-
Notifications
You must be signed in to change notification settings - Fork 6
/
Copy pathfast_protein_cluster.cpp
1093 lines (1054 loc) · 54 KB
/
fast_protein_cluster.cpp
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
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
#ifdef GPU
#define __CL_ENABLE_EXCEPTIONS
#include <CL/cl.hpp>
#endif
#ifdef OPENMP
#include <omp.h>
#endif
#include <time.h>
#include "lite.h"
#include "error_handlers.h"
#include <iostream>
#include "libcluster.hpp"
#ifdef GPU
char gtmscorecl_location[LINE_LENGTH], grmsdcl_location[LINE_LENGTH]; //path to tmscore.cl and rmsd.cl kernels
#endif
double gtimer1=0;
double gtimer2=0;
cpu_properties gcpu_info;
//main function parses options
//prune and cluster functions that do the work - templated to allow for compact representation of data in either or both steps
template <class T,class T2>
cluster_models_set<T2>* prune_it(int nthreads, prune_options *prune_options,cluster_options *cluster_options,unsigned int seed);
template <class T>
void cluster_it(int nthreads,cluster_options *cluster_options,cluster_models_set<T> *models,unsigned int seed);
double get_time();
//global cpu_properties
//needed to prevent OPENMP feature which causes slowdown with hyperthreading
//really should be part of classes calling hierarchical clustering for portablility
int main(int argc, char *argv[]){
const char *options_text = "Protinfo_cluster -i <input_file with decoy files or names> -o <output_file basename> --<options>\n\noptions (can be all upper case or all lower case)\n\nCompute_type\n--cpu CPU is used for computations - (default)\n--prune_cpu CPU is used for pruning computations\n--cluster_cpu CPU is used for cluster computations\n--gpu GPU is used for computations - this is the default\n--prune_gpu GPU is used for pruning computations\n--cluster_gpu GPU is used for cluster computations\n\nScore_type\n--rmsd RMSD is used as the similarity metric (default)\n--prune_rmsd RMSD is used as the similarity metric for pruning\n--cluster_rmsd RMSD is used as the similarity metric for clustering\n--tmscore TMSCORE is used as the similarity metric\n--prune_tmscore TMSCORE is used as the similarity metric for pruning\n--cluster_tmscore TMSCORE is used as the similarity metric for clustering\n\nsimd_type\n--scalar normal scalar operations will be used (default)\n--sse2 SSE2 vectorization will be used when possible (Pentium 4 or newer, AMDK8 or newer)\n--sse3 SSE3 vectorization will be used when possible (Pentium 4 Prescott or newer, AMD Athlon 64 or newer)\n--avx AVX vectorization will be used when possible (Sandy Bridge, Ivy Bridge, AMD Bulldozer/Piledriver)\n\ndistance_matrix_storage\n--float Use float (single precision 4 bytes) to represent distances (default)\n--prune_float Use float (single precision 4 bytes) to represent distances in pruning step\n--cluster_float Use float (single precision 4 bytes) to represent distances in cluster step\n--compact Use unsigned char (1 byte) to represent distances\n--prune_compact Use unsigned char (1 byte) to represent distances in prune step\n--cluster_compact Use unsigned char (1 byte) to represent distances in cluster step\n\nInput/output \n\nwrite matrix in binary form for clustering\n--write_matrix <matrix_file> \n--write_binary_matrix <matrix_file> \n--cluster_write_matrix <matrix_file> \n--cluster_write_binary_matrix <matrix_file> \n\nwrite distance matrix in binary form for pruning\n--PRUNE_WRITE_MATRIX <matrix_file> --prune_write_matrix <matrix_file> \n--PRUNE_WRITE_BINARY_MATRIX <matrix_file> --prune_write_binary_matrix <matrix_file> \n\nwrite distance matrix in text form\n--cluster_write_text_matrix <matrix_file> \n--prune_write_text_matrix <matrix_file> \n\nwrite distance matrix in compact form \n--cluster_write_compact_matrix <matrix_file> \n--prune_write_compact_matrix <matrix_file> \n\nwrite cluster distance matrix in binary form\n--write_matrix <matrix_file> \n--write_binary_matrix <matrix_file> \n--cluster_write_matrix <matrix_file> \n--cluster_write_binary_matrix <matrix_file> \n\nwrite prune distance matrix in binary form \n--prune_write_matrix <matrix_file> \n--prune_write_binary_matrix <matrix_file>\n\ninput binary coordinates instead pdb_list (application will look for <input_file>.coords and <input_file>.names\n--binary_coords --BINARY_COORDS\n--cluster_binary_coords --CLUSTER_BINARY_COORDS\n--prune_binary_coords --PRUNE_BINARY_COORDS\n\n\nCluster methods\n--nclusters number of clusters\n--prune_nclusters number of clusters to be used in pruning\n\nHierarchical\n--hcomplete complete linkage hierarchical clustering (max distance between clusters used)\n--haverage average linkage hierarchical clustering (average distance between clusters used)\n--hsingle single linkage hierarchical clustering (min distance between clusters used)\n\n--cluster_hcomplete clustering step: complete linkage hierarchical clustering (max distance between clusters used)\n--cluster_haverage clustering step: average linkage hierarchical clustering (average distance between clusters used)\n--cluster_hsingle