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
|
function comparisonData = compareAll(trials)
if (nargin < 1 || trials < 1)
trials = 5;
end
try
load('mongoose_data.mat');
catch fileNotFound
save('mongoose_data.mat');
comparisonData = {};
lastMatrixCompleted = 0;
j = 1;
end
index = ssget;
% Sort by nnz
nnzs = index.nnz;
[~,ids] = sortrows(nnzs');
for i = ids'
% Skip specific problematic matrices
if (i == 1772 || i == 2177 || i == 2249)
continue;
end
found = ID_present(comparisonData,i);
if (index.isReal(i) && ~found)
Prob = ssget(i);
A = Prob.A;
fprintf('Computing separator for %d: %s\n', i, Prob.name);
[~, n_cols] = size(A);
% If matrix is unsymmetric, form the augmented system
if (index.numerical_symmetry(i) < 1)
[m_rows, n_cols] = size(A);
A = [sparse(m_rows,m_rows) A; A' sparse(n_cols,n_cols)];
end
for use_weights = 0:1
% Sanitize the matrix (remove diagonal, make symmetric)
A = sanitize(A, ~use_weights);
% If the sanitization removed all vertices, skip this matrix
if nnz(A) < 2
comparisonData(j).problem_id = Prob.id;
j = j+1;
continue;
end
% Run Mongoose with various options to partition the graph.
for guessCutType = 0:2
for doCommunityMatching = 0:1
for matchingStrategy = 3:-1:0
% Community matching does not affect matching
% strategies Random or HEM
if (doCommunityMatching == 1 && (matchingStrategy == 0 || matchingStrategy == 1))
continue;
end
% if highest degree > 10*sqrt(n), skip Random/HEM
if (max(sum(sign(A))) > 10*sqrt(n_cols))
if (matchingStrategy == 0 || matchingStrategy == 1)
continue;
end
end
for coarsenLimit = [64, 256, 1024]
comparisonData(j).mongoose = 1;
comparisonData(j).problem_id = Prob.id;
comparisonData(j).problem_name = Prob.name;
comparisonData(j).problem_kind = Prob.kind;
comparisonData(j).problem_nnz = nnz(A);
comparisonData(j).problem_n = n_cols;
comparisonData(j).useWeights = use_weights;
comparisonData(j).guessCutType = guessCutType;
comparisonData(j).doCommunityMatching = doCommunityMatching;
comparisonData(j).matchingStrategy = matchingStrategy;
comparisonData(j).coarsenLimit = coarsenLimit;
fprintf('name = %s\n', Prob.name);
fprintf('use_weights = %d\n', use_weights);
fprintf('guessCutType = %d\n', guessCutType);
fprintf('doCommunityMatching = %d\n', doCommunityMatching);
fprintf('matchingStrategy = %d\n', matchingStrategy);
fprintf('coarsenLimit = %d\n', coarsenLimit);
for k = 1:trials
% Set up options struct for this run
O = edgecut_options();
O.randomSeed = 123456789;
O.guessCutType = guessCutType;
O.doCommunityMatching = doCommunityMatching;
O.matchingStrategy = matchingStrategy;
O.coarsenLimit = coarsenLimit;
tic;
partition = edgecut(A,O);
t = toc;
fprintf('Mongoose: %0.2f\n', t);
mongoose_times(j, k) = t;
part_A = find(partition);
part_B = find(1-partition);
perm = [part_A part_B];
p = length(part_A);
A_perm = A(perm, perm);
mongoose_cut_weight(j, k) = sum(sum(A_perm((p+1):n_cols, 1:p)));
mongoose_cut_size(j, k) = sum(sum(sign(A_perm((p+1):n_cols, 1:p))));
mongoose_imbalance(j, k) = abs(0.5-(length(part_A)/(length(part_A) + length(part_B))));
% If it took more than 30 minutes, only
% run once.
if (t > 1800)
break;
end
end
comparisonData(j).time = trimmean(mongoose_times(j, 1:k), 40);
comparisonData(j).cutWeight = trimmean(mongoose_cut_weight(j, 1:k), 40);
comparisonData(j).cutSize = trimmean(mongoose_cut_size(j, 1:k), 40);
comparisonData(j).cutImbalance = trimmean(mongoose_imbalance(j, 1:k), 40);
j = j+1;
end
end
end
end
% Run METIS to partition the graph.
for k = 1:trials
tic;
[part_A,part_B] = metispart(A, 0, 123456789);
t = toc;
fprintf('METIS: %0.2f\n', t);
metis_times(j, k) = t;
p = length(part_A);
perm = [part_A part_B];
A_perm = A(perm, perm);
metis_cut_weight(j, k) = sum(sum(A_perm((p+1):n_cols, 1:p)));
metis_cut_size(j, k) = sum(sum(sign(A_perm((p+1):n_cols, 1:p))));
metis_imbalance(j, k) = abs(0.5-(length(part_A)/(length(part_A) + length(part_B))));
end
comparisonData(j).problem_id = Prob.id;
comparisonData(j).problem_name = Prob.name;
comparisonData(j).problem_kind = Prob.kind;
comparisonData(j).problem_nnz = nnz(A);
comparisonData(j).problem_n = n_cols;
comparisonData(j).useWeights = use_weights;
comparisonData(j).mongoose = 0;
comparisonData(j).time = trimmean(metis_times(j, 1:k), 40);
comparisonData(j).cutWeight = trimmean(metis_cut_weight(j, 1:k), 40);
comparisonData(j).cutSize = trimmean(metis_cut_size(j, 1:k), 40);
comparisonData(j).cutImbalance = trimmean(metis_imbalance(j, 1:k), 40);
j = j+1;
end
end
lastMatrixCompleted = i;
save('mongoose_data.mat');
end
% Write data to file for future comparisons
writetable(struct2table(comparisonData), 'mongoose_data.csv');
end
function found = ID_present(comparisonData, id)
found = 0;
for i = 1:length(comparisonData)
if (comparisonData(i).problem_id == id)
found = 1;
return;
end
end
end
|
