-
Notifications
You must be signed in to change notification settings - Fork 0
Expand file tree
/
Copy pathfaq.html
More file actions
779 lines (725 loc) · 59.9 KB
/
faq.html
File metadata and controls
779 lines (725 loc) · 59.9 KB
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
<!doctype html>
<html class="no-js" lang="en" data-content_root="./">
<head><meta charset="utf-8">
<meta name="viewport" content="width=device-width,initial-scale=1">
<meta name="color-scheme" content="light dark"><meta name="viewport" content="width=device-width, initial-scale=1" />
<link rel="index" title="Index" href="genindex.html"><link rel="search" title="Search" href="search.html"><link rel="prev" title="Exceptions" href="api/exceptions.html">
<!-- Generated with Sphinx 9.0.4 and Furo 2025.12.19 -->
<title>Frequently Asked Questions - SPFlow 1.0.0</title>
<link rel="stylesheet" type="text/css" href="_static/pygments.css?v=d111a655" />
<link rel="stylesheet" type="text/css" href="_static/styles/furo.css?v=7bdb33bb" />
<link rel="stylesheet" type="text/css" href="_static/copybutton.css?v=76b2166b" />
<link rel="stylesheet" type="text/css" href="_static/styles/furo-extensions.css?v=8dab3a3b" />
<style>
body {
--color-code-background: #f2f2f2;
--color-code-foreground: #1e1e1e;
--color-brand-primary: #0066cc;
--color-brand-content: #0066cc;
}
@media not print {
body[data-theme="dark"] {
--color-code-background: #202020;
--color-code-foreground: #d0d0d0;
--color-brand-primary: #4da6ff;
--color-brand-content: #4da6ff;
}
@media (prefers-color-scheme: dark) {
body:not([data-theme="light"]) {
--color-code-background: #202020;
--color-code-foreground: #d0d0d0;
--color-brand-primary: #4da6ff;
--color-brand-content: #4da6ff;
}
}
}
</style></head>
<body>
<script>
document.body.dataset.theme = localStorage.getItem("theme") || "auto";
</script>
<svg xmlns="http://www.w3.org/2000/svg" style="display: none;">
<symbol id="svg-toc" viewBox="0 0 24 24">
<title>Contents</title>
<svg stroke="currentColor" fill="currentColor" stroke-width="0" viewBox="0 0 1024 1024">
<path d="M408 442h480c4.4 0 8-3.6 8-8v-56c0-4.4-3.6-8-8-8H408c-4.4 0-8 3.6-8 8v56c0 4.4 3.6 8 8 8zm-8 204c0 4.4 3.6 8 8 8h480c4.4 0 8-3.6 8-8v-56c0-4.4-3.6-8-8-8H408c-4.4 0-8 3.6-8 8v56zm504-486H120c-4.4 0-8 3.6-8 8v56c0 4.4 3.6 8 8 8h784c4.4 0 8-3.6 8-8v-56c0-4.4-3.6-8-8-8zm0 632H120c-4.4 0-8 3.6-8 8v56c0 4.4 3.6 8 8 8h784c4.4 0 8-3.6 8-8v-56c0-4.4-3.6-8-8-8zM115.4 518.9L271.7 642c5.8 4.6 14.4.5 14.4-6.9V388.9c0-7.4-8.5-11.5-14.4-6.9L115.4 505.1a8.74 8.74 0 0 0 0 13.8z"/>
</svg>
</symbol>
<symbol id="svg-menu" viewBox="0 0 24 24">
<title>Menu</title>
<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 24 24" fill="none" stroke="currentColor"
stroke-width="2" stroke-linecap="round" stroke-linejoin="round" class="feather-menu">
<line x1="3" y1="12" x2="21" y2="12"></line>
<line x1="3" y1="6" x2="21" y2="6"></line>
<line x1="3" y1="18" x2="21" y2="18"></line>
</svg>
</symbol>
<symbol id="svg-arrow-right" viewBox="0 0 24 24">
<title>Expand</title>
<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 24 24" fill="none" stroke="currentColor"
stroke-width="2" stroke-linecap="round" stroke-linejoin="round" class="feather-chevron-right">
<polyline points="9 18 15 12 9 6"></polyline>
</svg>
</symbol>
<symbol id="svg-sun" viewBox="0 0 24 24">
<title>Light mode</title>
<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 24 24" fill="none" stroke="currentColor"
stroke-width="1" stroke-linecap="round" stroke-linejoin="round" class="feather-sun">
<circle cx="12" cy="12" r="5"></circle>
<line x1="12" y1="1" x2="12" y2="3"></line>
<line x1="12" y1="21" x2="12" y2="23"></line>
<line x1="4.22" y1="4.22" x2="5.64" y2="5.64"></line>
<line x1="18.36" y1="18.36" x2="19.78" y2="19.78"></line>
<line x1="1" y1="12" x2="3" y2="12"></line>
<line x1="21" y1="12" x2="23" y2="12"></line>
<line x1="4.22" y1="19.78" x2="5.64" y2="18.36"></line>
<line x1="18.36" y1="5.64" x2="19.78" y2="4.22"></line>
</svg>
</symbol>
<symbol id="svg-moon" viewBox="0 0 24 24">
<title>Dark mode</title>
<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 24 24" fill="none" stroke="currentColor"
stroke-width="1" stroke-linecap="round" stroke-linejoin="round" class="icon-tabler-moon">
<path stroke="none" d="M0 0h24v24H0z" fill="none" />
<path d="M12 3c.132 0 .263 0 .393 0a7.5 7.5 0 0 0 7.92 12.446a9 9 0 1 1 -8.313 -12.454z" />
</svg>
</symbol>
<symbol id="svg-sun-with-moon" viewBox="0 0 24 24">
<title>Auto light/dark, in light mode</title>
<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 24 24" fill="none" stroke="currentColor"
stroke-width="1" stroke-linecap="round" stroke-linejoin="round"
class="icon-custom-derived-from-feather-sun-and-tabler-moon">
<path style="opacity: 50%" d="M 5.411 14.504 C 5.471 14.504 5.532 14.504 5.591 14.504 C 3.639 16.319 4.383 19.569 6.931 20.352 C 7.693 20.586 8.512 20.551 9.25 20.252 C 8.023 23.207 4.056 23.725 2.11 21.184 C 0.166 18.642 1.702 14.949 4.874 14.536 C 5.051 14.512 5.231 14.5 5.411 14.5 L 5.411 14.504 Z"/>
<line x1="14.5" y1="3.25" x2="14.5" y2="1.25"/>
<line x1="14.5" y1="15.85" x2="14.5" y2="17.85"/>
<line x1="10.044" y1="5.094" x2="8.63" y2="3.68"/>
<line x1="19" y1="14.05" x2="20.414" y2="15.464"/>
<line x1="8.2" y1="9.55" x2="6.2" y2="9.55"/>
<line x1="20.8" y1="9.55" x2="22.8" y2="9.55"/>
<line x1="10.044" y1="14.006" x2="8.63" y2="15.42"/>
<line x1="19" y1="5.05" x2="20.414" y2="3.636"/>
<circle cx="14.5" cy="9.55" r="3.6"/>
</svg>
</symbol>
<symbol id="svg-moon-with-sun" viewBox="0 0 24 24">
<title>Auto light/dark, in dark mode</title>
<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 24 24" fill="none" stroke="currentColor"
stroke-width="1" stroke-linecap="round" stroke-linejoin="round"
class="icon-custom-derived-from-feather-sun-and-tabler-moon">
<path d="M 8.282 7.007 C 8.385 7.007 8.494 7.007 8.595 7.007 C 5.18 10.184 6.481 15.869 10.942 17.24 C 12.275 17.648 13.706 17.589 15 17.066 C 12.851 22.236 5.91 23.143 2.505 18.696 C -0.897 14.249 1.791 7.786 7.342 7.063 C 7.652 7.021 7.965 7 8.282 7 L 8.282 7.007 Z"/>
<line style="opacity: 50%" x1="18" y1="3.705" x2="18" y2="2.5"/>
<line style="opacity: 50%" x1="18" y1="11.295" x2="18" y2="12.5"/>
<line style="opacity: 50%" x1="15.316" y1="4.816" x2="14.464" y2="3.964"/>
<line style="opacity: 50%" x1="20.711" y1="10.212" x2="21.563" y2="11.063"/>
<line style="opacity: 50%" x1="14.205" y1="7.5" x2="13.001" y2="7.5"/>
<line style="opacity: 50%" x1="21.795" y1="7.5" x2="23" y2="7.5"/>