clustering step: single linkage hierarchical clustering (min distance between clusters used)\n\n--prune_hcomplete prune using average distance to centers of clusters from complete linkage hierarchical clustering\n--prune_haverage prune using average distance to centers of clusters from average linkage hierarchical clustering\n--prune_hsingle prune using average distance to centers of clusters from single linkage hierarchical clustering\n\nkmeans \n--kmeans --cluster_kmeans use kmeans for clustering step (default) -cannot use for pruning\n--min_cluster_size <n> minimum cluster size (kmeans only)\n--max_iterations <n> max iterations for each random starting partition to converge to a final partition\n--total_seeds <n> number of different starting partitions to try\n--converge_seeds <n> number of starting partitions without improvement kmeans is said to have converged \n--percentile <P> used to calculate when the partition score is in the P percentile with confidence pvalue p\n--pvalue <p> P is a float value between 0-1 (higher is better value) and p is a positive float value\n--fixed centers <n> the final n clusters of starting partition are not random but the most distant from the other clusters\n--fine_parallel uses finer level of parallelism for kmeans rather than the default seed level parallelism - useful for large sets and small numbers of seeds\n\ndensity\n--density --cluster_density treat ensemble as single cluster and use min average distance to other structures to find center\n--prune_density use single cluster density as criteron for pruning (default) \n--density_only calculate density only - does not create similarity matrix so it is very memory efficient O(n)\n--sort_density sort the densities in the density_only mode\n\nkcenters\n--kcenters --cluster_kcenters use kcenters algorithm for clustering\n--prune_kcenters use distance from centers of kcenters to prune\n\npruning options\n--prune_until_size<size> only option for all but density based pruning - prunes until the size of the ensemble is <size>\n--prune_stop_at_size differs from prune_until_size by defining a minimum ensemble size regardless of other criteria\n--prune_outlier_ratio <r> instead of removing the worst structure at each step use a ratio i.e. a ratio of .9 means that 99% best structures are kept after each step\n--prune_to_density <density> prune until a threshhold average density <density> is reached\n--prune_stop_density <density> stop pruning when average density reaches <density>\n--prune_zmode used in conjunction with --prune_to_density/--prune_stop_density for stopping criterion\n instead of <density> absolute values of density - <density> is number of standard deviations from mean\ngpu_options\n--gpu_id <id> specifies gpu whem there is more than one available\n--prune_gpu_id <id> specifies gpu whem there is more than one available\n--cluster_gpu_id <id> specifies gpu whem there is more than one available\n\nmisc\n-i <filename> file with list of PDBs in normal (PDB_LIST) mode\n file with list of names when similarity matrix is being read in\n file with basename for binary_coords mode - program expects files <file>.coords <file>.names\n-o <output_basename> default is 'output'\n-S <seed> define integer seed to be used by random number generators\n-p <path> define the path to the tmscore.cl and rmsd.cl kernels\n--help output this text\n";
cluster_options cluster_options;
prune_options prune_options;
int
//acceleration variables
nthreads=1, //number of preferred threads -does not override system maxium thread settings
prune_flag=0,
density_only_flag=0, //simple density calculation does not require similarity matrix if nothing else is being done
all_atoms_flag=0, //all-atoms for RMSD calculation - when TMScore is to be calculated in prune or cluster step
//this is set back to 0 - change int the future to switch back and forth between all-atom and CA - already in reference code
srand (time(NULL)); //random number is used to seed the kmeans - might use better algorithm for parallel
bool sort_density=false; //in density_only_mode - determines whether the densities should be sorted
unsigned int seed=time(NULL);
double start_time=get_time();
//tell us what compile options are enabled
#ifdef SSE2
fprintf(stderr,"SSE2 enabled\n");
#endif
#ifdef SSE3
fprintf(stderr,"SSE3 enabled\n");
#endif
#ifdef AVX
fprintf(stderr,"AVX enabled\n");
#endif
#ifdef GPU
fprintf(stderr,"GPU enabled\n");
#ifdef NVIDIA
sprintf(gtmscorecl_location,"/usr/local/cuda/include/tmscore.cl");
sprintf(grmsdcl_location,"/usr/local/cuda/include/rmsd.cl");
#endif
#ifdef AMD
sprintf(gtmscorecl_location,"/opt/AMDAPP/include/tmscore.cl");
sprintf(grmsdcl_location,"/opt/AMDAPP/include/rmsd.cl");
#endif
#endif
#ifdef FAST_DIVISION
fprintf(stderr,"FAST_DIVISION enabled\n");
#endif
#ifdef OPEN_MP
omp_set_dynamic(1);
#endif
//options block
{
int i=1;
while(i<argc)
{
if(argv[i][0] == '-' && argv[i][1] != '\0')
{
char option=argv[i][1];
char switch_word[LINE_LENGTH];
//these have a --
switch(option)
{
case '-':
{
strcpy(switch_word,&(argv[i][2]));
