
|
|
|
|
|
|
|
BEGIN { |
|
16
|
6
|
|
|
6
|
|
1092
|
my $debug; |
|
17
|
13
|
|
|
13
|
0
|
902
|
sub DEBUG { $debug } |
|
18
|
|
|
|
|
|
|
|
|
19
|
6
|
|
100
|
|
|
45
|
$debug = $ENV{MATH_ROUND_FAIR_DEBUG} || 0; |
|
20
|
6
|
50
|
|
6
|
|
7308
|
use Devel::Assert DEBUG() ? ('-all -verbose') : (); |
|
|
6
|
|
|
|
|
13
|
|
|
|
6
|
|
|
|
|
24
|
|
|
21
|
|
|
|
|
|
|
|
|
22
|
|
|
|
|
|
|
# used in assertions |
|
23
|
6
|
100
|
|
5
|
|
16
|
eval q{use Perl6::Junction 'none'} if DEBUG(); |
|
|
5
|
|
|
|
|
3375
|
|
|
|
5
|
|
|
|
|
58413
|
|
|
|
5
|
|
|
|
|
1940
|
|
|
24
|
|
|
|
|
|
|
} |
|
25
|
|
|
|
|
|
|
|
|
26
|
|
|
|
|
|
|
|
|
27
|
|
|
|
|
|
|
=head1 NAME |
|
28
|
|
|
|
|
|
|
|
|
29
|
|
|
|
|
|
|
Math::Round::Fair - distribute rounding errors fairly |
|
30
|
|
|
|
|
|
|
|
|
31
|
|
|
|
|
|
|
=head1 SYNOPSIS |
|
32
|
|
|
|
|
|
|
|
|
33
|
|
|
|
|
|
|
use Math::Round::Fair 'round_fair', 'round_adjacent'; |
|
34
|
|
|
|
|
|
|
|
|
35
|
|
|
|
|
|
|
my $cents = 7; |
|
36
|
|
|
|
|
|
|
my @weights = (1, 2, 3, 2, 1); |
|
37
|
|
|
|
|
|
|
my @allocation = round_fair($cents, @weights); |
|
38
|
|
|
|
|
|
|
|
|
39
|
|
|
|
|
|
|
print "@allocation\n"; |
|
40
|
|
|
|
|
|
|
|
|
41
|
|
|
|
|
|
|
# output will be one of the following: |
|
42
|
|
|
|
|
|
|
# 0 1 3 2 1 |
|
43
|
|
|
|
|
|
|
# 0 2 2 2 1 |
|
44
|
|
|
|
|
|
|
# 0 2 3 1 1 |
|
45
|
|
|
|
|
|
|
# 0 2 3 2 0 |
|
46
|
|
|
|
|
|
|
# 1 1 2 2 1 |
|
47
|
|
|
|
|
|
|
# 1 1 3 1 1 |
|
48
|
|
|
|
|
|
|
# 1 1 3 2 0 |
|
49
|
|
|
|
|
|
|
# 1 2 2 1 1 |
|
50
|
|
|
|
|
|
|
# 1 2 2 2 0 |
|
51
|
|
|
|
|
|
|
|
|
52
|
|
|
|
|
|
|
my @total; |
|
53
|
|
|
|
|
|
|
for ( 1..900 ) { |
|
54
|
|
|
|
|
|
|
@allocation = round_fair($cents, @weights); |
|
55
|
|
|
|
|
|
|
@total[$_] += @allocation[$_] for 0..$#allocation; |
|
56
|
|
|
|
|
|
|
} |
|
57
|
|
|
|
|
|
|
print "@total\n"; |
|
58
|
|
|
|
|
|
|
|
|
59
|
|
|
|
|
|
|
# output will be *near* 700 1400 2100 1400 700, e.g.: |
|
60
|
|
|
|
|
|
|
# 698 1411 2096 1418 677 |
|
61
|
|
|
|
|
|
|
|
|
62
|
|
|
|
|
|
|
|
|
63
|
|
|
|
|
|
|
my @rounded = round_adjacent(0.95, 0.65, 0.41, 0.99); |
|
64
|
|
|
|
|
|
|
# @rounded will be one of the following: |
|
65
|
|
|
|
|
|
|
# 59% of the time: 1, 1, 0, 1 |
|
66
|
|
|
|
|
|
|
# 35% of the time: 1, 0, 1, 1 |
|
67
|
|
|
|
|
|
|
# 5% of the time: 0, 1, 1, 1 |
|
68
|
|
|
|
|
|
|
# 1% of the time: 1, 1, 1, 0 |
|
69
|
|
|
|
|
|
|
|
|
70
|
|
|
|
|
|
|
|
|
71
|
|
|
|
|
|
|
=head1 DESCRIPTION |
|
72
|
|
|
|
|
|
|
|
|
73
|
|
|
|
|
|
|
This module provides two exportable functions, C, which |
|
74
|
|
|
|
|
|
|