<line style="opacity: 50%" x1="15.316" y1="10.184" x2="14.464" y2="11.036"/>
<line style="opacity: 50%" x1="20.711" y1="4.789" x2="21.563" y2="3.937"/>
<circle style="opacity: 50%" cx="18" cy="7.5" r="2.169"/>
</svg>
</symbol>
<symbol id="svg-pencil" viewBox="0 0 24 24">
<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 24 24" fill="none" stroke="currentColor"
stroke-width="1" stroke-linecap="round" stroke-linejoin="round" class="icon-tabler-pencil-code">
<path d="M4 20h4l10.5 -10.5a2.828 2.828 0 1 0 -4 -4l-10.5 10.5v4" />
<path d="M13.5 6.5l4 4" />
<path d="M20 21l2 -2l-2 -2" />
<path d="M17 17l-2 2l2 2" />
</svg>
</symbol>
<symbol id="svg-eye" viewBox="0 0 24 24">
<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 24 24" fill="none" stroke="currentColor"
stroke-width="1" stroke-linecap="round" stroke-linejoin="round" class="icon-tabler-eye-code">
<path stroke="none" d="M0 0h24v24H0z" fill="none" />
<path d="M10 12a2 2 0 1 0 4 0a2 2 0 0 0 -4 0" />
<path
d="M11.11 17.958c-3.209 -.307 -5.91 -2.293 -8.11 -5.958c2.4 -4 5.4 -6 9 -6c3.6 0 6.6 2 9 6c-.21 .352 -.427 .688 -.647 1.008" />
<path d="M20 21l2 -2l-2 -2" />
<path d="M17 17l-2 2l2 2" />
</svg>
</symbol>
</svg>
<input type="checkbox" class="sidebar-toggle" name="__navigation" id="__navigation" aria-label="Toggle site navigation sidebar">
<input type="checkbox" class="sidebar-toggle" name="__toc" id="__toc" aria-label="Toggle table of contents sidebar">
<label class="overlay sidebar-overlay" for="__navigation"></label>
<label class="overlay toc-overlay" for="__toc"></label>
<a class="skip-to-content muted-link" href="#furo-main-content">Skip to content</a>
<div class="page">
<header class="mobile-header">
<div class="header-left">
<label class="nav-overlay-icon" for="__navigation">
<span class="icon"><svg><use href="#svg-menu"></use></svg></span>
</label>
</div>
<div class="header-center">
<a href="index.html"><div class="brand">SPFlow 1.0.0</div></a>
</div>
<div class="header-right">
<div class="theme-toggle-container theme-toggle-header">
<button class="theme-toggle" aria-label="Toggle Light / Dark / Auto color theme">
<svg class="theme-icon-when-auto-light"><use href="#svg-sun-with-moon"></use></svg>
<svg class="theme-icon-when-auto-dark"><use href="#svg-moon-with-sun"></use></svg>
<svg class="theme-icon-when-dark"><use href="#svg-moon"></use></svg>
<svg class="theme-icon-when-light"><use href="#svg-sun"></use></svg>
</button>
</div>
<label class="toc-overlay-icon toc-header-icon" for="__toc">
<span class="icon"><svg><use href="#svg-toc"></use></svg></span>
</label>
</div>
</header>
<aside class="sidebar-drawer">
<div class="sidebar-container">
<div class="sidebar-sticky"><a class="sidebar-brand" href="index.html">
<span class="sidebar-brand-text">SPFlow 1.0.0</span>
</a><form class="sidebar-search-container" method="get" action="search.html" role="search">
<input class="sidebar-search" placeholder="Search" name="q" aria-label="Search">
<input type="hidden" name="check_keywords" value="yes">
<input type="hidden" name="area" value="default">
</form>
<div id="searchbox"></div><div class="sidebar-scroll"><div class="sidebar-tree">
<ul>
<li class="toctree-l1"><a class="reference internal" href="getting_started.html">Getting Started</a></li>
<li class="toctree-l1"><a class="reference internal" href="concepts.html">Concepts</a></li>
</ul>
<ul>
<li class="toctree-l1 has-children"><a class="reference internal" href="guides/index.html">Guides</a><input aria-label="Toggle navigation of Guides" class="toctree-checkbox" id="toctree-checkbox-1" name="toctree-checkbox-1" role="switch" type="checkbox"/><label for="toctree-checkbox-1"><span class="icon"><svg><use href="#svg-arrow-right"></use></svg></span></label><ul>
<li class="toctree-l2"><a class="reference internal" href="guides/user_guide.html">User Guide</a></li>
<li class="toctree-l2"><a class="reference internal" href="guides/dev_guide.html">Developer Guide</a></li>
<li class="toctree-l2"><a class="reference internal" href="guides/apc_mnist_guide.html">APC MNIST Training Example</a></li>
<li class="toctree-l2"><a class="reference internal" href="guides/sklearn.html">Using SPFlow with sklearn</a></li>
</ul>
</li>
<li class="toctree-l1 has-children"><a class="reference internal" href="zoo/index.html">Paper Zoo</a><input aria-label="Toggle navigation of Paper Zoo" class="toctree-checkbox" id="toctree-checkbox-2" name="toctree-checkbox-2" role="switch" type="checkbox"/><label for="toctree-checkbox-2"><span class="icon"><svg><use href="#svg-arrow-right"></use></svg></span></label><ul>
<li class="toctree-l2"><a class="reference internal" href="zoo/apc.html">Autoencoding Probabilistic Circuits (APC)</a></li>
<li class="toctree-l2"><a class="reference internal" href="zoo/rat_spn.html">Random and Tensorized Sum-Product Networks (RAT-SPN)</a></li>
<li class="toctree-l2"><a class="reference internal" href="zoo/einet.html">Einsum Networks (Einet)</a></li>
<li class="toctree-l2"><a class="reference internal" href="zoo/hclt.html">Hidden Chow-Liu Trees (HCLT)</a></li>
<li class="toctree-l2"><a class="reference internal" href="zoo/conv_pc.html">Convolutional Probabilistic Circuits (ConvPc)</a></li>
<li class="toctree-l2"><a class="reference internal" href="zoo/cms.html">Continuous Mixtures (CMs)</a></li>
<li class="toctree-l2"><a class="reference internal" href="zoo/socs.html">Sum of Compatible Squares (SOCS)</a></li>
<li class="toctree-l2"><a class="reference internal" href="zoo/pic.html">Probabilistic Integral Circuits (PICs)</a></li>
</ul>
</li>
</ul>
<ul class="current">
<li class="toctree-l1 has-children"><a class="reference internal" href="api/index.html">API Documentation</a><input aria-label="Toggle navigation of API Documentation" class="toctree-checkbox" id="toctree-checkbox-3" name="toctree-checkbox-3" role="switch" type="checkbox"/><label for="toctree-checkbox-3"><span class="icon"><svg><use href="#svg-arrow-right"></use></svg></span></label><ul>
<li class="toctree-l2"><a class="reference internal" href="api/base_modules.html">Base Classes</a></li>
<li class="toctree-l2"><a class="reference internal" href="api/data.html">Data Structures</a></li>
<li class="toctree-l2"><a class="reference internal" href="api/module_shape.html">Module Shape</a></li>
<li class="toctree-l2"><a class="reference internal" href="api/dsl.html">DSL (Example Construction)</a></li>
<li class="toctree-l2"><a class="reference internal" href="api/sums.html">Sum Modules</a></li>