if(!strcmp(switch_word,"nthreads") || !strcmp(switch_word,"NTHREADS"))
{
i++;
nthreads=atoi(argv[i]);
}
//compute type
else if(!strcmp(switch_word,"CPU") || !strcmp(switch_word,"cpu"))
{
cluster_options.compute=cCPU;
prune_options.compute=cCPU;
}
else if(!strcmp(switch_word,"PRUNE_CPU") || !strcmp(switch_word,"prune_cpu"))
{
prune_flag=1;
prune_options.compute=cCPU;
}
else if(!strcmp(switch_word,"CLUSTER_CPU") || !strcmp(switch_word,"cluster_cpu") )
{
cluster_options.compute=cCPU;
}
else if(!strcmp(switch_word,"GPU") ||!strcmp(switch_word,"gpu") )
{
cluster_options.compute=cGPU;
prune_options.compute=cGPU;
}
else if(!strcmp(switch_word,"PRUNE_GPU") || !strcmp(switch_word,"prune_gpu"))
{
prune_flag=1;
prune_options.compute=cGPU;
}
else if(!strcmp(switch_word,"cluster_gpu") || !strcmp(switch_word,"CLUSTER_GPU"))
{
cluster_options.compute=cGPU;
}
//score_type
else if(!strcmp(switch_word,"RMSD") || !strcmp(switch_word,"rmsd"))
{
cluster_options.score_type=RMSD;
prune_options.score_type=RMSD;
}
else if(!strcmp(switch_word,"PRUNE_RMSD") || !strcmp(switch_word,"prune_rmsd"))
{
prune_options.score_type=RMSD;
}
else if(!strcmp(switch_word,"CLUSTER_RMSD") || !strcmp(switch_word,"cluster_rmsd"))
{
cluster_options.score_type=RMSD;
}
else if(!strcmp(switch_word,"TMSCORE") || !strcmp(switch_word,"tmscore"))
{
cluster_options.score_type=TMSCORE;
prune_options.score_type=TMSCORE;
}
else if(!strcmp(switch_word,"PRUNE_TMSCORE") || !strcmp(switch_word,"prune_tmscore"))
{
prune_options.score_type=TMSCORE;
}
else if(!strcmp(switch_word,"CLUSTER_TMSCORE") || !strcmp(switch_word,"cluster_tmscore"))
{
cluster_options.score_type=TMSCORE;
}
//SIMD TYPE
#ifdef SSE2
else if(!strcmp(switch_word,"SSE2") || !strcmp(switch_word,"sse2"))
{
cluster_options.simd_type=SSE2_;
prune_options.simd_type=SSE2_;
}
else if(!strcmp(switch_word,"PRUNE_SSE2") || !strcmp(switch_word,"prune_sse2"))
{
prune_flag=1;
prune_options.simd_type=SSE2_;
}
else if(!strcmp(switch_word,"CLUSTER_SSE2") || !strcmp(switch_word,"cluster_sse2"))
{
cluster_options.simd_type=SSE2_;
}
#endif
#ifdef SSE3
else if(!strcmp(switch_word,"SSE3")|| !strcmp(switch_word,"sse3"))
{
cluster_options.simd_type=SSE3_;
prune_options.simd_type=SSE3_;
}
else if(!strcmp(switch_word,"PRUNE_SSE3") || !strcmp(switch_word,"prune_sse3"))
{
prune_options.simd_type=SSE3_;
}
else if(!strcmp(switch_word,"CLUSTER_SSE3")|| !strcmp(switch_word,"prune_sse3"))
{
cluster_options.simd_type=SSE3_;
}
#endif
#ifdef AVX
else if(!strcmp(switch_word,"AVX") || !strcmp(switch_word,"avx"))
{
cluster_options.simd_type=AVX_;
prune_options.simd_type=AVX_;
}
else if(!strcmp(switch_word,"PRUNE_AVX") || !strcmp(switch_word,"prune_avx"))
{
prune_flag=1;
prune_options.simd_type=AVX_;
}
else if(!strcmp(switch_word,"CLUSTER_AVX") || !strcmp(switch_word,"cluster_avx"))
{
cluster_options.simd_type=AVX_;
}
#endif
//distance matrix type - compact or float - default is float - fairly easy to add double later
else if(!strcmp(switch_word,"COMPACT") || !strcmp(switch_word,"compact"))
{
cluster_options.distance_matrix_type=COMPACT;
prune_options.distance_matrix_type=COMPACT;
}
else if(!strcmp(switch_word,"PRUNE_COMPACT") || !strcmp(switch_word,"prune_compact"))
{
prune_flag++;
prune_options.distance_matrix_type=COMPACT;
}
else if(!strcmp(switch_word,"CLUSTER_COMPACT") || !strcmp(switch_word,"cluster_compact"))
{
cluster_options.distance_matrix_type=COMPACT;
}
else if(!strcmp(switch_word,"FLOAT") || !strcmp(switch_word,"float"))
{
cluster_options.distance_matrix_type=FLOAT;
prune_options.distance_matrix_type=FLOAT;
}
else if(!strcmp(switch_word,"PRUNE_FLOAT") || !strcmp(switch_word,"prune_float"))
{
prune_flag++;
prune_options.distance_matrix_type=FLOAT;
}
else if(!strcmp(switch_word,"PRUNE_LOG") || !strcmp(switch_word,"prune_log"))
{
prune_options.keep_log=1;
}
else if(!strcmp(switch_word,"CLUSTER_FLOAT")|| !strcmp(switch_word,"cluster_float") )
{
cluster_options.distance_matrix_type=FLOAT;
}
//input/output_matrix_type
else if (!strcmp(switch_word,"read_binary_matrix") ||
!strcmp(switch_word,"READ_BINARY_MATRIX") ||
!strcmp(switch_word,"READ_MATRIX") ||
!strcmp(switch_word,"read_matrix")
)
{
i++;
cluster_options.read_matrix_type=BINARY;
strcpy(cluster_options.read_matrix_filename,argv[i]);
}
else if (!strcmp(switch_word,"read_text_matrix") ||!strcmp(switch_word,"cluster_read_text_matrix") ||
!strcmp(switch_word,"READ_TEXT_MATRIX") ||!strcmp(switch_word,"CLUSTER_READ_TEXT_MATRIX") )
{
i++;
cluster_options.read_matrix_type=TEXT;
strcpy(cluster_options.read_matrix_filename,argv[i]);
}
else if (!strcmp(switch_word,"prune_read_binary_matrix") || !strcmp(switch_word,"PRUNE_READ_BINARY_MATRIX") )
{
i++;
prune_options.read_matrix_type=BINARY;
strcpy(prune_options.read_matrix_filename,argv[i]);
}
else if (!strcmp(switch_word,"prune_read_text_matrix") || !strcmp(switch_word,"PRUNE_READ_TEXT_MATRIX") )
{
i++;
prune_options.read_matrix_type=TEXT;
strcpy(prune_options.read_matrix_filename,argv[i]);
}
else if (!strcmp(switch_word,"cluster_read_binary_matrix") ||
!strcmp(switch_word,"CLUSTER_READ_BINARY_MATRIX") )
{