allocates an integer value, fairly distributing rounding errors, and |
|
75
|
|
|
|
|
|
|
C, which takes a list of real numbers and rounds them up, or |
|
76
|
|
|
|
|
|
|
down, to an adjacent integer, fairly. Both functions return a list of fairly |
|
77
|
|
|
|
|
|
|
rounded integer values. |
|
78
|
|
|
|
|
|
|
|
|
79
|
|
|
|
|
|
|
C and C round up, or down, randomly, where the |
|
80
|
|
|
|
|
|
|
probability of rounding up is equal to the fraction to round. For example, 0.5 |
|
81
|
|
|
|
|
|
|
will round to 1.0 with a probability of 0.5. 0.3 will round to 1.0 3 out of 10 |
|
82
|
|
|
|
|
|
|
times and to zero 7 out of 10 times, on average. |
|
83
|
|
|
|
|
|
|
|
|
84
|
|
|
|
|
|
|
Consider the problem of distributing one indivisible item, for example a penny, |
|
85
|
|
|
|
|
|
|
across three evenly weighted accounts, A, B, and C. |
|
86
|
|
|
|
|
|
|
|
|
87
|
|
|
|
|
|
|
Using a naive approach, none of the accounts will receive an allocation since |
|
88
|
|
|
|
|
|
|
the allocated portion to each is 1/3 and 1/3 rounds to zero. We are left with |
|
89
|
|
|
|
|
|
|
1 unallocated item. |
|
90
|
|
|
|
|
|
|
|
|
91
|
|
|
|
|
|
|
Another approach is to adjust the basis at each step. We start with 1 item to |
|
92
|
|
|
|
|
|
|
allocate to 3 accounts. 1/3 rounds to 0, so account A receives no allocation, |
|
93
|
|
|
|
|
|
|
and we drop it from consideration. Now, we have 2 accounts and one item to |
|
94
|
|
|
|
|
|
|
allocate. 1/2 rounds to 1, so we allocate 1 item to account B. Account C |
|
95
|
|
|
|
|
|
|
gets no allocation since there is nothing left to allocate. |
|
96
|
|
|
|
|
|
|
|
|
97
|
|
|
|
|
|
|
But what happens if we allocate one item to the same three accounts 10,000 |
|
98
|
|
|
|
|
|
|
times? Ideally, two accounts should end up with 3,333 items and one should end |
|
99
|
|
|
|
|
|
|
up with 3,334 items. |
|
100
|
|
|
|
|
|
|
|
|
101
|
|
|
|
|
|
|
Using the naive approach, all three accounts receive no allocation since at |
|
102
|
|
|
|
|
|
|
each round the allocation is 1/3 which rounds to zero. Using the second method, |
|
103
|
|
|
|
|
|
|
account A and account C will receive no allocation, and account B will receive |
|
104
|
|
|
|
|
|
|
a total allocation of 10,000 items. Account B always receives the benefit of |
|
105
|
|
|
|
|
|
|
the rounding errors using the second method. |
|
106
|
|
|
|
|
|
|
|
|
107
|
|
|
|
|
|
|
The algorithm employed by this module uses randomness to ensure a fair distribution of |
|
108
|
|
|
|
|
|
|
rounding errors. In our example problem, we start with 1 item to allocate. We |
