forked from johannesgerer/jburkardt-f
-
Notifications
You must be signed in to change notification settings - Fork 1
/
kmeans.html
677 lines (617 loc) · 19 KB
/
kmeans.html
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
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
<html>
<head>
<title>
KMEANS - the K-Means Data Clustering Problem
</title>
</head>
<body bgcolor="#EEEEEE" link="#CC0000" alink="#FF3300" vlink="#000055">
<h1 align = "center">
KMEANS <br> the K-Means Data Clustering Problem
</h1>
<hr>
<p>
<b>KMEANS</b>
is a FORTRAN90 library which
handles the K-Means problem,
which organizes a set of N points in M dimensions into K clusters;
</p>
<p>
In the K-Means problem, a set of N points X(I) in M-dimensions is
given. The goal is to arrange these points into K clusters,
with each cluster having a representative point Z(J), usually
chosen as the centroid of the points in the cluster. <pre>
Z(J) = Sum ( all X(I) in cluster J ) X(I) /
Sum ( all X(I) in cluster J ) 1.
</pre>
The energy of cluster J is <pre>
E(J) = Sum ( all X(I) in cluster J ) || X(I) - Z(J) ||^2
</pre>
</p>
<p>
For a given set of clusters, the total energy is then simply
the sum of the cluster energies E(J). The goal is to choose the
clusters in such a way that the total energy is minimized.
Usually, a point X(I) goes into the cluster with the closest
representative point Z(J). So to define the clusters, it's
enough simply to specify the locations of the cluster representatives.
</p>
<p>
This is actually a fairly hard problem. Most algorithms do
reasonably well, but cannot guarantee that the best solution
has been found. It is very common for algorithms to get
stuck at a solution which is merely a "local minimum".
For such a local minimum, every slight rearrangement of
the solution makes the energy go up; however a major
rearrangement would result in a big drop in energy.
</p>
<p>
A simple algorithm for the problem is known as the "H-Means algorithm".
It alternates between two procedures:
<ul>
<li>
Using the given cluster centers, assign each point to the
cluster with the nearest center;
</li>
<li>
Using the given cluster assignments, replace each cluster
center by the centroid or average of the points in the cluster.
</li>
</ul>
These steps are repeated until no points are moved, or some
other termination criterion is reached.
</p>
<p>
A more sophisticated algorithm, known as the "K-Means algorithm",
takes advantage of the fact that it is possible to quickly determine
the decrease in energy caused by moving a point from its current cluster
to another. It repeats the following procedure:
<ul>
<li>
For each point, move it to another cluster if that would lower
the energy. If you move a point, immediately update the
cluster centers of the two affected clusters.
</li>
</ul>
This procedure is repeated until no points are moved, or some
other termination criterion is reached.
</p>
<h3 align = "center">
The Weighted K-Means Problem
</h3>
<p>
A natural extension of the K-Means problem allows us to include
some more information, namely, a set of <i>weights</i> associated
with the data points. These might represent a measure of importance,
a frequency count, or some other information. The intent is that
a point with a weight of 5.0 is twice as "important" as a point with
a weight of 2.5, for instance. This gives rise to the "weighted"
K-Means problem.
</p>
<p>
In the <i>weighted K-Means problem</i>, we are given a set of N points
X(I) in M-dimensions, and a corresponding set of nonnegative weights
W(I). The goal is to arrange the points into K clusters,
with each cluster having a representative point Z(J), usually
chosen as the weighted centroid of the points in the cluster: <pre>
Z(J) = Sum ( all X(I) in cluster J ) W(I) * X(I) /
Sum ( all X(I) in cluster J ) W(I).
</pre>
The weighted energy of cluster J is <pre>
E(J) = Sum ( all X(I) in cluster J ) W(I) * || X(I) - Z(J) ||^2
</pre>
</p>
<h3 align = "center">
Licensing:
</h3>
<p>
The computer code and data files described and made available on this web page
are distributed under
<a href = "../../txt/gnu_lgpl.txt">the GNU LGPL license.</a>
</p>
<h3 align = "center">
Languages:
</h3>
<p>
<b>KMEANS</b> is available in
<a href = "../../cpp_src/kmeans/kmeans.html">a C++ version</a> and
<a href = "../../f_src/kmeans/kmeans.html">a FORTRAN90 version</a> and
<a href = "../../m_src/kmeans/kmeans.html">a MATLAB version.</a>
</p>
<h3 align = "center">
Related Data and Programs:
</h3>
<p>
<a href = "../../f_src/asa058/asa058.html">
ASA058</a>,
a FORTRAN90 library which
implements the K-means algorithm of Sparks.