i++;
cluster_options.read_matrix_type=BINARY;
strcpy(cluster_options.read_matrix_filename,argv[i]);
}
else if (!strcmp(switch_word,"read_compact_matrix") ||!strcmp(switch_word,"cluster_read_compact_matrix") ||
!strcmp(switch_word,"READ_COMPACT_MATRIX") ||!strcmp(switch_word,"CLUSTER_READ_COMPACT_MATRIX") )
{
i++;
cluster_options.read_matrix_type=CHAR;
strcpy(cluster_options.read_matrix_filename,argv[i]);
}
else if (!strcmp(switch_word,"prune_read_compact_matrix") || !strcmp(switch_word,"PRUNE_READ_COMPACT_MATRIX") )
{
i++;
prune_options.read_matrix_type=CHAR;
strcpy(prune_options.read_matrix_filename,argv[i]);
}
//write flags
else if (!strcmp(switch_word,"write_binary_matrix") ||!strcmp(switch_word,"cluster_write_binary_matrix") ||
!strcmp(switch_word,"WRITE_BINARY_MATRIX") ||!strcmp(switch_word,"CLUSTER_WRITE_BINARY_MATRIX") ||
!strcmp(switch_word,"WRITE_MATRIX") ||!strcmp(switch_word,"CLUSTER_WRITE_MATRIX")
)
{
i++;
cluster_options.write_matrix_type=BINARY;
strcpy(cluster_options.write_matrix_filename,argv[i]);
}
else if (!strcmp(switch_word,"write_text_matrix") ||!strcmp(switch_word,"cluster_write_text_matrix") ||
!strcmp(switch_word,"WRITE_TEXT_MATRIX") ||!strcmp(switch_word,"CLUSTER_WRITE_TEXT_MATRIX") )
{
i++;
cluster_options.write_matrix_type=TEXT;
strcpy(cluster_options.write_matrix_filename,argv[i]);
}
else if (!strcmp(switch_word,"prune_write_binary_matrix") || !strcmp(switch_word,"PRUNE_WRITE_BINARY_MATRIX") ||
!strcmp(switch_word,"prune_write_matrix") || !strcmp(switch_word,"PRUNE_WRITE_MATRIX") )
{
i++;
prune_options.write_matrix_type=BINARY;
strcpy(prune_options.write_matrix_filename,argv[i]);
}
else if (!strcmp(switch_word,"prune_write_text_matrix") || !strcmp(switch_word,"PRUNE_WRITE_TEXT_MATRIX"))
{
i++;
prune_options.write_matrix_type=TEXT;
strcpy(prune_options.write_matrix_filename,argv[i]);
}
else if (!strcmp(switch_word,"write_compact_matrix") ||!strcmp(switch_word,"cluster_write_compact_matrix") ||
!strcmp(switch_word,"WRITE_COMPACT_MATRIX") ||!strcmp(switch_word,"CLUSTER_WRITE_COMPACT_MATRIX") )
{
i++;
cluster_options.write_matrix_type=CHAR;
strcpy(cluster_options.write_matrix_filename,argv[i]);
}
else if (!strcmp(switch_word,"prune_write_compact_matrix") || !strcmp(switch_word,"PRUNE_WRITE_COMPACT_MATRIX") )
{
i++;
prune_options.write_matrix_type=CHAR;
strcpy(prune_options.write_matrix_filename,argv[i]);
}
//cluster flags
else if (!strcmp(switch_word,"hcomplete") ||!strcmp(switch_word,"HCOMPLETE") ||
!strcmp(switch_word,"cluster_hcomplete") ||!strcmp(switch_word,"CLUSTER_HCOMPLETE") )
{
cluster_options.method=HCOMPLETE;
}
else if (!strcmp(switch_word,"hsingle") ||!strcmp(switch_word,"HSINGLE") ||
!strcmp(switch_word,"cluster_hsingle") ||!strcmp(switch_word,"cluster_HSINGLE") )
{
cluster_options.method=HSINGLE;
}
else if (!strcmp(switch_word,"haverage") ||!strcmp(switch_word,"HAVERAGE") ||
!strcmp(switch_word,"cluster_haverage") ||!strcmp(switch_word,"cluster_HAVERAGE") )
{
cluster_options.method=HAVERAGE;
}
else if (!strcmp(switch_word,"prune_hcomplete") ||!strcmp(switch_word,"PRUNE_HCOMPLETE"))
{
prune_options.method=HCOMPLETE;
}
else if (!strcmp(switch_word,"prune_hsingle") ||!strcmp(switch_word,"PRUNE_HSINGLE"))
{
prune_options.method=HSINGLE;
}
else if (!strcmp(switch_word,"prune_haverage") ||!strcmp(switch_word,"PRUNE_HAVERAGE"))
{
prune_options.method=HAVERAGE;
}
else if (!strcmp(switch_word,"kcenters") ||!strcmp(switch_word,"KCENTERS") ||
!strcmp(switch_word,"cluster_kcenters") ||!strcmp(switch_word,"CLUSTER_KCENTERS") )
{
cluster_options.method=KCENTERS;
}
else if (!strcmp(switch_word,"prune_kcenters") ||!strcmp(switch_word,"PRUNE_KCENTERS"))
{
prune_options.method=KCENTERS;
}
else if (!strcmp(switch_word,"prune_density") ||!strcmp(switch_word,"PRUNE_DENSITY"))
{
prune_options.method=DENSITY;
}
else if (!strcmp(switch_word,"kmeans") ||!strcmp(switch_word,"KMEANS") ||
!strcmp(switch_word,"cluster_kmeans") ||!strcmp(switch_word,"CLUSTER_KMEANS") )
{
cluster_options.method=KMEANS;
}
else if (!strcmp(switch_word,"DENSITY") ||!strcmp(switch_word,"density") ||
!strcmp(switch_word,"cluster_density") ||!strcmp(switch_word,"CLUSTER_DENSITY") )
{
cluster_options.method=DENSITY;
}
//cluster option flags
else if (!strcmp(switch_word,"nclusters")||!strcmp(switch_word,"NCLUSTERS"))
{
i++;
cluster_options.nclusters=atoi(argv[i]);
prune_options.nclusters=atoi(argv[i]);
}
else if (!strcmp(switch_word,"cluster_nclusters") ||!strcmp(switch_word,"CLUSTER_NCLUSTERS") )
{
i++;
cluster_options.nclusters=atoi(argv[i]);
}
else if (!strcmp(switch_word,"prune_nclusters") ||!strcmp(switch_word,"PRUNE_NCLUSTERS") )
{
i++;
prune_options.nclusters=atoi(argv[i]);
}
//kmeans specific
else if (!strcmp(switch_word,"min_cluster_size") || !strcmp(switch_word,"MIN_CLUSTER_SIZE"))
{
i++;
cluster_options.min_cluster_size=atoi(argv[i]);
}
else if (!strcmp(switch_word,"max_iterations") || !strcmp(switch_word,"MAX_ITERATIONS"))
{
i++;
cluster_options.maximum_iterations=atoi(argv[i]);
}
else if (!strcmp(switch_word,"total_seeds") || !strcmp(switch_word,"TOTAL_SEEDS"))
{
i++;