|
109
|
|
|
|
|
|
|
calculate account A's share, 1/3. Since it is less than one item, we give it a |
|
110
|
|
|
|
|
|
|
1/3 chance of rounding up (and, therefore, a 2/3 chance of rounding down). It |
|
111
|
|
|
|
|
|
|
wins the allocation 1/3 of the time. 2/3 of the time we continue to B. We |
|
112
|
|
|
|
|
|
|
calculate B's allocation as 1/2 (since there are only 2 accounts remaining and |
|
113
|
|
|
|
|
|
|
one item to allocate). B rounds up 1/2 of 2/3 (or 1/3) of the time and down |
|
114
|
|
|
|
|
|
|
1/2 of 2/3 (or 1/3) of the time. If neither A nor B rounds up (which occurs |
|
115
|
|
|
|
|
|
|
2/3 * 1/2, or 1/3 of the time), C's allocation is calculated as 1/1 since we |
|
116
|
|
|
|
|
|
|
have one item to allocate and only one account to allocate it to. So, 1/3 of |
|
117
|
|
|
|
|
|
|
the time C receives the benefit of the rounding error. We never end up with |
|
118
|
|
|
|
|
|
|
any unallocated items. |
|
119
|
|
|
|
|
|
|
|
|
120
|
|
|
|
|
|
|
This algorithm works for any number of weighted allocations. |
|
121
|
|
|
|
|
|
|
|
|
122
|
|
|
|
|
|
|
=over 4 |
|
123
|
|
|
|
|
|
|
|
|
124
|
|
|
|
|
|
|
=item round_fair($value, @weights) |
|
125
|
|
|
|
|
|
|
|
|
126
|
|
|
|
|
|
|
Returns a list of integer values that sum to C<$value> where each return value |
|
127
|
|
|
|
|
|
|
is a portion of C<$value> allocated by the respective weights in C<@weights>. |
|
128
|
|
|
|
|
|
|
The number of return values is equal to the number of elements in C<@weights> |
|
129
|
|
|
|
|
|
|
|
|
130
|
|
|
|
|
|
|
C<$value> must be an integer. |
|
131
|
|
|
|
|
|
|
|
|
132
|
|
|
|
|
|
|
=cut |
|
133
|
|
|
|
|
|
|
|
|
134
|
|
|
|
|
|
|
sub round_fair { |
|
135
|
52
|
|
|
52
|
1
|
268
|
my $value = shift; |
|
136
|
|
|
|
|
|
|
|
|
137
|
52
|
50
|
33
|
|
|
221
|
croak "Value to be allocated must be an integer >= 0" unless int($value) == $value && $value >= 0; |
|
138
|
|
|
|
|
|
|
|
|
139
|
52
|
50
|
|
|
|
106
|
return ($value) if @_ == 1; |
|
140
|
52
|
50
|
|
|
|
92
|
return (0) x @_ if $value == 0; |
|
141
|
|
|
|
|
|
|
|
|
142
|
52
|
|
|
|
|
50
|
my $basis = 0; |
|
143
|
52
|
|
|
|
|
76
|
for my $w ( @_ ) { |
|
144
|
214
|
50
|
|
|
|
306
|
croak "Weights must be > 0" unless $w > 0; |
|
145
|
214
|
|
|
|
|
307
|
$basis += $w; |
|
146
|
|
|
|
|
|
|
} |
|
147
|
|
|
|
|
|
|
|
|
148
|
52
|
|
|
|
|
68
|
my $sum = 0; |
|
149
|
52
|
|
|
|
|
79
|
my @in = map { my $r = $value * $_ / $basis; $sum += $r; $r } @_; |
|
|
214
|
|
|
|
|
268
|
|
|
|
214
|
|
|
|
|
202
|
|
|
|
214
|
|
|
|
|
322
|
|
|
150
|
|
|
|
|
|
|
|