</p>
<p>
<a href = "../../f_src/asa136/asa136.html">
ASA136</a>,
a FORTRAN90 library which
implements the Hartigan and Wong clustering algorithm.
</p>
<p>
<a href = "../../f_src/cities/cities.html">
CITIES</a>,
a FORTRAN90 library which
handles various problems associated with a set of "cities" on a map.
</p>
<p>
<a href = "../../datasets/cities/cities.html">
CITIES</a>,
a dataset directory which
contains sets of data defining groups of cities.
</p>
<p>
<a href = "../../f_src/cluster_energy/cluster_energy.html">
CLUSTER_ENERGY</a>,
a FORTRAN90 program which
groups data into a given number of clusters to minimize the energy.
</p>
<p>
<a href = "../../f_src/lau_np/lau_np.html">
LAU_NP</a>,
a FORTRAN90 library which
contains heuristic algorithms for the K-center and K-median problems.
</p>
<p>
<a href = "../../f_src/point_merge/point_merge.html">
POINT_MERGE</a>,
a FORTRAN90 library which
considers N points in M dimensional space, and counts or indexes
the unique or "tolerably unique" items.
</p>
<p>
<a href = "../../f_src/spaeth/spaeth.html">
SPAETH</a>,
a FORTRAN90 library which
can cluster data according to various principles.
</p>
<p>
<a href = "../../datasets/spaeth/spaeth.html">
SPAETH</a>,
a dataset directory which
contains a set of test data.
</p>
<p>
<a href = "../../f_src/spaeth2/spaeth2.html">
SPAETH2</a>,
a FORTRAN90 library which
can cluster data according to various principles.
</p>
<p>
<a href = "../../datasets/spaeth2/spaeth2.html">
SPAETH2</a>,
a dataset directory which
contains a set of test data.
</p>
<h3 align = "center">
Reference:
</h3>
<p>
<ol>
<li>
John Hartigan, Manchek Wong,<br>
Algorithm AS 136:
A K-Means Clustering Algorithm,<br>
Applied Statistics,<br>
Volume 28, Number 1, 1979, pages 100-108.
</li>
<li>
Wendy Martinez, Angel Martinez,<br>
Computational Statistics Handbook with MATLAB,<br>
Chapman and Hall / CRC, 2002.
</li>
<li>
David Sparks,<br>
Algorithm AS 58:
Euclidean Cluster Analysis,<br>
Applied Statistics,<br>
Volume 22, Number 1, 1973, pages 126-130.
</li>
</ol>
</p>
<h3 align = "center">
Source Code:
</h3>
<p>
<ul>
<li>
<a href = "kmeans.f90">kmeans.f90</a>, the source code.
</li>
<li>
<a href = "kmeans.sh">kmeans.sh</a>,
commands to compile the source code.
</li>
</ul>
</p>
<h3 align = "center">
Examples and Tests:
</h3>
<p>
<ul>
<li>
<a href = "kmeans_prb.f90">kmeans_prb.f90</a>, a sample problem.
</li>
<li>
<a href = "kmeans_prb.sh">kmeans_prb.sh</a>,
commands to compile, link and run the sample program.
</li>
<li>
<a href = "kmeans_prb_output.txt">kmeans_prb_output.txt</a>,
the output file.
</li>
</ul>
</p>
<p>
There are data files read by the sample code:
<ul>
<li>
<a href = "points_100.txt">points_100.txt</a>, 100 2D points,
used as a case study by the sample problem.
</li>
<li>
<a href = "points_100.png">points_100.png</a>,
an image of the points.
</li>
<li>
<a href = "ruspini.txt">ruspini.txt</a>, 75 points in 2D,
with integer coordinates, and 0 < X < 120, 0 < Y < 160,
which might naturally be grouped into 4 sets.
</li>
<li>
<a href = "weights_equal_100.txt">weights_equal_100.txt</a>,
100 equal weights, for doing equally weighted calculations when
a program expects weights.
</li>
<li>
<a href = "weights_unequal_100.txt">weights_unequal_100.txt</a>,
100 weights, not all equal, for testing programs that can use weights.
</li>
</ul>
</p>
<p>
<b>TEST01</b> applies HMEANS_01 to points_100.txt:
<ul>
<li>
<a href = "test01_clusters.txt">test01_clusters.txt</a>
the cluster assignments.
</li>
<li>
<a href = "test01_centers.txt">test01_centers.txt</a>
the cluster centers.
</li>
</ul>
</p>
<p>
<b>TEST02</b> applies HMEANS_02 to points_100.txt:
<ul>
<li>
<a href = "test02_clusters.txt">test02_clusters.txt</a>
the cluster assignments.
</li>
<li>
<a href = "test02_centers.txt">test02_centers.txt</a>
the cluster centers.