cluster_options.total_iterations=atoi(argv[i]);
}
else if (!strcmp(switch_word,"converge_seeds") || !strcmp(switch_word,"CONVERGE_SEEDS") )
{//number of repetitions without improving score that score converges
i++;
cluster_options.nsolutions_after_best_score=atoi(argv[i]);
}
else if (!strcmp(switch_word,"percentile") || !strcmp(switch_word,"PERCENTILE"))
{//number of repetitions without improving score that score converges
i++;
cluster_options.percentile=atof(argv[i]);
}
else if (!strcmp(switch_word,"pvalue") || !strcmp(switch_word,"PVALUE"))
{
i++;
cluster_options.pvalue=atof(argv[i]);
}
else if (!strcmp(switch_word,"fixed_centers")|| !strcmp(switch_word,"FIXED_CENTERS"))
{
i++;
cluster_options.nfixed_centers=atoi(argv[i]);
}
else if (!strcmp(switch_word,"fine_parallel")|| !strcmp(switch_word,"FINE_PARALLEL"))
{
cluster_options.fine_parallel=1;
}
//prune density specific options
else if(!strcmp(switch_word,"prune_zmode") || !strcmp(switch_word,"PRUNE_ZMODE"))
{
prune_flag=1;
prune_options.prune_zmode=1;
}
else if(!strcmp(switch_word,"prune_to_density") || !strcmp(switch_word,"PRUNE_TO_DENSITY"))
{
i++;
prune_options.prune_to=atof(argv[i]);
prune_flag=1;
}
else if (!strcmp(switch_word,"PRUNE_STOP_DENSITY") || !strcmp(switch_word,"prune_stop_density"))
{
i++;
prune_options.prune_stop=atof(argv[i]);
prune_flag=1;
}
else if (!strcmp(switch_word,"prune_outlier_ratio") || !strcmp(switch_word,"RRUNE_OUTLIER_RATIO"))
{
i++;
prune_options.prune_outlier_ratio=atof(argv[i]);
prune_flag=1;
}
else if (!strcmp(switch_word,"prune_stop_at_size") || !strcmp(switch_word,"PRUNE_STOP_AT_SIZE"))
{
i++;
prune_options.prune_min_size=atoi(argv[i]);
prune_flag=1;
}
else if (!strcmp(switch_word,"prune_until_size") || !strcmp(switch_word,"PRUNE_UNTIL_SIZE") )
{
i++;
prune_options.prune_max_size=atoi(argv[i]);
prune_flag=1;
}
//density only
else if (!strcmp(switch_word,"density_only") || !strcmp(switch_word,"DENSITY_ONLY") )
{
density_only_flag=1; //calculate and output the density only
}
else if (!strcmp(switch_word,"sort_density") || !strcmp(switch_word,"SORT_DENSITY") )
{
sort_density=true; //this only applies to density_only mode
}
//coords mode
else if(!strcmp(switch_word,"binary_coords") || !strcmp(switch_word,"BINARY_COORDS") )
{
prune_options.input_type=BINARY_COORDS;
cluster_options.input_type=BINARY_COORDS;
}
else if(!strcmp(switch_word,"cluster_binary_coords") || !strcmp(switch_word,"CLUSTER_BINARY_COORDS") )
{
cluster_options.input_type=BINARY_COORDS;
}
else if(!strcmp(switch_word,"prune_binary_coords") || !strcmp(switch_word,"PRUNE_BINARY_COORDS") )
{
prune_options.input_type=BINARY_COORDS;
}
else if(!strcmp(switch_word,"all_atoms") || !strcmp(switch_word,"ALL_ATOMS") )
{
all_atoms_flag=1;
prune_options.all_atoms=1;
cluster_options.all_atoms=1;
}
else if (!strcmp(switch_word,"help") || !strcmp(switch_word,"HELP") )
{
fprintf(stderr,"****OPTIONS*****\n%s",options_text);
exit(FALSE);
}
#ifdef GPU
//gpu_flags
else if (!strcmp(switch_word,"gpu_id") || !strcmp(switch_word,"GPU_ID") )
{
i++;
prune_options.gpu_id=atoi(argv[i]);
cluster_options.gpu_id=atoi(argv[i]);
}
else if (!strcmp(switch_word,"prune_gpu_id") || !strcmp(switch_word,"PRUNE_GPU_ID") )
{
i++;
prune_options.gpu_id=atoi(argv[i]);
}
else if (!strcmp(switch_word,"cluster_gpu_id") || !strcmp(switch_word,"CLUSTER_GPU_ID") )
{
i++;
cluster_options.gpu_id=atoi(argv[i]);
}
#endif
else
{
fprintf(stderr,"unrecognized option %s\n",argv[i]);
fprintf(stderr,"%s",options_text);
exit(FALSE);
}
}
break;
case 'i' :
{
i++;
strcpy(prune_options.input_filename, argv[i]);
strcpy(cluster_options.input_filename, argv[i]);
}
break;
case 'o' :
{
i++;
strcpy(prune_options.output_filename, argv[i]);
strcpy(cluster_options.output_filename, argv[i]);
}
break;
case 'S' :
{
i++;
seed = (unsigned int) atoi(argv[i]);
}
break;
#ifdef GPU
case 'p' :
{
//path to tmscore.cl rmsd.cl
i++;
sprintf(gtmscorecl_location,"%s/tmscore.cl",argv[i]);
sprintf(grmsdcl_location,"%s/tmscore.cl",argv[i]);
}
#endif
default :
{
fprintf(stderr,"unrecognized option %s\n",argv[i]);
fprintf(stderr,"%s",options_text);
exit(FALSE);
}
}
i++;
}
else
{
fprintf(stderr,"unrecognized option %s\n",argv[i]);
fprintf(stderr,"%s",options_text);
exit(FALSE);
}
}
}
#ifdef OPENMP
{
int max_threads=omp_get_max_threads();
nthreads=(max_threads<nthreads)? max_threads : nthreads;
fprintf(stderr,"Using %d threads\n",nthreads);
}
#endif
if(all_atoms_flag){
if(cluster_options.score_type==TMSCORE || prune_options.score_type==TMSCORE){
fprintf(stderr,"all_atoms option presently only supported for RMSD - will continue with using only CA coords\n");
all_atoms_flag=0;
prune_options.all_atoms=0;
cluster_options.all_atoms=0;
}
}
if (density_only_flag){
//do not create a distance matrix - calculate the density vector directly and output
//can generate the density de novo from pdb files or from a pre-exisiting matrix file
//if a matrix is read in - the density vector is still calculated directly without memory being allocated for a temp distance matrix
//uses the other options passed to cluster options wrt to file names, metric etc...