<li class="toctree-l2"><a class="reference internal" href="api/products.html">Product Modules</a></li>
<li class="toctree-l2"><a class="reference internal" href="api/conv.html">Convolutional Modules</a></li>
<li class="toctree-l2"><a class="reference internal" href="api/leaves.html">Leaf Modules</a></li>
<li class="toctree-l2"><a class="reference internal" href="api/operations.html">Operations</a></li>
<li class="toctree-l2"><a class="reference internal" href="api/learning.html">Learning and Training</a></li>
<li class="toctree-l2"><a class="reference internal" href="api/scope.html">Scope Management</a></li>
<li class="toctree-l2"><a class="reference internal" href="api/utilities.html">Utilities</a></li>
<li class="toctree-l2"><a class="reference internal" href="api/measures.html">Measures</a></li>
<li class="toctree-l2"><a class="reference internal" href="api/wrappers.html">Wrapper Modules</a></li>
<li class="toctree-l2"><a class="reference internal" href="api/interfaces.html">Interfaces</a></li>
<li class="toctree-l2"><a class="reference internal" href="api/exceptions.html">Exceptions</a></li>
</ul>
</li>
<li class="toctree-l1 current current-page"><a class="current reference internal" href="#">Frequently Asked Questions</a></li>
</ul>
</div>
</div>
</div>
</div>
</aside>
<div class="main">
<div class="content">
<div class="article-container">
<a href="#" class="back-to-top muted-link">
<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 24 24">
<path d="M13 20h-2V8l-5.5 5.5-1.42-1.42L12 4.16l7.92 7.92-1.42 1.42L13 8v12z"></path>
</svg>
<span>Back to top</span>
</a>
<div class="content-icon-container">
<div class="view-this-page">
<a class="muted-link" href="_sources/faq.rst.txt" title="View this page">
<svg><use href="#svg-eye"></use></svg>
<span class="visually-hidden">View this page</span>
</a>
</div>
<div class="theme-toggle-container theme-toggle-content">
<button class="theme-toggle" aria-label="Toggle Light / Dark / Auto color theme">
<svg class="theme-icon-when-auto-light"><use href="#svg-sun-with-moon"></use></svg>
<svg class="theme-icon-when-auto-dark"><use href="#svg-moon-with-sun"></use></svg>
<svg class="theme-icon-when-dark"><use href="#svg-moon"></use></svg>
<svg class="theme-icon-when-light"><use href="#svg-sun"></use></svg>
</button>
</div>
<label class="toc-overlay-icon toc-content-icon" for="__toc">
<span class="icon"><svg><use href="#svg-toc"></use></svg></span>
</label>
</div>
<article role="main" id="furo-main-content">
<section id="frequently-asked-questions">
<h1>Frequently Asked Questions<a class="headerlink" href="#frequently-asked-questions" title="Link to this heading">¶</a></h1>
<p>This page answers common questions about SPFlow. For deeper explanations, see <a class="reference internal" href="concepts.html"><span class="doc">Concepts</span></a>.
For end-to-end tutorials, see the <a class="reference internal" href="guides/user_guide.html"><span class="doc">User Guide</span></a>.</p>
<hr class="docutils" />
<section id="general-questions">
<h2>General Questions<a class="headerlink" href="#general-questions" title="Link to this heading">¶</a></h2>
<section id="what-is-spflow">
<h3>What is SPFlow?<a class="headerlink" href="#what-is-spflow" title="Link to this heading">¶</a></h3>
<p>SPFlow is a Python library for building and learning <strong>Probabilistic Circuits (PCs)</strong>, including Sum-Product Networks (SPNs). These are deep generative and discriminative models that enable tractable (polynomial-time) probabilistic inference while maintaining expressive power.</p>
<p>SPFlow is built on <a class="reference external" href="https://pytorch.org/">PyTorch</a>, providing GPU acceleration and seamless integration with modern deep learning workflows.</p>
</section>
<section id="what-version-of-python-is-required">
<h3>What version of Python is required?<a class="headerlink" href="#what-version-of-python-is-required" title="Link to this heading">¶</a></h3>
<p>SPFlow requires <strong>Python 3.10+</strong> and PyTorch 2.0+.</p>
</section>
<section id="how-do-i-install-spflow">
<h3>How do I install SPFlow?<a class="headerlink" href="#how-do-i-install-spflow" title="Link to this heading">¶</a></h3>
<p>See the <a class="reference internal" href="getting_started.html"><span class="doc">Getting Started</span></a> guide for installation instructions. Quick summary:</p>
<div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="n">pip</span> <span class="n">install</span> <span class="n">spflow</span>
</pre></div>
</div>
</section>
</section>
<hr class="docutils" />
<section id="architecture-concepts">
<h2>Architecture & Concepts<a class="headerlink" href="#architecture-concepts" title="Link to this heading">¶</a></h2>
<section id="what-are-the-main-module-types-in-spflow">
<h3>What are the main module types in SPFlow?<a class="headerlink" href="#what-are-the-main-module-types-in-spflow" title="Link to this heading">¶</a></h3>
<p>SPFlow provides several core module types:</p>
<ul class="simple">
<li><p><strong>Leaves</strong>: Probability distributions at the terminals (Normal, Categorical, Bernoulli, etc.)</p></li>
<li><p><strong>Products</strong>: Combine independent distributions (Product, OuterProduct, ElementwiseProduct)</p></li>
<li><p><strong>Sums</strong>: Weighted mixtures of distributions (Sum, ElementwiseSum)</p></li>
<li><p><strong>Specialized architectures</strong>: RAT-SPN, ConvPc for images (see <a class="reference internal" href="zoo/index.html"><span class="doc">Paper Zoo</span></a>)</p></li>
</ul>
<p>See the <a class="reference internal" href="api/index.html"><span class="doc">API Reference</span></a> for complete documentation.</p>
</section>
<section id="what-is-a-scope">
<h3>What is a Scope?<a class="headerlink" href="#what-is-a-scope" title="Link to this heading">¶</a></h3>
<p>A <strong>Scope</strong> defines which input variables (features) a module operates on. Scopes are what make sums/products well-defined
and enforce decomposability/compatibility constraints.</p>
<p>See <a class="reference internal" href="concepts.html#concepts-scopes-and-decomposability"><span class="std std-ref">Scopes and Decomposability</span></a>.</p>
</section>
<section id="what-are-repetitions">
<h3>What are repetitions?<a class="headerlink" href="#what-are-repetitions" title="Link to this heading">¶</a></h3>
<p><strong>Repetitions</strong> are independent parameterizations of the same structure.