|
151
|
|
|
|
|
|
|
# First, create the extra entry for the total, so that the sum of |
|
152
|
|
|
|
|
|
|
# the new array is zero. |
|
153
|
52
|
|
|
|
|
75
|
push @in, -$sum; |
|
154
|
|
|
|
|
|
|
|
|
155
|
52
|
|
|
|
|
104
|
my $out = _round_adjacent_arrayref(\@in); |
|
156
|
52
|
|
|
|
|
66
|
pop @$out; # Discard the entry for the total |
|
157
|
|
|
|
|
|
|
|
|
158
|
52
|
|
|
|
|
160
|
return @$out; |
|
159
|
|
|
|
|
|
|
} |
|
160
|
|
|
|
|
|
|
|
|
161
|
|
|
|
|
|
|
=item round_adjacent(@input_values) |
|
162
|
|
|
|
|
|
|
|
|
163
|
|
|
|
|
|
|
Returns a list of integer values, each of which is numerically |
|
164
|
|
|
|
|
|
|
adjacent to the corresponding element of @input_values, and whose total is |
|
165
|
|
|
|
|
|
|
numerically adjacent to the total of @input_values. |
|
166
|
|
|
|
|
|
|
|
|
167
|
|
|
|
|
|
|
The expected value of each output value is equal to the corresponding element |
|
168
|
|
|
|
|
|
|
of @input_values (within a small error margin due to the limited machine |
|
169
|
|
|
|
|
|
|
precision). |
|
170
|
|
|
|
|
|
|
|
|
171
|
|
|
|
|
|
|
=cut |
|
172
|
|
|
|
|
|
|
|
|
173
|
|
|
|
|
|
|
sub round_adjacent { |
|
174
|
11002
|
100
|
|
11002
|
1
|
549496
|
return () unless @_; # identity |
|
175
|
|
|
|
|
|
|
|
|
176
|
|
|
|
|
|
|
# First, create the extra entry for the total, so that the sum of |
|
177
|
|
|
|
|
|
|
# the new array is zero. |
|
178
|
11001
|
|
|
|
|
29270
|
push @_, -sum(@_); |
|
179
|
|
|
|
|
|
|
|
|
180
|
|
|
|
|
|
|
# use a reference to eliminate an unnecessary copy |
|
181
|
11001
|
|
|
|
|
20226
|
my $out = _round_adjacent_arrayref(\@_); |
|
182
|
|
|
|
|
|
|
|
|
183
|
11001
|
|
|
|
|
13537
|
pop @$out; # Discard the entry for the total |
|
184
|
11001
|
|
|
|
|
43021
|
return @$out; |
|
185
|
|
|
|
|
|
|
} |
|
186
|
|
|
|
|
|
|
|
|
187
|
|
|
|
|
|
|
sub _round_adjacent_arrayref { |
|
188
|
11053
|
|
|
11053
|
|
12493
|
my $in = shift; |
|
189
|
|
|
|
|
|
|
|
|
190
|
|
|
|
|
|
|
# Next, shuffle the order, so that the input order has no effect |
|
191
|
|
|
|
|
|
|
# on the randomness characteristics. |
|
192
|
6
|
|
|
6
|
|
5433
|
my @order = shuffle($[ .. $#{$in}); |
|
|
6
|
|
|
|
|
2884
|
|
|
|
6
|
|
|
|
|
493
|
|
|
|
11053
|
|
|
|
|
11878
|
|
|
|
11053
|
|
|
|
|
47166
|
|
|
193
|
11053
|
|
|
|
|
59562
|
@$in = map $in->[$_], @order; |
|
194
|
|
|
|
|
|
|
|
|
195
|
11053
|
|
|
|
|
22868
|
my $out = _round_adjacent_core($in); |
|
196
|
|
|
|
|
|
|
|
|
197
|
11053
|
|
|
|
|
10678
|
assert(sum(@$out) == 0); |
|
198
|
|
|
|
|
|
|
|
|
199
|
|
|
|
|
|
|