FILE *fp;
cluster_models_set<float> *models;
if(cluster_options.read_matrix_type != NO_MATRIX){
models= new cluster_models_set<float>("READ_MATRIX",nthreads,cluster_options.score_type,cluster_options.all_atoms,cluster_options.input_filename,cluster_options.read_matrix_filename,
cluster_options.read_matrix_type,cluster_options.distance_matrix_type,(int*)0,cluster_options.subset_filename,1);
}
else if(cluster_options.input_type == BINARY_COORDS){
models=new cluster_models_set<float>("READ_BINARY_COORDS",nthreads,cluster_options.score_type,cluster_options.all_atoms,cluster_options.compute,cluster_options.distance_matrix_type,
cluster_options.input_filename,(int*) 0,cluster_options.subset_filename,cluster_options.simd_type,1,cluster_options.gpu_id);
}
else{ models=new cluster_models_set<float>("READ_PDB_LIST",nthreads,cluster_options.score_type,cluster_options.all_atoms,cluster_options.compute,cluster_options.distance_matrix_type,
cluster_options.input_filename,(int*) 0,cluster_options.subset_filename,cluster_options.simd_type,1,cluster_options.gpu_id);
}
char temp_filename[FILENAME_LENGTH];;
sprintf(temp_filename,"%s.density",cluster_options.output_filename);
open_file(&fp, temp_filename, "w", 0);
models->print_density(fp,sort_density);
close_file(&fp, temp_filename,0);
if(models) delete models;
return(TRUE);
}
void *vprune_models=0;
if(prune_flag){
if(prune_options.distance_matrix_type == COMPACT){
if(cluster_options.distance_matrix_type == COMPACT){
vprune_models=prune_it<unsigned char,unsigned char>(nthreads,&prune_options,&cluster_options,seed);
}
else{
vprune_models=prune_it<unsigned char,float>(nthreads,&prune_options,&cluster_options,seed);
}
}
else if(prune_options.distance_matrix_type == FLOAT){
if(cluster_options.distance_matrix_type == COMPACT){
vprune_models=prune_it<float,unsigned char>(nthreads,&prune_options,&cluster_options,seed);
}
else{
vprune_models=prune_it<float,float>(nthreads,&prune_options,&cluster_options,seed);
}
}
if (prune_options.write_matrix_type != NO_MATRIX){
FILE *matrix_fp;
char matrix_filename[FILENAME_LENGTH];
sprintf(matrix_filename,"%s.post_prune",prune_options.write_matrix_filename);
open_file(&matrix_fp, matrix_filename, "w", "main");
char temp_filename[FILENAME_LENGTH];;
sprintf(temp_filename,"%s.post_prune.names",prune_options.output_filename);
if(prune_options.distance_matrix_type == COMPACT){
cluster_models_set<unsigned char>* models=(cluster_models_set<unsigned char>*)vprune_models;
models->write_names_to_file(temp_filename,0,0);
if (prune_options.write_matrix_type == BINARY){
models->dmatrix->write_matrix_to_binary_file(matrix_fp);
}
else if(prune_options.write_matrix_type == TEXT){
models->dmatrix->write_matrix_to_text_file(matrix_fp);
}
else if(prune_options.write_matrix_type == CHAR ){
models->dmatrix->write_matrix_to_compact_file(matrix_fp);
}
close_file(&matrix_fp, prune_options.write_matrix_filename,"main");
}
else{
cluster_models_set<float>* models=(cluster_models_set<float>*)vprune_models;
models->write_names_to_file(temp_filename,0,0);
if (prune_options.write_matrix_type == BINARY){
models->dmatrix->write_matrix_to_binary_file(matrix_fp);
}
else if(prune_options.write_matrix_type == TEXT){
models->dmatrix->write_matrix_to_text_file(matrix_fp);
}
else if(prune_options.write_matrix_type == CHAR ){
models->dmatrix->write_matrix_to_compact_file(matrix_fp);
}
}
}
}
if(cluster_options.distance_matrix_type == COMPACT){
cluster_models_set<unsigned char> *models=(cluster_models_set<unsigned char>*)vprune_models;
cluster_it<unsigned char>(nthreads,&cluster_options,models,seed);
}
else if(cluster_options.distance_matrix_type == FLOAT){
cluster_models_set<float> *models=(cluster_models_set<float>*)vprune_models;
cluster_it<float>(nthreads,&cluster_options,models,seed);
}
}
template <class T,class T2> //types are those of the prune and cluster classes
cluster_models_set<T2>* prune_it(int nthreads, prune_options *prune_options,cluster_options *cluster_options,unsigned int seed){
fprintf(stderr,"pruning the set\n");
void *vmodels=0;
mapped_cluster_set<T>* prune_cluster=0;
//if the distance_matrix type is the same in pruning and clustering models is returned otherwise cmodels is returned
//there is an obscure way of doing this
cluster_models_set<T>* models=0;
cluster_models_set<T2>* out_models=0;
if(prune_options->read_matrix_type != NO_MATRIX){
int *inverse_map=0;
models=new cluster_models_set<T>("READ_MATRIX",nthreads,prune_options->score_type,prune_options->all_atoms,prune_options->input_filename,prune_options->read_matrix_filename,
prune_options->read_matrix_type,prune_options->distance_matrix_type,(int*) 0,prune_options->subset_filename,0);
}
else if(prune_options->input_type==BINARY_COORDS){ models=new cluster_models_set<T>("READ_BINARY_COORDS",nthreads,prune_options->score_type,prune_options->all_atoms,prune_options->compute,prune_options->distance_matrix_type,
prune_options->input_filename,(int*)0,prune_options->subset_filename,prune_options->simd_type,0,prune_options->gpu_id);
}
else{ models=new cluster_models_set<T>("READ_PDB_LIST",nthreads,prune_options->score_type,prune_options->all_atoms,prune_options->compute,prune_options->distance_matrix_type,
prune_options->input_filename,(int*)0,prune_options->subset_filename,prune_options->simd_type,0,prune_options->gpu_id);
}
//write out the pre_prune matrices
if (prune_options->write_matrix_type != NO_MATRIX){
FILE *matrix_fp;
char matrix_filename[FILENAME_LENGTH];
sprintf(matrix_filename,"%s.pre_prune",prune_options->write_matrix_filename);
open_file(&matrix_fp,matrix_filename, "w", "main");
char temp_filename[FILENAME_LENGTH];;
sprintf(temp_filename,"%s.pre_prune.names",prune_options->output_filename);
if(prune_options->distance_matrix_type == COMPACT){
models->write_names_to_file(temp_filename,0,0);
if (prune_options->write_matrix_type == BINARY){
models->dmatrix->write_matrix_to_binary_file(matrix_fp);
}
else if(prune_options->write_matrix_type == TEXT){
models->dmatrix->write_matrix_to_text_file(matrix_fp);
}
else if(prune_options->write_matrix_type == CHAR ){
models->dmatrix->write_matrix_to_compact_file(matrix_fp);
}
close_file(&matrix_fp, prune_options->write_matrix_filename,"main");
}
else{
models->write_names_to_file(temp_filename,0,0);
if (prune_options->write_matrix_type == BINARY){
models->dmatrix->write_matrix_to_binary_file(matrix_fp);
}
else if(prune_options->write_matrix_type == TEXT){
models->dmatrix->write_matrix_to_text_file(matrix_fp);
}
else if(prune_options->write_matrix_type == CHAR ){
