28:. Instead of drawing each repeated variable individually, a plate or rectangle is used to group variables into a subgraph that repeat together, and a number is drawn on the plate to represent the number of repetitions of the subgraph in the plate. The assumptions are that the subgraph is duplicated that many times, the variables in the subgraph are indexed by the repetition number, and any links that cross a plate boundary are replicated once for each subgraph repetition.
357:
37:
403:
A number of extensions have been created by various authors to express more information than simply the conditional relationships. However, few of these have become standard. Perhaps the most commonly used extension is to use rectangles in place of circles to indicate non-random variables—either
443:
Categorical variables that act as "switches", and which pick one or more other random variables to condition on from a large set of such variables (e.g. mixture components), are indicated with a special type of arrow containing a squiggly line and ending in a T
367:
using plate notation. Smaller squares indicate fixed parameters; larger circles indicate random variables. Filled-in shapes indicate known values. The indication means a vector of size
95:
342:
312:
274:
206:
168:
240:
134:
526:
465:
434:
are similarly indicated by putting the matrix size in brackets in the middle of the node, with commas separating row size from column size.
361:
609:
48:
41:
604:
415:
The diagram on the right shows a few more non-standard conventions used in some articles in
Knowledge (XXG) (e.g.
599:
395:— the value of this variable selects, for the other incoming variables, which value to use out of the size-
60:
518:
464:
drawing packages, but also as part of graphical user interfaces to
Bayesian statistics programs such as
55:
that models how documents in a corpus are topically related. There are two variables not in any plate;
284:. The directed edges between variables indicate dependencies between the variables: for example, each
545:
437:
380:
109:
times, once for each document. The inner plate represents the variables associated with each of the
277:
569:
535:
73:
17:
469:
356:
561:
416:
36:
553:
317:
287:
249:
181:
143:
52:
218:
112:
70:
The outermost plate represents all the variables related to a specific document, including
431:
409:
281:
25:
549:
405:
593:
447:
Boldface is consistently used for vector or matrix nodes (but not categorical nodes).
424:
364:
573:
493:
412:), or variables whose values are computed deterministically from a random variable.
67:
is the parameter of the uniform
Dirichlet prior on the per-topic word distribution.
440:
are indicated by placing their size (without a bracket) in the middle of the node.
246:. The circle representing the individual words is shaded, indicating that each
427:
are indicated by putting the vector size in brackets in the middle of the node.
565:
215:
in the corner represents the repetition of the variables in the inner plate
105:
in the corner of the plate indicates that the variables inside are repeated
280:, and the other circles are empty, indicating that the other variables are
557:
540:
461:
355:
35:
473:
457:
24:
is a method of representing variables that repeat in a
320:
290:
252:
221:
184:
146:
115:
76:
63:prior on the per-document topic distributions, and
336:
306:
268:
234:
200:
162:
128:
89:
519:"Operations for Learning with Graphical Models"
456:Plate notation has been implemented in various
8:
527:Journal of Artificial Intelligence Research
539:
408:given a fixed value (or computed through
387:outcomes. The squiggly line coming from
325:
319:
295:
289:
257:
251:
226:
220:
189:
183:
151:
145:
120:
114:
81:
75:
484:
242:times, once for each word in document
97:, the topic distribution for document
7:
517:Buntine, Wray L. (December 1994).
492:Ghahramani, Zoubin (August 2007).
170:is the topic distribution for the
14:
534:. AI Access Foundation: 159–225.
391:ending in a crossbar indicates a
59:is the parameter of the uniform
1:
47:In this example, we consider
430:Variables that are actually
423:Variables that are actually
498:(Speech). TĂĽbingen, Germany
404:parameters to be computed,
90:{\displaystyle \theta _{i}}
49:Latent Dirichlet allocation
42:Latent Dirichlet allocation
626:
371:; means a matrix of size
399:array of possible values.
208:is the actual word used.
