122:(VAEs): VAEs are generative models that simultaneously learn to encode and decode data. The latent space in VAEs acts as an embedding space. By training VAEs on high-dimensional data, such as images or audio, the model learns to encode the data into a compact latent representation. VAEs are known for their ability to generate new data samples from the learned latent space.
116:: Siamese networks are a type of neural network architecture commonly used for similarity-based embedding. They consist of two identical subnetworks that process two input samples and produce their respective embeddings. Siamese networks are often used for tasks like image similarity, recommendation systems, and face recognition.
81:
high-dimensional, complex, and nonlinear, which may add to the difficulty of interpretation. Some visualization techniques have been developed to connect the latent space to the visual world, but there is often not a direct connection between the latent space interpretation and the model itself. Such techniques include
110:: GloVe (Global Vectors for Word Representation) is another widely used embedding model for NLP. It combines global statistical information from a corpus with local context information to learn word embeddings. GloVe embeddings are known for capturing both semantic and relational similarities between words.
139:
To embed multimodal data, specialized architectures such as deep multimodal networks or multimodal transformers are employed. These architectures combine different types of neural network modules to process and integrate information from various modalities. The resulting embeddings capture the
80:
The interpretation of the latent spaces of machine learning models is an active field of study, but latent space interpretation is difficult to achieve. Due to the black-box nature of machine learning models, the latent space may be completely unintuitive. Additionally, the latent space may be
104:: Word2Vec is a popular embedding model used in natural language processing (NLP). It learns word embeddings by training a neural network on a large corpus of text. Word2Vec captures semantic and syntactic relationships between words, allowing for meaningful computations like word analogies.
131:
Multimodality refers to the integration and analysis of multiple modes or types of data within a single model or framework. Embedding multimodal data involves capturing relationships and interactions between different data types, such as images, text, audio, and structured data.
135:
Multimodal embedding models aim to learn joint representations that fuse information from multiple modalities, allowing for cross-modal analysis and tasks. These models enable applications like image captioning, visual question answering, and multimodal sentiment analysis.
85:(t-SNE), where the latent space is mapped to two dimensions for visualization. Latent space distances lack physical units, so the interpretation of these distances may depend on the application.
167:
Social
Systems: Embedding techniques can be used to learn latent representations of social systems such as internal migration systems, academic citation networks, and world trade networks.
164:
Healthcare: Embedding techniques have been applied to electronic health records, medical imaging, and genomic data for disease prediction, diagnosis, and treatment.
82:
152:
Information
Retrieval: Embedding techniques enable efficient similarity search and recommendation systems by representing data points in a compact space.
97:. These models learn the embeddings by leveraging statistical techniques and machine learning algorithms. Here are some commonly used embedding models:
50:
in which items resembling each other are positioned closer to one another. Position within the latent space can be viewed as being defined by a set of
155:
Natural
Language Processing: Word embeddings have revolutionized NLP tasks like sentiment analysis, machine translation, and document classification.
333:
Arvanitidis, Georgios; Hansen, Lars Kai; Hauberg, SĂžren (13 December 2021). "Latent Space Oddity: on the
Curvature of Deep Generative Models".
93:
Several embedding models have been developed to perform this transformation to create latent space embeddings given a set of data items and a
437:
213:
161:
Recommendation
Systems: Embeddings help capture user preferences and item characteristics, enabling personalized recommendations.
77:, and they can then be used as feature spaces in machine learning models, including classifiers and other supervised predictors.
355:
158:
Computer Vision: Image and video embeddings enable tasks like object recognition, image retrieval, and video summarization.
533:"Investigating internal migration with network analysis and latent space representations: an application to Turkey"
715:
148:
Embedding latent space and multimodal embedding models have found numerous applications across various domains:
193:
66:
203:
119:
113:
198:
140:
complex relationships between different data types, facilitating multimodal analysis and understanding.
710:
590:"Detecting trends in academic research from a citation network using network representation learning"
183:
286:
Li, Ziqiang; Tao, Rentuo; Wang, Jie; Li, Fu; Niu, Hongjing; Yue, Mingdao; Li, Bin (February 2021).