</li>
</ul>
</p>
<p>
<b>TEST03</b> applies KMEANS_01 to points_100.txt:
<ul>
<li>
<a href = "test03_clusters.txt">test03_clusters.txt</a>
the cluster assignments.
</li>
<li>
<a href = "test03_centers.txt">test03_centers.txt</a>
the cluster centers.
</li>
</ul>
</p>
<p>
<b>TEST04</b> applies KMEANS_02 to points_100.txt:
<ul>
<li>
<a href = "test04_clusters.txt">test04_clusters.txt</a>
the cluster assignments.
</li>
<li>
<a href = "test04_centers.txt">test04_centers.txt</a>
the cluster centers.
</li>
</ul>
</p>
<p>
<b>TEST05</b> applies KMEANS_03 to points_100.txt:
<ul>
<li>
<a href = "test05_clusters.txt">test05_clusters.txt</a>
the cluster assignments.
</li>
<li>
<a href = "test05_centers.txt">test05_centers.txt</a>
the cluster centers.
</li>
</ul>
</p>
<p>
<b>TEST06</b> applies HMEANS_01 + KMEANS_01 to points_100.txt:
<ul>
<li>
<a href = "test06_clusters.txt">test06_clusters.txt</a>
the cluster assignments.
</li>
<li>
<a href = "test06_centers.txt">test06_centers.txt</a>
the cluster centers.
</li>
</ul>
</p>
<p>
<b>TEST07</b> applies HMEANS_01 + KMEANS_02 to points_100.txt:
<ul>
<li>
<a href = "test07_clusters.txt">test07_clusters.txt</a>
the cluster assignments.
</li>
<li>
<a href = "test07_centers.txt">test07_centers.txt</a>
the cluster centers.
</li>
</ul>
</p>
<p>
<b>TEST08</b> applies KMEANS_01 + KMEANS_03 to points_100.txt:
<ul>
<li>
<a href = "test08_clusters.txt">test08_clusters.txt</a>
the cluster assignments.
</li>
<li>
<a href = "test08_centers.txt">test08_centers.txt</a>
the cluster centers.
</li>
</ul>
</p>
<p>
<b>TEST09</b> applies HMEANS_W_01 to points_100.txt and weights_equal_100.txt:
<ul>
<li>
<a href = "test09_clusters.txt">test09_clusters.txt</a>
the cluster assignments.
</li>
<li>
<a href = "test09_centers.txt">test09_centers.txt</a>
the cluster centers.
</li>
</ul>
</p>
<p>
<b>TEST10</b> applies HMEANS_W_02 to points_100.txt and weights_equal_100.txt:
<ul>
<li>
<a href = "test10_clusters.txt">test10_clusters.txt</a>
the cluster assignments.
</li>
<li>
<a href = "test10_centers.txt">test10_centers.txt</a>
the cluster centers.
</li>
</ul>
</p>
<p>
<b>TEST11</b> applies KMEANS_W_01 to points_100.txt and weights_equal_100.txt:
<ul>
<li>
<a href = "test11_clusters.txt">test11_clusters.txt</a>
the cluster assignments.
</li>
<li>
<a href = "test11_centers.txt">test11_centers.txt</a>
the cluster centers.
</li>
</ul>
</p>
<p>
<b>TEST12</b> applies KMEANS_W_03 to points_100.txt and weights_equal_100.txt:
<ul>
<li>
<a href = "test12_clusters.txt">test12_clusters.txt</a>
the cluster assignments.
</li>
<li>
<a href = "test12_centers.txt">test12_centers.txt</a>
the cluster centers.
</li>
</ul>
</p>
<p>
<b>TEST13</b> applies HMEANS_W_01 to points_100.txt and weights_unequal_100.txt:
<ul>
<li>
<a href = "test13_clusters.txt">test13_clusters.txt</a>
the cluster assignments.
</li>
<li>
<a href = "test13_centers.txt">test13_centers.txt</a>
the cluster centers.
</li>
</ul>
</p>
<p>
<b>TEST14</b> applies HMEANS_W_02 to points_100.txt and weights_unequal_100.txt:
<ul>
<li>
<a href = "test14_clusters.txt">test14_clusters.txt</a>
the cluster assignments.
</li>
<li>
<a href = "test14_centers.txt">test14_centers.txt</a>
the cluster centers.
</li>
</ul>
</p>
<p>
<b>TEST15</b> applies KMEANS_W_01 to points_100.txt and weights_unequal_100.txt:
<ul>
<li>
<a href = "test15_clusters.txt">test15_clusters.txt</a>
the cluster assignments.