452:Software implementation
400:
338:
337:{\displaystyle z_{ij}}
308:
307:{\displaystyle w_{ij}}
270:
269:{\displaystyle w_{ij}}
236:
202:
201:{\displaystyle w_{ij}}
164:
163:{\displaystyle z_{ij}}
130:
91:
44:
610:Mathematical notation
438:Categorical variables
362:multivariate Gaussian
359:
339:
309:
271:
237:
235:{\displaystyle N_{j}}
203:
165:
131:
129:{\displaystyle N_{i}}
92:
39:
381:categorical variable
318:
288:
250:
219:
182:
174:th word in document
144:
113:
74:
550:1994cs.......12102B
40:Plate notation for
401:
379:; K alone means a
334:
304:
266:
232:
198:
160:
136:words in document
126:
87:
45:
18:Bayesian inference
605:Bayesian networks
417:variational Bayes
617:
600:Graphical models
585:
584:
582:
580:
543:
523:
514:
508:
507:
505:
503:
495:Graphical models
489:
343:
341:
340:
335:
333:
332:
313:
311:
310:
305:
303:
302:
282:latent variables
275:
273:
272:
267:
265:
264:
241:
239:
238:
233:
231:
230:
207:
205:
204:
199:
197:
196:
169:
167:
166:
161:
159:
158:
135:
133:
132:
127:
125:
124:
96:
94:
93:
88:
86:
85:
53:Bayesian network
625:
624:
620:
619:
618:
616:
615:
614:
590:
589:
588:
578:
576:
558:10.1613/jair.62
521:
516:
515:
511:
501:
499:
491:
490:
486:
482:
454:
432:random matrices
410:empirical Bayes
406:hyperparameters
354:
321:
316:
315:
291:
286:
285:
253:
248:
247:
222:
217:
216:
185:
180:
179:
147:
142:
141:
116:
111:
110:
77:
72:
71:
34:
26:graphical model
12:
11:
5:
623:
621:
613:
612:
607:
602:
592:
591:
587:
586:
509:
483:
481:
478:
453:
450:
449:
448:
445:
441:
435:
428:
425:random vectors
353:
350:
331:
328:
324:
301:
298:
294:
263:
260:
256:
229:
225:
195:
192:
188:
157:
154:
150:
123:
119:
84:
80:
33:
30:
22:plate notation
13:
10:
9:
6:
4:
3:
2:
622:
611:
608:
606:
603:
601:
598:
597:
595:
575:
571:
567:
563:
559:
555:
551:
547:
542:
537:
533:
529:
528:
520:
513:
510:
497:
496:
488:
485:
479:
477:
475:
471:
467:
463:
459:
451:
446:
442:
439:
436:
433:
429:
426:
422:
421:
420:
418:
413:
411:
407:
398:
394:
390:
386:
382:
378:
374:
370:
366:
365:mixture model
363:
358:
351:
349:
347:
329:
326:
322:
299:
296:
292:
283:
279:
261:
258:
254:
245:
227:
223:
214:
209:
193:
190:
186:
177:
173:
155:
152:
148:
139:
121:
117:
108:
104:
100:
82:
78:
68:
66:
62:
58:
54:
50:
43:
38:
31:
29:
27:
23:
19:
577:. Retrieved
531:
525:
512:
500:. Retrieved
494:
487:
455:
414:
402:
396:
392:
388:
384:
376:
372:
368:
345:
243:
212:
210:
175:
171:
137:
106:
102:
98:
69:
64:
56:
46:
21:
15:
579:21 February
502:21 February
314:depends on
594:Categories
541:cs/9412102
480:References
470:BayesiaLab
352:Extensions
278:observable
566:1076-9757
444:junction.
360:Bayesian
79:θ
61:Dirichlet
574:11672931
546:Bibcode
101:. The
32:Example
572:
564:
393:switch
375:×
178:, and
570:S2CID
536:arXiv
522:(PDF)
462:LaTeX
383:with
581:2008
562:ISSN
504:2008
474:PyMC
472:and
468:and
466:BUGS
344:and
211:The
51:, a
554:doi
458:TeX
419:):
276:is
16:In
596::
568:.
560:.
552:.
544:.
530:.
524:.
476:.
348:.
140::
20:,
583:.
556::
548::
538::
532:2
506:.
460:/
397:K
389:z
385:K
377:D
373:D
369:K
346:β
330:j
327:i
323:z
300:j
297:i
293:w
262:j
259:i
255:w
244:i
228:j
224:N
213:N
194:j
191:i
187:w
176:i
172:j
156:j
153:i
149:z
138:i
122:i
118:N
107:M
103:M
99:i
83:i
65:β
57:α
Text is available under the Creative Commons Attribution-ShareAlike License. Additional terms may apply.