models->dmatrix->write_matrix_to_compact_file(matrix_fp);
}
}
}
if(prune_options->method ==DENSITY){//no partition is needed if density is used
FILE *log_fp=0;
char log_filename[FILENAME_LENGTH];
if(prune_options->keep_log){
sprintf(log_filename,"%s.pruned",prune_options->output_filename);
open_file(&log_fp, log_filename, "w", "main");
}
prune_cluster = new mapped_cluster_set<T>(models,1,1,seed);
prune_cluster->models->idensity_filter(prune_options->prune_zmode,prune_options->prune_to,prune_options->prune_stop,prune_options->prune_min_size,prune_options->prune_max_size,prune_options->prune_outlier_ratio,log_fp);
if(prune_options->keep_log){
close_file(&log_fp, log_filename,"main");
sprintf(log_filename,"%s.density",prune_options->output_filename);
open_file(&log_fp, log_filename, "w", "main");
prune_cluster->models->print_density(log_fp);
close_file(&log_fp, log_filename,"main");
}
}
else{
if(!prune_options->prune_max_size){
fprintf(stderr,"For pruning by any method other than density, the prune_until_size option is the only recognized and must be set\n");
exit(FALSE);
}
fprintf(stderr,"pruning from %d structures to %d structures\n",models->nmodels,prune_options->prune_max_size);
double start_cluster=get_time();
//hierarchical clustering
if(prune_options->method == HSINGLE || prune_options->method == HCOMPLETE ||prune_options->method == HAVERAGE ){
int *history=0;
int initial_nmodels=models->nmodels;
prune_cluster= new mapped_cluster_set<T>(models,models->nmodels,1,seed,0);
history=new int [initial_nmodels*2];
if(prune_options->method == HCOMPLETE){
fprintf(stderr,"using complete linkage/max-distance between elements as cluster distance\n");
prune_cluster->best_partition->reduce_by_agglomeration_complete_linkage(prune_options->nclusters,prune_cluster->models,history,0,1,nthreads);
}
else if (prune_options->method == HSINGLE){
fprintf(stderr,"using single linkage/min-distance between elements as cluster distance\n");
prune_cluster->best_partition->reduce_by_agglomeration_single_linkage(prune_options->nclusters,prune_cluster->models,history,0,1,nthreads);
}
else if (prune_options->method == HAVERAGE){
fprintf(stderr,"using average distance between elements as cluster distance\n");
prune_cluster->best_partition->reduce_by_agglomeration_average(prune_options->nclusters,prune_cluster->models,history,0,1,nthreads);
}
if(history)
{//print out history of groups joined
FILE *fp;
char temp_filename[FILENAME_LENGTH];;
sprintf(temp_filename,"%s.prune.agglomeration.history",prune_options->output_filename);
fprintf(stderr,"writing out history file\n");
open_file(&fp,temp_filename, "w", 0);
for(int n=0;n<initial_nmodels;n++)
fprintf(fp,"%d %d %d\n",n,history[2*n],history[2*n+1]);
close_file(&fp,temp_filename, 0);
if(history) delete [] history;
}
}
else if(prune_options->method == KCENTERS){
prune_cluster= new mapped_cluster_set<T>(models,prune_options->nclusters,1,seed);
prune_cluster->best_partition->kcenters(100,100,prune_cluster->models, &cluster_partition<T>::assign_to_lowest_average_distance);
}
prune_cluster->print_centers(stderr,prune_cluster->best_partition);
prune_cluster->filter_by_average_centroid_distance(prune_options->prune_max_size);
}
//write the pruned state out to the real models in the class
//save them as out_models
//is the pruned matrix a simple subset of the original?
if((prune_options->score_type == cluster_options->score_type &&
prune_options->distance_matrix_type == cluster_options->distance_matrix_type)){
fprintf(stderr,"simple subset\n");
prune_cluster->models->base_set->write_state_to_base(prune_cluster->models->map,prune_cluster->models->nmodels);
if(prune_cluster)delete prune_cluster->best_partition;
vmodels=prune_cluster->models->base_set;
return((cluster_models_set<T2>*) vmodels);
}
else{
//prepare new distance matrix
//are the distance matrix class type the same
if(prune_options->distance_matrix_type == cluster_options->distance_matrix_type){
fprintf(stderr,"same type class %d\n",prune_options->distance_matrix_type);
if(cluster_options->read_matrix_type != NO_MATRIX) //read new matrix
prune_cluster->models->base_set->write_state_to_base(cluster_options->read_matrix_filename,cluster_options->score_type,cluster_options->read_matrix_type,prune_cluster->models->map,prune_cluster->models->nmodels);
else
prune_cluster->models->base_set->write_state_to_base(cluster_options->compute,cluster_options->score_type,cluster_options->simd_type,prune_cluster->models->map,prune_cluster->models->nmodels,nthreads,prune_options->gpu_id);
if(prune_cluster->best_partition)delete prune_cluster->best_partition;
vmodels=prune_cluster->models->base_set;
return((cluster_models_set<T2>*) vmodels);
}
else{//both are class type and metric type are different
fprintf(stderr,"different type class\n");
cluster_models_set<T2>* out_models=new cluster_models_set<T2>();
if(cluster_options->read_matrix_type != NO_MATRIX) //read new matrix
prune_cluster->models->base_set->write_state_to_different_type_base(out_models,cluster_options->read_matrix_filename,cluster_options->score_type,cluster_options->read_matrix_type,prune_cluster->models->map,prune_cluster->models->nmodels);
else prune_cluster->models->base_set->write_state_to_different_type_base(out_models,cluster_options->compute,cluster_options->score_type,cluster_options->simd_type,prune_cluster->models->map,prune_cluster->models->nmodels,nthreads,cluster_options->gpu_id);
if(prune_cluster)delete prune_cluster;
return(out_models);
}
}
}
template <class T>
void cluster_it(int nthreads,cluster_options *cluster_options,cluster_models_set<T> *models,unsigned int seed){
//check whether a distance matrix and models needs to be generated
if(!models){
if(cluster_options->read_matrix_type != NO_MATRIX){
models=new cluster_models_set<T>("READ_MATRIX",nthreads,cluster_options->score_type,cluster_options->all_atoms,cluster_options->input_filename,cluster_options->read_matrix_filename,
cluster_options->read_matrix_type,cluster_options->distance_matrix_type,(int*)0,cluster_options->subset_filename,0);
}
else if (cluster_options->input_type == BINARY_COORDS){models=new cluster_models_set<T>("READ_BINARY_COORDS",nthreads,cluster_options->score_type,cluster_options->all_atoms,cluster_options->compute,cluster_options->distance_matrix_type,
cluster_options->input_filename,(int*) 0,cluster_options->subset_filename,cluster_options->simd_type,0,cluster_options->gpu_id);
}
else{ models=new cluster_models_set<T>("READ_PDB_LIST",nthreads,cluster_options->score_type,cluster_options->all_atoms,cluster_options->compute,cluster_options->distance_matrix_type,