# put the output back into original order |
|
200
|
11053
|
|
|
|
|
9893
|
my @r; |
|
201
|
11053
|
|
|
|
|
84125
|
$r[$order[$_]] = $out->[$_] for $[ .. $#order; |
|
202
|
|
|
|
|
|
|
|
|
203
|
11053
|
|
|
|
|
35240
|
return \@r; |
|
204
|
|
|
|
|
|
|
} |
|
205
|
|
|
|
|
|
|
|
|
206
|
|
|
|
|
|
|
# Like _round_adjacent_arrayref, except that the inputs must sum to zero, and the |
|
207
|
|
|
|
|
|
|
# input order may affect the variance and correlations, etc. |
|
208
|
|
|
|
|
|
|
sub _round_adjacent_core { |
|
209
|
11053
|
|
|
11053
|
|
11717
|
my $in = shift; |
|
210
|
|
|
|
|
|
|
|
|
211
|
11053
|
|
|
|
|
9501
|
assert(scalar @$in); # @$in must not be empty |
|
212
|
|
|
|
|
|
|
|
|
213
|
11053
|
|
|
|
|
15382
|
my $eps1 = 4.0 * DBL_EPSILON() * (1 + @$in); |
|
214
|
11053
|
|
|
|
|
11139
|
my $eps = $eps1; |
|
215
|
11053
|
|
|
|
|
13297
|
my @fp = map { my $ip = floor($_); $_ - $ip } @$in; |
|
|
70270
|
|
|
|
|
95160
|
|
|
|
70270
|
|
|
|
|
118278
|
|
|
216
|
|
|
|
|
|
|
|
|
217
|
11053
|
|
|
|
|
12915
|
assert(none(@fp) < 0.0); |
|
218
|
|
|
|
|
|
|
|
|
219
|
|
|
|
|
|
|
# TBD: Maybe accuracy or fairness can be improved by |
|
220
|
|
|
|
|
|
|
# re-adjusting after every iteration. This would slow it |
|
221
|
|
|
|
|
|
|
# down significantly, though. |
|
222
|
11053
|
|
|
|
|
17885
|
_adjust_input(\@fp); |
|
223
|
|
|
|
|
|
|
|
|
224
|
11053
|
|
|
|
|
13313
|
my @out; |
|
225
|
11053
|
|
|
|
|
10873
|
INPUT: while() { |
|
|
11053
|
|
|
|
|
10188
|
|
|
226
|
70270
|
|
|
|
|
77023
|
$eps += $eps1; |
|
|
11053
|
|
|
|
|
22648
|
|
|
227
|
|
|
|
|
|
|
|
|
228
|
70270
|
|
|
|
|
63601
|
assert(_check_invariants($eps, $in, \@fp)); |
|
229
|
|
|
|
|
|
|
|
|
230
|
11053
|
|
|
11053
|
|
12400
|
# Calculate the next output. Discard the next input in the |
|
231
|
|
|
|
|
|
|
# process. |
|
232
|
70270
|
|
|
|
|
79956
|
my $p0 = shift @fp; # Probability of having to overpay |
|
233
|
70270
|
100
|
|
|
|
132633
|
my $r0 = rand()<$p0 ? 1 : 0; # 1 if selected to overpay; else 0 |
|
234
|
70270
|
|
|
|
|
169667
|
push @out, floor(shift @$in) + $r0; |
|
|
11053
|
|
|
|
|
19906
|
|
|
235
|
11053
|
100
|
|
|
|
37553
|
|
|
236
|
70270
|
100
|
|
|
|
130914
|
last unless @fp; |
|
|
4016
|
|
|
|
|
7229
|
|
|
237
|
|
|
|
|
|
|
|
|
238
|
4016
|
50
|
33
|
|
|
18880
|
# Now adjust the remaining fractional parts. |
|
239
|
|
|
|
|
|
|
|
|
240
|
|
|
|
|
|
|
# $slack[i] = min( $p0 * $fp[i], (1-$p0) * (1-$fp[i]) ). |
|
241
|
59217
|
|
|
|
|
61327
|
my @slack; |
|
|
4016
|
|
|
|
|
4411
|
|
|
242
|
59217
|
100
|
|
|
|
57856
|
my $tslack = 0.0; |