They usually show up as <code class="docutils literal notranslate"><span class="pre">R</span></code> in <code class="docutils literal notranslate"><span class="pre">model.to_str()</span></code> and are tracked in module shapes.</p>
<p>See <a class="reference internal" href="concepts.html#concepts-shapes-and-dimensions"><span class="std std-ref">Shapes and Dimensions</span></a>.</p>
</section>
<section id="what-is-the-difference-between-sum-and-elementwisesum">
<h3>What is the difference between Sum and ElementwiseSum?<a class="headerlink" href="#what-is-the-difference-between-sum-and-elementwisesum" title="Link to this heading">¶</a></h3>
<ul class="simple">
<li><p><strong>Sum</strong>: Computes weighted mixtures over all input channels. Output has <code class="docutils literal notranslate"><span class="pre">out_channels</span></code> channels, each being a weighted combination of all input channels.</p></li>
<li><p><strong>ElementwiseSum</strong>: Sums corresponding channels across multiple input modules element-wise. Requires all inputs to have the same scope and channel count.</p></li>
</ul>
</section>
<section id="what-is-the-difference-between-product-outerproduct-and-elementwiseproduct">
<h3>What is the difference between Product, OuterProduct, and ElementwiseProduct?<a class="headerlink" href="#what-is-the-difference-between-product-outerproduct-and-elementwiseproduct" title="Link to this heading">¶</a></h3>
<ul class="simple">
<li><p><strong>Product</strong>: Combines inputs by computing products across all features. The inputs must have disjoint scopes.</p></li>
<li><p><strong>OuterProduct</strong>: Computes the outer product of split inputs. Takes input split into groups and produces all combinations.</p></li>
<li><p><strong>ElementwiseProduct</strong>: Multiplies corresponding elements across multiple input modules. Requires inputs with compatible shapes.</p></li>
</ul>
</section>
</section>
<hr class="docutils" />
<section id="model-building">
<h2>Model Building<a class="headerlink" href="#model-building" title="Link to this heading">¶</a></h2>
<section id="how-do-i-create-a-simple-spn">
<h3>How do I create a simple SPN?<a class="headerlink" href="#how-do-i-create-a-simple-spn" title="Link to this heading">¶</a></h3>
<p>Here’s a minimal example:</p>
<div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="kn">import</span><span class="w"> </span><span class="nn">torch</span>
<span class="kn">from</span><span class="w"> </span><span class="nn">spflow.modules.sums</span><span class="w"> </span><span class="kn">import</span> <span class="n">Sum</span>
<span class="kn">from</span><span class="w"> </span><span class="nn">spflow.modules.products</span><span class="w"> </span><span class="kn">import</span> <span class="n">Product</span>
<span class="kn">from</span><span class="w"> </span><span class="nn">spflow.modules.leaves</span><span class="w"> </span><span class="kn">import</span> <span class="n">Normal</span>
<span class="kn">from</span><span class="w"> </span><span class="nn">spflow.meta</span><span class="w"> </span><span class="kn">import</span> <span class="n">Scope</span>
<span class="c1"># Create leaves for 2 features</span>
<span class="n">scope</span> <span class="o">=</span> <span class="n">Scope</span><span class="p">([</span><span class="mi">0</span><span class="p">,</span> <span class="mi">1</span><span class="p">])</span>
<span class="n">leaves</span> <span class="o">=</span> <span class="n">Normal</span><span class="p">(</span><span class="n">scope</span><span class="o">=</span><span class="n">scope</span><span class="p">,</span> <span class="n">out_channels</span><span class="o">=</span><span class="mi">4</span><span class="p">)</span>
<span class="c1"># Stack product and sum layers</span>
<span class="n">product</span> <span class="o">=</span> <span class="n">Product</span><span class="p">(</span><span class="n">inputs</span><span class="o">=</span><span class="n">leaves</span><span class="p">)</span>
<span class="n">model</span> <span class="o">=</span> <span class="n">Sum</span><span class="p">(</span><span class="n">inputs</span><span class="o">=</span><span class="n">product</span><span class="p">,</span> <span class="n">out_channels</span><span class="o">=</span><span class="mi">1</span><span class="p">)</span>
<span class="c1"># Use the model</span>
<span class="n">data</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">randn</span><span class="p">(</span><span class="mi">32</span><span class="p">,</span> <span class="mi">2</span><span class="p">)</span>
<span class="n">log_ll</span> <span class="o">=</span> <span class="n">model</span><span class="o">.</span><span class="n">log_likelihood</span><span class="p">(</span><span class="n">data</span><span class="p">)</span>
</pre></div>
</div>
<p>See the <a class="reference internal" href="getting_started.html"><span class="doc">Getting Started</span></a> guide for more examples.</p>
</section>
<section id="what-leaf-distributions-are-available">
<h3>What leaf distributions are available?<a class="headerlink" href="#what-leaf-distributions-are-available" title="Link to this heading">¶</a></h3>
<p>SPFlow includes many univariate distributions:</p>
<p><strong>Continuous</strong>: Normal, LogNormal, Exponential, Laplace, Gamma, Uniform</p>
<p><strong>Discrete</strong>: Categorical, Bernoulli, Binomial, Poisson, Geometric, NegativeBinomial, Hypergeometric</p>
<p>See <a class="reference internal" href="api/leaves.html"><span class="doc">Leaf Modules</span></a> for complete documentation.</p>
</section>
<section id="how-do-i-use-rat-spn">
<h3>How do I use RAT-SPN?<a class="headerlink" href="#how-do-i-use-rat-spn" title="Link to this heading">¶</a></h3>
<p>RAT-SPN (Randomized And Tensorized SPN) automatically builds a deep circuit from hyperparameters:</p>
<div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="kn">from</span><span class="w"> </span><span class="nn">spflow.zoo.rat</span><span class="w"> </span><span class="kn">import</span> <span class="n">RatSPN</span>
<span class="kn">from</span><span class="w"> </span><span class="nn">spflow.modules.leaves</span><span class="w"> </span><span class="kn">import</span> <span class="n">Normal</span>
<span class="kn">from</span><span class="w"> </span><span class="nn">spflow.meta</span><span class="w"> </span><span class="kn">import</span> <span class="n">Scope</span>
<span class="c1"># Create leaves</span>
<span class="n">scope</span> <span class="o">=</span> <span class="n">Scope</span><span class="p">(</span><span class="nb">list</span><span class="p">(</span><span class="nb">range</span><span class="p">(</span><span class="mi">64</span><span class="p">)))</span>
<span class="n">leaves</span> <span class="o">=</span> <span class="n">Normal</span><span class="p">(</span><span class="n">scope</span><span class="o">=</span><span class="n">scope</span><span class="p">,</span> <span class="n">out_channels</span><span class="o">=</span><span class="mi">4</span><span class="p">,</span> <span class="n">num_repetitions</span><span class="o">=</span><span class="mi">2</span><span class="p">)</span>
<span class="c1"># Build RAT-SPN</span>
<span class="n">model</span> <span class="o">=</span> <span class="n">RatSPN</span><span class="p">(</span>
<span class="n">leaf_modules</span><span class="o">=</span><span class="p">[</span><span class="n">leaves</span><span class="p">],</span>
<span class="n">n_root_nodes</span><span class="o">=</span><span class="mi">1</span><span class="p">,</span>
<span class="n">n_region_nodes</span><span class="o">=</span><span class="mi">8</span><span class="p">,</span>
<span class="n">num_repetitions</span><span class="o">=</span><span class="mi">2</span><span class="p">,</span>
<span class="n">depth</span><span class="o">=</span><span class="mi">3</span>
<span class="p">)</span>
</pre></div>