cluster_options->input_filename,(int*) 0,cluster_options->subset_filename,cluster_options->simd_type,0,cluster_options->gpu_id);
}
}
//do we write the matrix
if (cluster_options->write_matrix_type != NO_MATRIX){
FILE *matrix_fp;
open_file(&matrix_fp, cluster_options->write_matrix_filename, "w", "cluster_it");
char temp_filename[FILENAME_LENGTH];;
sprintf(temp_filename,"%s.names",cluster_options->output_filename);
models->write_names_to_file(temp_filename,0,0);
if (cluster_options->write_matrix_type == BINARY){
models->dmatrix->write_matrix_to_binary_file(matrix_fp);
}
else if(cluster_options->write_matrix_type == TEXT){
models->dmatrix->write_matrix_to_text_file(matrix_fp);
}
else if(cluster_options->write_matrix_type == CHAR ){
models->dmatrix->write_matrix_to_compact_file(matrix_fp);
}
else{
fprintf(stderr,"unrecognized write matrix type - %d binary format written\n",cluster_options->write_matrix_type);
models->dmatrix->write_matrix_to_binary_file(matrix_fp);
}
close_file(&matrix_fp,cluster_options->write_matrix_filename,"cluster_it");
}
//which cluster method is to be used
if(cluster_options->method == DENSITY || cluster_options->nclusters ==1){
mapped_cluster_models_set<T> cluster = mapped_cluster_models_set<T>(models);
FILE *log_fp;
char log_filename[FILENAME_LENGTH];
sprintf(log_filename,"%s.density",cluster_options->output_filename);
open_file(&log_fp, log_filename, "w", "main");
fprintf(stderr,"saving density to %s\n",log_filename);
cluster.print_density(log_fp);
close_file(&log_fp,log_filename,"main");
if(models)delete(models);
return;
}
else if(cluster_options->method == KCENTERS || (cluster_options->method == KMEANS && cluster_options->nfixed_centers >= cluster_options->nclusters)){
double start_cluster=get_time();
mapped_cluster_set<T> kc_cluster=mapped_cluster_set<T>(models,cluster_options->nclusters,1,seed);
kc_cluster.best_partition->kcenters(100,100,kc_cluster.models, &cluster_partition<T>::assign_to_lowest_average_distance);
fprintf(stderr, "%8.3f seconds elapsed to cluster\n",get_time()-start_cluster);
kc_cluster.print_centers(stderr,kc_cluster.best_partition);
FILE *output_fp;
char temp_filename[FILENAME_LENGTH];
sprintf(temp_filename,"%s.cluster.stats",cluster_options->output_filename);
open_file(&output_fp, temp_filename, "w", "main");
kc_cluster.print_centers(output_fp,kc_cluster.best_partition);
kc_cluster.print_cluster_stats(output_fp,kc_cluster.best_partition);
close_file(&output_fp,temp_filename,"main");
sprintf(temp_filename,"%s.clusters",cluster_options->output_filename);
open_file(&output_fp, temp_filename, "w", "main");
kc_cluster.print_density(output_fp,kc_cluster.best_partition);
close_file(&output_fp,temp_filename,"main");
if(models)delete(models);
return;
}
//hierarchical clustering
else if(cluster_options->method == HCOMPLETE || cluster_options->method == HAVERAGE || cluster_options->method == HSINGLE ){
fprintf(stderr,"hierarchical clustering\n");
double start_cluster=get_time();
int *history=new int[models->nmodels*2];
for(int i=0;i<models->nmodels*2;i++)
history[i]=-1;
mapped_cluster_set<T> hcluster=mapped_cluster_set<T>((cluster_models_set<T>*) models,models->nmodels,1,seed,0);
if(cluster_options->method == HCOMPLETE){
fprintf(stderr,"using complete linkage/max-distance between elements as cluster distance\n");
hcluster.best_partition->reduce_by_agglomeration_complete_linkage(cluster_options->nclusters,hcluster.models,history,0,1,nthreads);
}
else if (cluster_options->method == HSINGLE){
fprintf(stderr,"using single linkage/min-distance between elements as cluster distance\n");
hcluster.best_partition->reduce_by_agglomeration_single_linkage(cluster_options->nclusters,hcluster.models,history,0,1,nthreads);
}
else{
fprintf(stderr,"using average distance between elements as cluster distance\n");
hcluster.best_partition->reduce_by_agglomeration_average(cluster_options->nclusters,hcluster.models,history,0,1,nthreads);
}
if(history){//print out history of groups joined
FILE *fp;
char temp_filename[FILENAME_LENGTH];;
sprintf(temp_filename,"%s.agglomeration.history",cluster_options->output_filename);
fprintf(stderr,"writing out history file\n");
open_file(&fp,temp_filename, "w", 0);
for(int n=0;n<models->nmodels;n++)
fprintf(fp,"%d %d %d\n",n,history[2*n],history[2*n+1]);
close_file(&fp,temp_filename, 0);
if(history) delete [] history;
}
hcluster.print_centers(stderr,hcluster.best_partition);
fprintf(stderr, "%8.3f seconds elapsed to cluster\n",get_time()-start_cluster);
FILE *output_fp;
char temp_filename[FILENAME_LENGTH];
sprintf(temp_filename,"%s.cluster.stats",cluster_options->output_filename);
open_file(&output_fp, temp_filename, "w", "main");
hcluster.print_centers(output_fp,hcluster.best_partition);
hcluster.print_cluster_stats(output_fp,hcluster.best_partition);
close_file(&output_fp,temp_filename,"main");
sprintf(temp_filename,"%s.clusters",cluster_options->output_filename);
open_file(&output_fp, temp_filename, "w", "main");
hcluster.print_density(output_fp,hcluster.best_partition);
close_file(&output_fp,temp_filename,"main");
if(models)delete models;
return;
}
//kmeans clustering
else if(cluster_options->method == KMEANS){
double start_cluster=get_time();
//fine parallelism - one kmeans seed at a time
if(cluster_options->fine_parallel){
#ifdef OPENMP
nthreads=(nthreads<omp_get_max_threads())?nthreads:omp_get_max_threads();
#endif
if(!(cluster_options->total_iterations) && cluster_options->nsolutions_after_best_score <= 0 ){
//convergence criterion set using pvalue and percential limits
cluster_options->nsolutions_after_best_score=(int)(log(cluster_options->pvalue)/log(cluster_options->percentile))+1;
fprintf(stderr,"%d scores before convergence to ensure final score is in %8.5f percentile with pvalue less than %8.5f\n",cluster_options->nsolutions_after_best_score,cluster_options->percentile,cluster_options->pvalue);
}
mapped_cluster_set<T> kcluster=mapped_cluster_set<T>(models,cluster_options->nclusters,1,seed);
//always update
kcluster.best_partition->recalc_density_ratio=1.0;
float best_score=(kcluster.greater_is_better)? -FLT_MAX : FLT_MAX; //absolute best_score as a function of densities
bool change_flag=false,done =false; //has there been a change in best score
int iterations=0,old_iterations=0,iterations_between_best_scores=0; //these variables control termination
int limit;
if (cluster_options->total_iterations) limit=cluster_options->total_iterations;