218:
208:
188:
94:
513:
487:
451:
367:
334:
315:
268:
686:
668:
629:
611:
570:
552:
505:
443:
433:
393:
Proceedings of the 2014 Conference on
Empirical Methods in Natural Language Processing (EMNLP)
307:
260:
65:
from which the data points are drawn, making the construction of a latent space an example of
676:
660:
647:
GarcĂa-PĂ©rez, Guillermo; Boguñå, MariĂĄn; Allard, Antoine; Serrano, M. Ăngeles (2016-09-16).
619:
601:
560:
544:
497:
425:
396:
299:
252:
240:
178:
74:
70:
424:, Methods in Molecular Biology, vol. 2190, New York, NY: Springer US, pp. 73â94,
51:
475:
681:
648:
624:
589:
565:
58:
704:
517:
455:
319:
272:
62:
649:"The hidden hyperbolic geometry of international trade: World Trade Atlas 1870â2013"
606:
588:
Asatani, Kimitaka; Mori, Junichiro; Ochi, Masanao; Sakata, Ichiro (2018-05-21).
429:
354:
Mikolov, Tomas; Sutskever, Ilya; Chen, Kai; Corrado, Greg S; Dean, Jeff (2013).
548:
287:
17:
417:
395:. Doha, Qatar: Association for Computational Linguistics. pp. 1532â1543.
672:
615:
556:
532:
509:
356:"Distributed Representations of Words and Phrases and their Compositionality"
311:
303:
264:
43:
690:
633:
574:
447:
387:
Pennington, Jeffrey; Socher, Richard; Manning, Christopher (October 2014).
401:
101:
61:
of the latent space is chosen to be lower than the dimensionality of the
47:
501:
664:
256:
241:"Latent Space Cartography: Visual Analysis of Vector Space Embeddings"
388:
492:
339:
372:
107:
27:
Embedding of data within a manifold based on a similarity function
288:"Interpreting the Latent Space of GANs via Measuring Decoupling"
239:
Liu, Yang; Jun, Eunice; Li, Qisheng; Heer, Jeffrey (June 2019).
54:
that emerge from the resemblances from the objects.
360:Advances in Neural Information Processing Systems
474:Kingma, Diederik P.; Welling, Max (2019-11-27).
416:Chicco, Davide (2021), Cartwright, Hugh (ed.),
389:"Glove: Global Vectors for Word Representation"
476:"An Introduction to Variational Autoencoders"
8:
531:GĂŒrsoy, Furkan; Badur, Bertan (2022-10-06).
292:IEEE Transactions on Artificial Intelligence
83:t-distributed stochastic neighbor embedding
480:Foundations and Trends in Machine Learning
680:
623:
605:
564:
491:
400:
371:
338:
69:, which can also be viewed as a form of
231:
418:"Siamese Neural Networks: An Overview"
7:
73:. Latent spaces are usually fit via
537:Social Network Analysis and Mining
214:Nonlinear dimensionality reduction
25:
1:
607:10.1371/journal.pone.0197260
430:10.1007/978-1-0716-0826-5_3
46:of a set of items within a
732:
549:10.1007/s13278-022-00974-w
422:Artificial Neural Networks
366:. Curran Associates, Inc.
304:10.1109/TAI.2021.3071642
194:Latent semantic analysis
120:Variational Autoencoders
67:dimensionality reduction
245:Computer Graphics Forum
204:Ordination (statistics)
199:Latent variable model
184:Clustering algorithm
36:latent feature space
402:10.3115/v1/D14-1162
219:Self-organizing map
209:Manifold hypothesis
189:Intrinsic dimension
95:similarity function
57:In most cases, the
653:Scientific Reports
502:10.1561/2200000056
34:, also known as a
665:10.1038/srep33441
439:978-1-0716-0826-5
257:10.1111/cgf.13672
16:(Redirected from
723:
716:Cluster analysis
695:
694:
684:
644:
638:
637:
627:
609:
585:
579:
578:
568:
528:
522:
521:
495:
471:
465:
464:
463:
462:
413:
407:
406:
404:
384:
378:
377:
375:
351:
345:
344:
342:
330:
324:
323:
283:
277:
276:
236:
179:Induced topology
114:Siamese Networks
89:Embedding models
75:machine learning
71:data compression
52:latent variables
21:
731:
730:
726:
725:
724:
722:
721:
720:
701:
700:
699:
698:
646:
645:
641:
600:(5): e0197260.