</div>
<p>See <a class="reference internal" href="zoo/rat_spn.html"><span class="doc">Random and Tensorized Sum-Product Networks (RAT-SPN)</span></a> for details.</p>
</section>
<section id="does-spflow-have-image-specific-modules">
<h3>Does SPFlow have image-specific modules?<a class="headerlink" href="#does-spflow-have-image-specific-modules" title="Link to this heading">¶</a></h3>
<p>Yes! Use the <code class="docutils literal notranslate"><span class="pre">ConvPc</span></code> module for image data with spatial structure:</p>
<div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="kn">import</span><span class="w"> </span><span class="nn">torch</span>
<span class="kn">from</span><span class="w"> </span><span class="nn">spflow.zoo.conv</span><span class="w"> </span><span class="kn">import</span> <span class="n">ConvPc</span>
<span class="kn">from</span><span class="w"> </span><span class="nn">spflow.modules.leaves</span><span class="w"> </span><span class="kn">import</span> <span class="n">Normal</span>
<span class="kn">from</span><span class="w"> </span><span class="nn">spflow.meta</span><span class="w"> </span><span class="kn">import</span> <span class="n">Scope</span>
<span class="c1"># Create leaf layer for 28x28 grayscale images (e.g., MNIST)</span>
<span class="n">height</span><span class="p">,</span> <span class="n">width</span> <span class="o">=</span> <span class="mi">28</span><span class="p">,</span> <span class="mi">28</span>
<span class="n">scope</span> <span class="o">=</span> <span class="n">Scope</span><span class="p">(</span><span class="nb">list</span><span class="p">(</span><span class="nb">range</span><span class="p">(</span><span class="n">height</span> <span class="o">*</span> <span class="n">width</span><span class="p">)))</span>
<span class="n">leaf</span> <span class="o">=</span> <span class="n">Normal</span><span class="p">(</span><span class="n">scope</span><span class="o">=</span><span class="n">scope</span><span class="p">,</span> <span class="n">out_channels</span><span class="o">=</span><span class="mi">8</span><span class="p">,</span> <span class="n">num_repetitions</span><span class="o">=</span><span class="mi">1</span><span class="p">)</span>
<span class="c1"># Build convolutional PC</span>
<span class="n">model</span> <span class="o">=</span> <span class="n">ConvPc</span><span class="p">(</span>
<span class="n">leaf</span><span class="o">=</span><span class="n">leaf</span><span class="p">,</span>
<span class="n">input_height</span><span class="o">=</span><span class="n">height</span><span class="p">,</span>
<span class="n">input_width</span><span class="o">=</span><span class="n">width</span><span class="p">,</span>
<span class="n">depth</span><span class="o">=</span><span class="mi">3</span><span class="p">,</span>
<span class="n">channels</span><span class="o">=</span><span class="mi">16</span><span class="p">,</span>
<span class="n">kernel_size</span><span class="o">=</span><span class="mi">2</span><span class="p">,</span>
<span class="n">num_repetitions</span><span class="o">=</span><span class="mi">1</span><span class="p">,</span>
<span class="p">)</span>
</pre></div>
</div>
<p>For adapting existing models to image data, use <code class="docutils literal notranslate"><span class="pre">ImageWrapper</span></code>:</p>
<div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="kn">from</span><span class="w"> </span><span class="nn">spflow.modules.wrapper</span><span class="w"> </span><span class="kn">import</span> <span class="n">ImageWrapper</span>
<span class="c1"># Wrap any SPFlow model for image data</span>
<span class="n">wrapped</span> <span class="o">=</span> <span class="n">ImageWrapper</span><span class="p">(</span><span class="n">model</span><span class="p">,</span> <span class="n">num_channel</span><span class="o">=</span><span class="mi">1</span><span class="p">,</span> <span class="n">height</span><span class="o">=</span><span class="mi">28</span><span class="p">,</span> <span class="n">width</span><span class="o">=</span><span class="mi">28</span><span class="p">)</span>
<span class="c1"># Now works with 4D tensors: (batch, channels, height, width)</span>
<span class="n">image_data</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">randn</span><span class="p">(</span><span class="mi">32</span><span class="p">,</span> <span class="mi">1</span><span class="p">,</span> <span class="mi">28</span><span class="p">,</span> <span class="mi">28</span><span class="p">)</span>
<span class="n">log_ll</span> <span class="o">=</span> <span class="n">wrapped</span><span class="o">.</span><span class="n">log_likelihood</span><span class="p">(</span><span class="n">image_data</span><span class="p">)</span>
</pre></div>
</div>
<p><code class="docutils literal notranslate"><span class="pre">ConvPc</span></code> currently supports <code class="docutils literal notranslate"><span class="pre">num_repetitions</span> <span class="pre">==</span> <span class="pre">1</span></code> only.</p>
<p>See <a class="reference internal" href="api/conv.html"><span class="doc">Convolutional Modules</span></a>, <a class="reference internal" href="zoo/conv_pc.html"><span class="doc">Convolutional Probabilistic Circuits (ConvPc)</span></a> and <a class="reference internal" href="api/wrappers.html"><span class="doc">Wrapper Modules</span></a> for complete documentation.</p>
</section>
</section>
<hr class="docutils" />
<section id="training-learning">
<h2>Training & Learning<a class="headerlink" href="#training-learning" title="Link to this heading">¶</a></h2>
<section id="how-do-i-train-a-model">
<h3>How do I train a model?<a class="headerlink" href="#how-do-i-train-a-model" title="Link to this heading">¶</a></h3>
<p>SPFlow provides two main training approaches:</p>
<p><strong>Gradient Descent</strong>:</p>
<div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="kn">from</span><span class="w"> </span><span class="nn">torch.utils.data</span><span class="w"> </span><span class="kn">import</span> <span class="n">DataLoader</span><span class="p">,</span> <span class="n">TensorDataset</span>
<span class="kn">from</span><span class="w"> </span><span class="nn">spflow.learn</span><span class="w"> </span><span class="kn">import</span> <span class="n">train_gradient_descent</span>
<span class="n">dataloader</span> <span class="o">=</span> <span class="n">DataLoader</span><span class="p">(</span><span class="n">TensorDataset</span><span class="p">(</span><span class="n">train_data</span><span class="p">),</span> <span class="n">batch_size</span><span class="o">=</span><span class="mi">64</span><span class="p">,</span> <span class="n">shuffle</span><span class="o">=</span><span class="kc">True</span><span class="p">)</span>
<span class="n">train_gradient_descent</span><span class="p">(</span>
<span class="n">model</span><span class="p">,</span>
<span class="n">dataloader</span><span class="p">,</span>
<span class="n">epochs</span><span class="o">=</span><span class="mi">100</span><span class="p">,</span>
<span class="n">lr</span><span class="o">=</span><span class="mf">0.01</span>
<span class="p">)</span>
</pre></div>
</div>
<p><strong>Expectation-Maximization</strong>:</p>
<div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="kn">from</span><span class="w"> </span><span class="nn">spflow.learn</span><span class="w"> </span><span class="kn">import</span> <span class="n">expectation_maximization</span>
<span class="n">expectation_maximization</span><span class="p">(</span><span class="n">model</span><span class="p">,</span> <span class="n">train_data</span><span class="p">,</span> <span class="n">max_steps</span><span class="o">=</span><span class="mi">50</span><span class="p">)</span>
</pre></div>
</div>
</section>
<section id="what-is-the-difference-between-gradient-descent-and-em">