</li>
<li>
<a href = "test15_centers.txt">test15_centers.txt</a>
the cluster centers.
</li>
</ul>
</p>
<p>
<b>TEST16</b> applies KMEANS_W_03 to points_100.txt and weights_unequal_100.txt:
<ul>
<li>
<a href = "test16_clusters.txt">test16_clusters.txt</a>
the cluster assignments.
</li>
<li>
<a href = "test16_centers.txt">test16_centers.txt</a>
the cluster centers.
</li>
</ul>
</p>
<h3 align = "center">
List of Routines:
</h3>
<p>
<ul>
<li>
<b>CH_CAP</b> capitalizes a single character.
</li>
<li>
<b>CH_EQI</b> is a case insensitive comparison of two characters for equality.
</li>
<li>
<b>CH_TO_DIGIT</b> returns the integer value of a base 10 digit.
</li>
<li>
<b>CLUSTER_ENERGY_COMPUTE</b> computes the energy of the clusters.
</li>
<li>
<b>CLUSTER_INITIALIZE_1</b> initializes the clusters to data points.
</li>
<li>
<b>CLUSTER_INITIALIZE_2</b> initializes the cluster centers to random values.
</li>
<li>
<b>CLUSTER_INITIALIZE_3</b> initializes the cluster centers to random values.
</li>
<li>
<b>CLUSTER_INITIALIZE_4</b> initializes the cluster centers to random values.
</li>
<li>
<b>CLUSTER_INITIALIZE_5</b> initializes the cluster centers to random values.
</li>
<li>
<b>CLUSTER_PRINT_SUMMARY</b> prints a summary of data about a clustering.
</li>
<li>
<b>CLUSTER_VARIANCE_COMPUTE</b> computes the variance of the clusters.
</li>
<li>
<b>FILE_COLUMN_COUNT</b> counts the number of columns in the first line of a file.
</li>
<li>
<b>FILE_ROW_COUNT</b> counts the number of row records in a file.
</li>
<li>
<b>GET_UNIT</b> returns a free FORTRAN unit number.
</li>
<li>
<b>HMEANS_01</b> applies the H-Means algorithm.
</li>
<li>
<b>HMEANS_02</b> applies the H-Means algorithm.
</li>
<li>
<b>HMEANS_W_01</b> applies the weighted H-Means algorithm.
</li>
<li>
<b>HMEANS_W_02</b> applies the weighted H-Means algorithm.
</li>
<li>
<b>I4_UNIFORM</b> returns a scaled pseudorandom I4.
</li>
<li>
<b>I4MAT_WRITE</b> writes an I4MAT file.
</li>
<li>
<b>KMEANS_01</b> applies the K-Means algorithm.
</li>
<li>
<b>KMEANS_02</b> applies the K-Means algorithm.
</li>
<li>
<b>KMEANS_02_OPTRA</b> carries out the optimal transfer stage.
</li>
<li>
<b>KMEANS_02_QTRAN</b> carries out the quick transfer stage.
</li>
<li>
<b>KMEANS_03</b> applies the K-Means algorithm.
</li>
<li>
<b>KMEANS_W_01</b> applies the weighted K-Means algorithm.
</li>
<li>
<b>KMEANS_W_03</b> applies the weighted K-Means algorithm.
</li>
<li>
<b>R4_UNIFORM_01</b> returns a unit pseudorandom R4.
</li>
<li>
<b>R8_UNIFORM_01</b> returns a unit pseudorandom R8.
</li>
<li>
<b>R8MAT_DATA_READ</b> reads data from an R8MAT file.
</li>
<li>
<b>R8MAT_HEADER_READ</b> reads the header from an R8MAT file.
</li>
<li>
<b>R8MAT_UNIFORM_01</b> returns a unit pseudorandom R8MAT.
</li>
<li>
<b>R8MAT_WRITE</b> writes an R8MAT file.
</li>
<li>
<b>R8VEC_UNIFORM_01</b> returns a unit pseudorandom R8VEC.
</li>
<li>
<b>RANDOM_INITIALIZE</b> initializes the FORTRAN90 random number seed.
</li>
<li>
<b>S_TO_R8</b> reads an R8 from a string.
</li>
<li>
<b>S_TO_R8VEC</b> reads an R8VEC from a string.
</li>
<li>
<b>S_WORD_COUNT</b> counts the number of "words" in a string.
</li>
<li>
<b>TIMESTAMP</b> prints the current YMDHMS date as a time stamp.
</li>
</ul>
</p>
<p>
You can go up one level to <a href = "../f_src.html">
the FORTRAN90 source codes</a>.
</p>
<hr>
<i>
Last revised on 15 October 2009.
</i>
<!-- John Burkardt -->
</body>
</html>