587:
586:
582:
530:
529:
525:
473:
472:
468:
460:
458:
440:
415:
414:
410:
386:
385:
381:
353:
352:
348:
332:
331:
327:
285:
284:
280:
238:
237:
233:
228:
223:
174:
146:
129:
91:
40:embedding space
28:
23:
22:
18:Latent manifold
15:
12:
11:
5:
729:
727:
719:
718:
713:
703:
702:
697:
696:
639:
580:
523:
486:(4): 307â392.
466:
438:
408:
379:
346:
325:
278:
230:
229:
227:
224:
222:
221:
216:
211:
206:
201:
196:
191:
186:
181:
175:
173:
170:
169:
168:
165:
162:
159:
156:
153:
145:
142:
128:
125:
124:
123:
117:
111:
105:
90:
87:
59:dimensionality
26:
24:
14:
13:
10:
9:
6:
4:
3:
2:
728:
717:
714:
712:
709:
708:
706:
692:
688:
683:
678:
674:
670:
666:
662:
658:
654:
650:
643:
640:
635:
631:
626:
621:
617:
613:
608:
603:
599:
595:
591:
584:
581:
576:
572:
567:
562:
558:
554:
550:
546:
542:
538:
534:
527:
524:
519:
515:
511:
507:
503:
499:
494:
489:
485:
481:
477:
470:
467:
457:
453:
449:
445:
441:
435:
431:
427:
423:
419:
412:
409:
403:
398:
394:
390:
383:
380:
374:
369:
365:
361:
357:
350:
347:
341:
336:
329:
326:
321:
317:
313:
309:
305:
301:
297:
293:
289:
282:
279:
274:
270:
266:
262:
258:
254:
250:
246:
242:
235:
232:
225:
220:
217:
215:
212:
210:
207:
205:
202:
200:
197:
195:
192:
190:
187:
185:
182:
180:
177:
176:
171:
166:
163:
160:
157:
154:
151:
150:
149:
143:
141:
137:
133:
127:Multimodality
126:
121:
118:
115:
112:
109:
106:
103:
100:
99:
98:
96:
88:
86:
84:
78:
76:
72:
68:
64:
63:feature space
60:
55:
53:
49:
45:
41:
37:
33:
19:
659:(1): 33441.
656:
652:
642:
597:
593:
583:
540:
536:
526:
483:
479:
469:
459:, retrieved
421:
411:
392:
382:
363:
359:
349:
328:
298:(1): 58â70.
295:
291:
281:
251:(3): 67â78.
248:
244:
234:
147:
144:Applications
138:
134:
130:
92:
79:
56:
39:
35:
32:latent space
31:
29:
711:Data mining
705:Categories
543:(1): 150.
493:1906.02691
461:2023-06-26
340:1710.11379
226:References
673:2045-2322
616:1932-6203
557:1869-5469
518:174802445
510:1935-8237
456:221144012
373:1310.4546
320:234847784
312:2691-4581
273:189858337
265:0167-7055
44:embedding
691:27633649
634:29782521
594:PLOS ONE
575:36246429
448:32804361
172:See also
102:Word2Vec
48:manifold
42:, is an
682:5025783
625:5962067
566:9540093
689:
679:
671:
632:
622:
614:
573:
563:
555:
516:
508:
454:
446:
436:
318:
310:
271:
263:
514:S2CID
488:arXiv
452:S2CID
368:arXiv
335:arXiv
316:S2CID
269:S2CID
108:GloVe
687:PMID
669:ISSN
630:PMID
612:ISSN
571:PMID
553:ISSN
506:ISSN
444:PMID
434:ISBN
308:ISSN
261:ISSN
677:PMC
661:doi
620:PMC
602:doi
561:PMC
545:doi
498:doi
426:doi
397:doi
300:doi
253:doi
38:or
707::
685:.
675:.
667:.
655:.
651:.
628:.
618:.
610:.
598:13
596:.
592:.
569:.
559:.
551:.
541:12
539:.
535:.
512:.
504:.
496:.
484:12
482:.
478:.
450:,
442:,
432:,
420:,
391:.
364:26
362:.
358:.
314:.
306:.
294:.
290:.
267:.
259:.
249:38
247:.
243:.
30:A
693:.
663::
657:6
636:.
604::
577:.
547::
520:.
500::
490::
428::
405:.
399::
376:.
370::
343:.
337::
322:.
302::
296:2
275:.
255::
20:)
Text is available under the Creative Commons Attribution-ShareAlike License. Additional terms may apply.