<h3>What is the difference between gradient descent and EM?<a class="headerlink" href="#what-is-the-difference-between-gradient-descent-and-em" title="Link to this heading">¶</a></h3>
<p>Both methods use gradients in SPFlow’s implementation:</p>
<ul class="simple">
<li><p><strong>Gradient Descent</strong>: Standard PyTorch optimization. Suitable for most cases, especially when combined with other neural network components.</p></li>
<li><p><strong>Expectation-Maximization (EM)</strong>: A specialized algorithm that alternates between computing expected sufficient statistics and updating parameters. This is usually more stable and converges faster than gradient descent.</p></li>
</ul>
<p>Choose based on your use case; gradient descent is generally more flexible.</p>
</section>
<section id="how-do-i-use-structure-learning">
<h3>How do I use structure learning?<a class="headerlink" href="#how-do-i-use-structure-learning" title="Link to this heading">¶</a></h3>
<p>Use <code class="docutils literal notranslate"><span class="pre">learn_spn</span></code> to automatically learn circuit structure from data:</p>
<div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="kn">from</span><span class="w"> </span><span class="nn">spflow.learn</span><span class="w"> </span><span class="kn">import</span> <span class="n">learn_spn</span>
<span class="kn">from</span><span class="w"> </span><span class="nn">spflow.modules.leaves</span><span class="w"> </span><span class="kn">import</span> <span class="n">Normal</span>
<span class="kn">from</span><span class="w"> </span><span class="nn">spflow.meta</span><span class="w"> </span><span class="kn">import</span> <span class="n">Scope</span>
<span class="n">scope</span> <span class="o">=</span> <span class="n">Scope</span><span class="p">(</span><span class="nb">list</span><span class="p">(</span><span class="nb">range</span><span class="p">(</span><span class="mi">10</span><span class="p">)))</span>
<span class="n">leaves</span> <span class="o">=</span> <span class="n">Normal</span><span class="p">(</span><span class="n">scope</span><span class="o">=</span><span class="n">scope</span><span class="p">,</span> <span class="n">out_channels</span><span class="o">=</span><span class="mi">4</span><span class="p">)</span>
<span class="n">model</span> <span class="o">=</span> <span class="n">learn_spn</span><span class="p">(</span>
<span class="n">data</span><span class="p">,</span>
<span class="n">leaf_modules</span><span class="o">=</span><span class="n">leaves</span><span class="p">,</span>
<span class="n">out_channels</span><span class="o">=</span><span class="mi">1</span><span class="p">,</span>
<span class="n">min_instances_slice</span><span class="o">=</span><span class="mi">100</span>
<span class="p">)</span>
</pre></div>
</div>
<p>See <a class="reference internal" href="api/learning.html"><span class="doc">Learning and Training</span></a> for details and the <a class="reference internal" href="guides/user_guide.html"><span class="doc">User Guide</span></a> for end-to-end examples.</p>
</section>
</section>
<hr class="docutils" />
<section id="inference-sampling">
<h2>Inference & Sampling<a class="headerlink" href="#inference-sampling" title="Link to this heading">¶</a></h2>
<section id="how-do-i-compute-log-likelihood">
<h3>How do I compute log-likelihood?<a class="headerlink" href="#how-do-i-compute-log-likelihood" title="Link to this heading">¶</a></h3>
<p>Call the <code class="docutils literal notranslate"><span class="pre">log_likelihood</span></code> method on your model:</p>
<div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="n">log_likelihood</span> <span class="o">=</span> <span class="n">model</span><span class="o">.</span><span class="n">log_likelihood</span><span class="p">(</span><span class="n">data</span><span class="p">)</span>
<span class="c1"># Returns tensor of shape (batch_size, features, channels, repetitions)</span>
</pre></div>
</div>
</section>
<section id="how-do-i-sample-from-a-model">
<h3>How do I sample from a model?<a class="headerlink" href="#how-do-i-sample-from-a-model" title="Link to this heading">¶</a></h3>
<p>Use the <code class="docutils literal notranslate"><span class="pre">sample</span></code> method:</p>
<div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="c1"># Generate 100 unconditional samples</span>
<span class="n">samples</span> <span class="o">=</span> <span class="n">model</span><span class="o">.</span><span class="n">sample</span><span class="p">(</span><span class="n">num_samples</span><span class="o">=</span><span class="mi">100</span><span class="p">)</span>
</pre></div>
</div>
<p>For conditional sampling with evidence, use <code class="docutils literal notranslate"><span class="pre">sample_with_evidence</span></code>:</p>
<div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="c1"># Sample some features given others</span>
<span class="n">evidence</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">full</span><span class="p">((</span><span class="mi">10</span><span class="p">,</span> <span class="n">num_features</span><span class="p">),</span> <span class="nb">float</span><span class="p">(</span><span class="s1">'nan'</span><span class="p">))</span>
<span class="n">evidence</span><span class="p">[:,</span> <span class="mi">0</span><span class="p">]</span> <span class="o">=</span> <span class="mf">0.5</span> <span class="c1"># Condition on feature 0</span>
<span class="n">samples</span> <span class="o">=</span> <span class="n">model</span><span class="o">.</span><span class="n">sample_with_evidence</span><span class="p">(</span><span class="n">evidence</span><span class="o">=</span><span class="n">evidence</span><span class="p">)</span>
</pre></div>
</div>
</section>
<section id="what-is-mpe-most-probable-explanation-sampling">
<h3>What is MPE (Most Probable Explanation) sampling?<a class="headerlink" href="#what-is-mpe-most-probable-explanation-sampling" title="Link to this heading">¶</a></h3>
<p>MPE returns the most probable state of the model, useful for generating
clearer outputs and validating training. This is also known as MAP (Maximum
A Posteriori) sampling.</p>
<div class="admonition note">
<p class="admonition-title">Note</p>
<p>MAP sampling is different from MMAP (Marginal MAP) sampling, which
marginalizes over some variables while maximizing over others.</p>
</div>
<div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="c1"># Get the most probable sample</span>
<span class="n">mpe_sample</span> <span class="o">=</span> <span class="n">model</span><span class="o">.</span><span class="n">sample</span><span class="p">(</span><span class="n">num_samples</span><span class="o">=</span><span class="mi">1</span><span class="p">,</span> <span class="n">is_mpe</span><span class="o">=</span><span class="kc">True</span><span class="p">)</span>
</pre></div>
</div>
<p>MPE can be combined with evidence for conditional MPE:</p>
<div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="n">evidence</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">full</span><span class="p">((</span><span class="mi">1</span><span class="p">,</span> <span class="n">num_features</span><span class="p">),</span> <span class="nb">float</span><span class="p">(</span><span class="s1">'nan'</span><span class="p">))</span>
<span class="n">evidence</span><span class="p">[</span><span class="mi">0</span><span class="p">,</span> <span class="p">:</span><span class="mi">10</span><span class="p">]</span> <span class="o">=</span> <span class="n">observed_values</span><span class="p">[:</span><span class="mi">10</span><span class="p">]</span> <span class="c1"># Condition on first 10 features</span>
<span class="n">conditional_mpe</span> <span class="o">=</span> <span class="n">model</span><span class="o">.</span><span class="n">mpe</span><span class="p">(</span><span class="n">data</span><span class="o">=</span><span class="n">evidence</span><span class="p">)</span>