|
|
4016
|
|
|
|
|
6827
|
|
|
243
|
59217
|
|
|
|
|
60727
|
do { |
|
|
2616
|
|
|
|
|
14147
|
|
|
244
|
|
|
|
|
|
|
@slack = map { |
|
245
|
59217
|
|
|
|
|
67364
|
if ( 1 ) { |
|
|
502371
|
|
|
|
|
445559
|
|
|
|
1400
|
|
|
|
|
2080
|
|
|
246
|
502371
|
|
|
|
|
927402
|
my $slack = min $p0 * $_, (1 - $_) * (1.0 - $p0); |
|
|
1400
|
|
|
|
|
13129
|
|
|
247
|
502371
|
|
|
|
|
520563
|
$tslack += $slack; |
|
248
|
502371
|
|
|
|
|
726121
|
$slack; |
|
249
|
|
|
|
|
|
|
} |
|
250
|
|
|
|
|
|
|
else { |
|
251
|
|
|
|
|
|
|
# This is fewer FLOPS, but the perf benefit |
|
252
|
0
|
|
|
0
|
|
|
# is only 1% on a modern system, and it leads |
|
253
|
|
|
|
|
|
|
# to greater numerical errors for some reason. |
|
254
|
0
|
0
|
|
|
|
|
my $add = $p0 + $_; |
|
|
0
|
|
|
|
|
|
|
|
255
|
0
|
|
|
|
|
|
my $mult = $p0 * $_; |
|
256
|
0
|
|
|
|
|
|
$add > 1.0 ? 1.0 - $add + $mult : $mult |
|
257
|
0
|
|
|
|
|
|
} |
|
258
|
0
|
|
|
|
|
|
} @fp; |
|
259
|
|
|
|
|
|
|
}; |
|
260
|
|
|
|
|
|
|
|
|
261
|
0
|
|
|
|
|
|
# See bottom of file for proof of this property: |
|
262
|
59217
|
|
|
|
|
69392
|
assert($tslack + $eps >= $p0 * (1.0 - $p0)); |
|
|
0
|
|
|
|
|
|
|
|
263
|
|
|
|
|
|
|
|
|
264
|
0
|
|
|
|
|
|
# wrapped in assert to make it a noop when DEBUG() == 0 |
|
265
|
59217
|
|
|
|
|
51205
|
assert(do { warn "TSLACK = $tslack\n" if DEBUG() > 1; 1 }); |
|
266
|
|
|
|
|
|
|
|
|
267
|
59217
|
100
|
|
|
|
117893
|
if ( $tslack > $eps1 ) { |
|
268
|
37168
|
|
|
|
|
51714
|
$eps += 128.0 * $eps1 * $eps / $tslack; |
|
269
|
|
|
|
|
|
|
# NOTE: The expected value of gain is |
|
270
|
|
|
|
|
|
|
# $p0 * ($p0 - 1.0) /$tslack + |
|
271
|
|
|
|
|
|
|
# (1.0 - $p0) * $p0 / $tslack = 0 |
|
272
|
37168
|
|
|
|
|
33430
|
my $gain = do { |
|
273
|
37168
|
100
|
|
|
|
57291
|
if ( $r0 ) { |
|
274
|
|
|
|
|
|
|
# Last guy overpaid, so the probabilities for |
|
275
|
|
|
|
|
|
|
# subsequent payers drop. |
|
276
|
17549
|
|
|
|
|
23108
|
($p0 - 1.0) / $tslack; |
|
277
|
|
|
|
|
|
|
} |
|
278
|
|
|
|
|
|
|
else { |
|
279
|
|
|
|
|
|
|
# Last guy underpaid, so the probabilities for |
|
280
|
|
|
|
|
|
|
# subsequent payers rise. |
|
281
|
19619
|
|
|
|
|
23664
|
$p0 / $tslack; |
|
282
|
|
|
|
|
|
|
} |
|
283
|
|
|
|
|
|
|
}; |
|
284
|
|
|
|
|
|
|
|
|
285
|
|
|
|
|
|
|
# NOTE: The change in the sum of @fp due to this step |
|
286
|
|
|
|
|
|
|
# is $tslack * $gain, which is either $p0 or ($p0 - 1). |
|
287
|
|
|
|
|
|
|
# Either way, the sum remains an integer, because it |
|
288
|
|
|
|
|
|
|
# was reduced by $p0 when we shifted off the first |
|
289
|
|
|
|
|
|
|
# element early in the INPUT loop iteration. |
|
290
|
|
|
|
|
|
|
# Also note that each element of @fp stays in the range |
|
291
|
|
|
|
|
|
|
# [0,1] because if $r0, then slack($_, $p0) * -$gain <= |
|
292
|
|
|
|
|
|
|
# $p0 * $_ * (1.0 - $p0) / ($p0 * (1.0 - $p0)) == |
|
293
|
|
|
|
|
|
|
# $_, and otherwise slack($_, $p0) * $gain <= |
|
294
|
|
|
|
|
|
|
# (1 - $p0) * (1 - $_) * $p0 / ($p0 * (1.0 - $p0)) == |
|
295
|
|
|
|
|
|
|
# 1 - $_. |
|
296
|
|
|
|
|
|
|
# We modify in place here, for performance. |
|
297
|
37168
|
|
|
|
|
225908
|
$_ += shift(@slack) * $gain for @fp; |
|
298
|
|
|
|
|
|
|
} |
|
299
|
|
|
|
|
|
|
} |
|
300
|
|
|
|
|
|
|
assert(@$in == 0); |
|
301
|
|
|
|
|
|
|
return \@out; |
|
302
|
|
|
|
|
|
|
} |
|
303
|
|
|
|
|
|
|
|
|
304
|
|
|
|
|
|
|
sub _adjust_input { |
|
305
|
|
|
|
|
|
|
my $p = shift; |
|
306
|
|
|
|
|
|
|
|
|
307
|
|
|
|
|
|
|
# Adjust @$p to account for numerical errors due to small |
|
308
|
|
|
|
|
|
|
# difference of large numbers when the integer parts are big. |
|
309
|
|
|
|
|
|
|
my $sum = sum @$p; |
|
310
|
|
|
|
|
|
|
if ( $sum != floor($sum) ) { |
|
311
|
|
|
|
|
|
|
my $target = floor($sum + 0.5); |
|
312
|
|
|
|
|
|
|
|
|
313
|
|
|
|
|
|
|
die "Total loss of precision" |
|
314
|
|
|
|
|
|
|
unless abs($sum - $target) < 0.1 && $sum + 0.05 != $sum; |
|
315
|
|
|
|
|
|
|
|
|
316
|
|
|
|
|
|
|
my $adj = $target / $sum; |
|
317
|
|
|
|
|
|
|
if ( $adj <= 1.0 ) { |
|
318
|
|
|
|
|
|
|
$_ *= $adj for @$p; |
|
319
|
|
|
|
|
|
|
} else { |
|
320
|
|
|
|
|
|
|
$adj = (@$p - $target) / (@$p - $sum); |
|
321
|
|
|
|
|
|
|
$_ = 1.0 - (1.0-$_) * $adj for @$p; |
|
322
|
|
|
|
|
|
|
} |
|
323
|
|
|
|
|
|
|
} |
|
324
|
|
|
|
|
|
|
} |
|
325
|
|
|
|
|
|
|
|
|
326
|
|
|
|
|
|
|
sub _check_invariants { |
|
327
|
|
|
|
|
|
|
my ( $eps, $v, $fp ) = @_; |
|
328
|
|
|
|
|
|
|
|
|
329
|
|
|
|
|
|
|
if ( DEBUG() > 1 ) { |
|
330
|
|
|
|
|
|
|
warn sprintf "%d %f\n", floor($_), $_ for @$fp; |
|
331
|
|
|
|
|
|
|
} |
|
332
|
|
|
|
|
|
|
|
|
333
|
|
|
|
|
|
|
assert(@$v && @$v == @$fp); |
|
334
|
|
|
|
|
|
|
|
|
335
|
|
|
|
|
|
|
for ( @$fp ) { |
|
336
|
|
|
|
|
|
|
assert($_ >= -$eps); |
|
337
|
|
|
|
|
|
|
assert($_ <= 1.0 + $eps); |
|
338
|
|
|
|
|
|
|
} |
|
339
|
|
|
|
|
|
|
|
|
340
|
|
|
|
|
|
|
my $sum = sum @$fp; |
|
341
|
|
|
|
|
|
|
assert(abs($sum - floor($sum + 0.5)) < $eps * (1 + $sum)); |
|
342
|
|
|
|
|
|
|
|
|
343
|
|
|
|
|
|
|
1; |
|
344
|
|
|
|
|
|
|
} |
|
345
|
|
|
|
|
|
|
|
|
346
|
|
|
|
|
|
|
1; |
|
347
|
|
|
|
|
|
|
|
|
348
|
|
|
|
|
|
|
__END__ |