</pre></div>
</div>
</section>
<section id="how-do-i-handle-missing-data">
<h3>How do I handle missing data?<a class="headerlink" href="#how-do-i-handle-missing-data" title="Link to this heading">¶</a></h3>
<p>Use <code class="docutils literal notranslate"><span class="pre">torch.nan</span></code> in your data/evidence tensor to indicate missing values:</p>
<div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="c1"># Create data with missing values</span>
<span class="n">data</span> <span class="o">=</span> <span class="n">torch</span><span class="o">.</span><span class="n">randn</span><span class="p">(</span><span class="mi">100</span><span class="p">,</span> <span class="mi">5</span><span class="p">)</span>
<span class="n">data</span><span class="p">[</span><span class="mi">0</span><span class="p">,</span> <span class="mi">2</span><span class="p">]</span> <span class="o">=</span> <span class="nb">float</span><span class="p">(</span><span class="s1">'nan'</span><span class="p">)</span> <span class="c1"># Feature 2 is missing for sample 0</span>
<span class="n">data</span><span class="p">[</span><span class="mi">1</span><span class="p">,</span> <span class="mi">0</span><span class="p">:</span><span class="mi">2</span><span class="p">]</span> <span class="o">=</span> <span class="nb">float</span><span class="p">(</span><span class="s1">'nan'</span><span class="p">)</span> <span class="c1"># Features 0-1 missing for sample 1</span>
<span class="c1"># Log-likelihood handles missing data automatically</span>
<span class="n">log_ll</span> <span class="o">=</span> <span class="n">model</span><span class="o">.</span><span class="n">log_likelihood</span><span class="p">(</span><span class="n">data</span><span class="p">)</span>
</pre></div>
</div>
<p>SPFlow will marginalize over missing features when computing likelihoods.</p>
<p>See <a class="reference internal" href="concepts.html#concepts-missing-data-and-evidence"><span class="std std-ref">Missing Data and Evidence</span></a> for details and conditional sampling patterns.</p>
</section>
</section>
<hr class="docutils" />
<section id="visualization-debugging">
<h2>Visualization & Debugging<a class="headerlink" href="#visualization-debugging" title="Link to this heading">¶</a></h2>
<section id="how-do-i-visualize-a-circuit">
<h3>How do I visualize a circuit?<a class="headerlink" href="#how-do-i-visualize-a-circuit" title="Link to this heading">¶</a></h3>
<p>Use the <code class="docutils literal notranslate"><span class="pre">visualize</span></code> function:</p>
<div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="kn">from</span><span class="w"> </span><span class="nn">spflow.utils.visualization</span><span class="w"> </span><span class="kn">import</span> <span class="n">visualize</span>
<span class="n">visualize</span><span class="p">(</span>
<span class="n">model</span><span class="p">,</span>
<span class="n">output_path</span><span class="o">=</span><span class="s2">"/tmp/my_circuit"</span><span class="p">,</span>
<span class="nb">format</span><span class="o">=</span><span class="s2">"pdf"</span><span class="p">,</span>
<span class="n">show_scope</span><span class="o">=</span><span class="kc">True</span><span class="p">,</span>
<span class="n">show_shape</span><span class="o">=</span><span class="kc">True</span>
<span class="p">)</span>
</pre></div>
</div>
<p>Requires <a class="reference external" href="https://graphviz.org/">Graphviz</a> to be installed on your system.</p>
</section>
<section id="what-output-formats-are-supported">
<h3>What output formats are supported?<a class="headerlink" href="#what-output-formats-are-supported" title="Link to this heading">¶</a></h3>
<p>The visualization function supports multiple formats via Graphviz:</p>
<ul class="simple">
<li><p><strong>PDF</strong>: <code class="docutils literal notranslate"><span class="pre">format="pdf"</span></code> (recommended for papers)</p></li>
<li><p><strong>SVG</strong>: <code class="docutils literal notranslate"><span class="pre">format="svg"</span></code> (scalable, good for web)</p></li>
<li><p><strong>PNG</strong>: <code class="docutils literal notranslate"><span class="pre">format="png"</span></code> (raster image)</p></li>
</ul>
</section>
<section id="how-do-i-print-the-model-structure">
<h3>How do I print the model structure?<a class="headerlink" href="#how-do-i-print-the-model-structure" title="Link to this heading">¶</a></h3>
<p>Use the <code class="docutils literal notranslate"><span class="pre">to_str()</span></code> method for a text representation:</p>
<div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="nb">print</span><span class="p">(</span><span class="n">model</span><span class="o">.</span><span class="n">to_str</span><span class="p">())</span>
<span class="c1"># Example output:</span>
<span class="c1"># Sum [D=1, C=1] [weights: (1, 4, 1)] → scope: 0-1</span>
<span class="c1"># └─ Product [D=1, C=4] → scope: 0-1</span>
<span class="c1"># └─ Normal [D=2, C=4] → scope: 0-1</span>
</pre></div>
</div>
</section>
<section id="how-do-i-log-model-complexity-nodes-edges-parameters">
<h3>How do I log model complexity (nodes/edges/parameters)?<a class="headerlink" href="#how-do-i-log-model-complexity-nodes-edges-parameters" title="Link to this heading">¶</a></h3>
<p>Use <code class="docutils literal notranslate"><span class="pre">get_structure_stats</span></code> to compute deterministic structure statistics:</p>
<div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="kn">from</span><span class="w"> </span><span class="nn">spflow.utils.structure_stats</span><span class="w"> </span><span class="kn">import</span> <span class="n">get_structure_stats</span>
<span class="n">stats</span> <span class="o">=</span> <span class="n">get_structure_stats</span><span class="p">(</span><span class="n">model</span><span class="p">)</span>
<span class="nb">print</span><span class="p">(</span><span class="n">stats</span><span class="o">.</span><span class="n">num_nodes_total</span><span class="p">,</span> <span class="n">stats</span><span class="o">.</span><span class="n">num_edges_total</span><span class="p">,</span> <span class="n">stats</span><span class="o">.</span><span class="n">num_parameters_total</span><span class="p">)</span>
</pre></div>
</div>
<p>For a short text overview (similar to <code class="docutils literal notranslate"><span class="pre">to_str()</span></code>), use:</p>
<div class="highlight-default notranslate"><div class="highlight"><pre><span></span><span class="nb">print</span><span class="p">(</span><span class="n">model</span><span class="o">.</span><span class="n">print_structure_stats</span><span class="p">())</span>
</pre></div>
</div>
<p>The traversal matches <code class="docutils literal notranslate"><span class="pre">model.to_str()</span></code> (it skips internal <code class="docutils literal notranslate"><span class="pre">Cat</span></code>/<code class="docutils literal notranslate"><span class="pre">ModuleList</span></code> wrappers), is DAG-aware,
and counts parameters uniquely across shared subgraphs.</p>
</section>
<hr class="docutils" />
<section id="what-s-the-difference-between-spflow-v1-x-and-the-legacy-version">
<h3>What’s the difference between SPFlow v1.x and the legacy version?<a class="headerlink" href="#what-s-the-difference-between-spflow-v1-x-and-the-legacy-version" title="Link to this heading">¶</a></h3>
<p>SPFlow v1.0 is a <strong>complete rewrite</strong> using PyTorch as the primary backend. Key differences:</p>
<ul class="simple">
<li><p>Modern PyTorch architecture for GPU acceleration</p></li>
<li><p>Significantly improved performance</p></li>
<li><p>Enhanced modular design with composable layers</p></li>
</ul>
<p>The pre-v1.0.0 version is still available:</p>
<ul class="simple">
<li><p>On PyPI: <code class="docutils literal notranslate"><span class="pre">pip</span> <span class="pre">install</span> <span class="pre">spflow==0.0.48</span></code></p></li>
<li><p>In the <code class="docutils literal notranslate"><span class="pre">legacy</span></code> branch of the GitHub repository</p></li>
</ul>
<p>Models from the legacy version are <strong>not compatible</strong> with v1.x and need to be rebuilt.</p>
</section>
</section>
<section id="migration-from-legacy">
<h2>Migration from Legacy<a class="headerlink" href="#migration-from-legacy" title="Link to this heading">¶</a></h2>
<section id="how-do-i-migrate-from-spflow-0-x-to-1-x">
<h3>How do I migrate from SPFlow 0.x to 1.x?<a class="headerlink" href="#how-do-i-migrate-from-spflow-0-x-to-1-x" title="Link to this heading">¶</a></h3>
<p>SPFlow 1.0 is a complete rewrite. Key changes:</p>
<ol class="arabic simple">
<li><p><strong>PyTorch-based</strong>: All modules are <code class="docutils literal notranslate"><span class="pre">nn.Module</span></code> subclasses</p></li>
<li><p><strong>Layered composition</strong>: Build circuits by stacking modules</p></li>
<li><p><strong>New API</strong>: Method names and signatures have changed</p></li>
<li><p><strong>GPU support</strong>: Native CUDA acceleration</p></li>
</ol>
<p>There is no automatic migration path. You will need to:</p>
<ol class="arabic simple">
<li><p>Reinstall: <code class="docutils literal notranslate"><span class="pre">pip</span> <span class="pre">install</span> <span class="pre">--upgrade</span> <span class="pre">spflow</span></code> (uninstall legacy first if needed)</p></li>
<li><p>Rebuild your models using the new API</p></li>
<li><p>Retrain your models</p></li>
</ol>
</section>
<section id="are-old-models-compatible-with-spflow-v1-x">
<h3>Are old models compatible with SPFlow v1.x?<a class="headerlink" href="#are-old-models-compatible-with-spflow-v1-x" title="Link to this heading">¶</a></h3>
<p><strong>No.</strong> Models saved with SPFlow 0.x cannot be loaded in SPFlow 1.x due to the complete architectural rewrite.</p>
<p>You must rebuild and retrain your models using the new API. See the <a class="reference internal" href="guides/user_guide.html"><span class="doc">User Guide</span></a> for comprehensive examples.</p>
</section>
</section>
</section>
</article>
</div>
<footer>
<div class="related-pages">
<a class="prev-page" href="api/exceptions.html">
<svg class="furo-related-icon"><use href="#svg-arrow-right"></use></svg>
<div class="page-info">
<div class="context">
<span>Previous</span>
</div>
<div class="title">Exceptions</div>
</div>
</a>
</div>
<div class="bottom-of-page">
<div class="left-details">
<div class="copyright">
Copyright © 2025, SPFlow Contributors
</div>
Made with <a href="https://www.sphinx-doc.org/">Sphinx</a> and <a class="muted-link" href="https://pradyunsg.me">@pradyunsg</a>'s
<a href="https://github.com/pradyunsg/furo">Furo</a>
</div>
<div class="right-details">
</div>
</div>
</footer>
</div>
<aside class="toc-drawer">
<div class="toc-sticky toc-scroll">
<div class="toc-title-container">
<span class="toc-title">
On this page
</span>
</div>
<div class="toc-tree-container">
<div class="toc-tree">
<ul>
<li><a class="reference internal" href="#">Frequently Asked Questions</a><ul>
<li><a class="reference internal" href="#general-questions">General Questions</a><ul>
<li><a class="reference internal" href="#what-is-spflow">What is SPFlow?</a></li>
<li><a class="reference internal" href="#what-version-of-python-is-required">What version of Python is required?</a></li>
<li><a class="reference internal" href="#how-do-i-install-spflow">How do I install SPFlow?</a></li>
</ul>
</li>
<li><a class="reference internal" href="#architecture-concepts">Architecture & Concepts</a><ul>
<li><a class="reference internal" href="#what-are-the-main-module-types-in-spflow">What are the main module types in SPFlow?</a></li>
<li><a class="reference internal" href="#what-is-a-scope">What is a Scope?</a></li>
<li><a class="reference internal" href="#what-are-repetitions">What are repetitions?</a></li>
<li><a class="reference internal" href="#what-is-the-difference-between-sum-and-elementwisesum">What is the difference between Sum and ElementwiseSum?</a></li>
<li><a class="reference internal" href="#what-is-the-difference-between-product-outerproduct-and-elementwiseproduct">What is the difference between Product, OuterProduct, and ElementwiseProduct?</a></li>
</ul>
</li>
<li><a class="reference internal" href="#model-building">Model Building</a><ul>
<li><a class="reference internal" href="#how-do-i-create-a-simple-spn">How do I create a simple SPN?</a></li>
<li><a class="reference internal" href="#what-leaf-distributions-are-available">What leaf distributions are available?</a></li>
<li><a class="reference internal" href="#how-do-i-use-rat-spn">How do I use RAT-SPN?</a></li>
<li><a class="reference internal" href="#does-spflow-have-image-specific-modules">Does SPFlow have image-specific modules?</a></li>
</ul>
</li>
<li><a class="reference internal" href="#training-learning">Training & Learning</a><ul>
<li><a class="reference internal" href="#how-do-i-train-a-model">How do I train a model?</a></li>
<li><a class="reference internal" href="#what-is-the-difference-between-gradient-descent-and-em">What is the difference between gradient descent and EM?</a></li>
<li><a class="reference internal" href="#how-do-i-use-structure-learning">How do I use structure learning?</a></li>
</ul>
</li>
<li><a class="reference internal" href="#inference-sampling">Inference & Sampling</a><ul>
<li><a class="reference internal" href="#how-do-i-compute-log-likelihood">How do I compute log-likelihood?</a></li>
<li><a class="reference internal" href="#how-do-i-sample-from-a-model">How do I sample from a model?</a></li>
<li><a class="reference internal" href="#what-is-mpe-most-probable-explanation-sampling">What is MPE (Most Probable Explanation) sampling?</a></li>
<li><a class="reference internal" href="#how-do-i-handle-missing-data">How do I handle missing data?</a></li>
</ul>
</li>
<li><a class="reference internal" href="#visualization-debugging">Visualization & Debugging</a><ul>
<li><a class="reference internal" href="#how-do-i-visualize-a-circuit">How do I visualize a circuit?</a></li>
<li><a class="reference internal" href="#what-output-formats-are-supported">What output formats are supported?</a></li>
<li><a class="reference internal" href="#how-do-i-print-the-model-structure">How do I print the model structure?</a></li>
<li><a class="reference internal" href="#how-do-i-log-model-complexity-nodes-edges-parameters">How do I log model complexity (nodes/edges/parameters)?</a></li>
<li><a class="reference internal" href="#what-s-the-difference-between-spflow-v1-x-and-the-legacy-version">What’s the difference between SPFlow v1.x and the legacy version?</a></li>
</ul>
</li>
<li><a class="reference internal" href="#migration-from-legacy">Migration from Legacy</a><ul>
<li><a class="reference internal" href="#how-do-i-migrate-from-spflow-0-x-to-1-x">How do I migrate from SPFlow 0.x to 1.x?</a></li>
<li><a class="reference internal" href="#are-old-models-compatible-with-spflow-v1-x">Are old models compatible with SPFlow v1.x?</a></li>
</ul>
</li>
</ul>
</li>
</ul>
</div>
</div>
</div>
</aside>
</div>
</div><script src="_static/documentation_options.js?v=8d563738"></script>
<script src="_static/doctools.js?v=fd6eb6e6"></script>
<script src="_static/sphinx_highlight.js?v=6ffebe34"></script>
<script src="_static/scripts/furo.js?v=46bd48cc"></script>
<script src="_static/clipboard.min.js?v=a7894cd8"></script>
<script src="_static/copybutton.js?v=f281be69"></script>
<script crossorigin="anonymous" integrity="sha256-Ae2Vz/4ePdIu6ZyI/5ZGsYnb+m0JlOmKPjt6XZ9JJkA=" src="https://cdnjs.cloudflare.com/ajax/libs/require.js/2.3.4/require.min.js"></script>
</body>
</html>