435:, at a precise location in the visual field (the scene)". It has been experimentally determined, for example, after mapping the angular position of some objects in the visual field, there will be a one-to-one correspondence of cells in the scene to the angular positions of those objects. Hawkins predicts that when the features of a visual scene are known in a memory, anticipatory cells should fire
25:
333:) of patterns and developing invariant representations. Higher levels of the cortical hierarchy predict the future on a longer time scale, or over a wider range of sensory input. Lower levels interpret or control limited domains of experience, or sensory or effector systems. Connections from the higher level states predispose some selected transitions in the lower-level state machines.
480:
will be active during a learned sequence. Hawkins posits that these cells will remain active for the duration of the learned sequence, even if the remainder of the cortical column is shifting state. Since we do not know the encoding of the sequence, we do not yet know the definition of
349:
is a basic element in the framework. Hawkins places particular emphasis on the role of the interconnections from peer columns, and the activation of columns as a whole. He strongly implies that a column is the cortex's physical representation of a state in a state machine.
303:
Hawkins' basic idea is that the brain is a mechanism to predict the future, specifically, hierarchical regions of the brain predict their future input sequences. Perhaps not always far in the future, but far enough to be of real use to an organism. As such, the brain is a
361:. For example, for the purposes of his framework, the nerve impulses can be taken to form a temporal sequence (but phase encoding could be a possible implementation of such a sequence; these details are immaterial for the framework).
514:
6. Hawkins' novel prediction is that certain cells are inhibited during a learned sequence. A class of cells in layers 2 and 3 should NOT fire during a learned sequence, the axons of these "exception cells" should fire
129:
447:
3. In layers 2 and 3, predictive activity (neural firing) should stop propagating at specific cells, corresponding to a specific prediction. Hawkins does not rule out anticipatory cells in layers 4 and 5.
924:
357:, but merely signals that the natural process has performed Hawkins' functional decomposition in a different, unexpected way, as Hawkins' motivation is to create intelligent
353:
As an engineer, any specific failure to find a natural occurrence of some process in his framework does not signal a fault in the memory-prediction framework
464:
are in layer 2, physically adjacent to layer 1. Hawkins does not rule out the existence of layer 3 cells with dendrites in layer 1, which might perform as
675:
929:
815:
527:
7. If an unusual event occurs (the learned sequence fails), the "exception cells" should fire, propagating up the cortical hierarchy to the
759:
George, Dileep; Hawkins, Jeff (2005). "A Hierarchical
Bayesian Model of Invariant Pattern Recognition in the Visual Cortex": 1812–1817.
884:
658:
674:
Fogassi, Leonardo; Ferrari, Pier
Francesco; Gesierich, Benno; Rozzi, Stefano; Chersi, Fabian; Rizzolatti, Giacomo (April 29, 2005).
222:
108:
871:
42:
919:
89:
339:
is part of the framework, in which the event of learning physically alters neurons and connections, as learning takes place.
46:
61:
608:
244:
539:
8. Hawkins predicts a cascade of predictions, when recognition occurs, propagating down the cortical column (with each
68:
797:- Research papers and programs presenting experimental results with Bayesian models of the Memory-Prediction Framework
614:
298:
282:
308:
794:
75:
899:
377:
as a prototype for some example predictions, such as
Predictions 2, 8, 10, and 11. Other predictions cite the
122:
On
Intelligence: How a New Understanding of the Brain will Lead to the Creation of Truly Intelligent Machines
270:
On
Intelligence: How a New Understanding of the Brain will Lead to the Creation of Truly Intelligent Machines
57:
35:
519:. This prevents flooding the brain with the usual sensations, leaving only exceptions for post-processing.
760:
323:
305:
774:
722:
584:
690:
765:
839:
714:
342:
460:. Hawkins calls the cells which fire in this sequence "name cells". Hawkins suggests that these
706:
654:
650:
643:
611:, a technology by Hawkins's startup Numenta Inc. to replicate the properties of the neocortex.
571:. Hawkins posits a temporal window (presuming time-encoded firing) which is necessary for his
330:
229:
217:
82:
892:
856:
831:
698:
336:
278:
255:
146:
787:
498:
428:
395:
378:
346:
694:
588:
556:
452:"Name cells" at layers 2 and 3 should preferentially connect to layer 6 cells of cortex
875:
913:
432:
411:
374:
319:
903:
718:
599:
11. Hawkins predicts that "name cells" will be found in all regions of the cortex.
274:
142:
551:
Pyramidal cells should detect coincidences of synaptic activity on thin dendrites
528:
493:(i.e., the name cells simultaneously "light up") across an array of name cells.
490:
174:
24:
431:, Hawkins predicts, for example, "we should find anticipatory cells in or near
326:
state machine, the machine responds to future events predicted from past data.
403:
370:
164:
800:
702:
544:
202:
710:
443:
Prediction should stop propagating in the cortical column at layers 2 and 3
489:; Hawkins suggests that the ON pattern may be as simple as a simultaneous
560:
312:
236:
860:
835:
579:
Learned representations move down the cortical hierarchy, with training
568:
540:
358:
329:
The hierarchy is capable of memorizing frequently observed sequences (
572:
564:
676:"Parietal lobe: from action organization to intention understanding"
286:
803:- An open source project for modeling Memory-Prediction Framework
827:
751:
128:
847:
Colwell, B. (2005). "Machine
Intelligence Meets Neuroscience".
322:
actually controls the behavior of the organism. Since it is a
385:
An
Appendix of 11 Testable Predictions, beginning on page 237:
18:
510:"Exception cells" should remain OFF during a learned sequence
456:
4. Learned sequences of firings comprise a representation of
230:
587:(IT) level has learned a sequence, that eventually cells in
365:
Predictions of the theory of the memory-prediction framework
390:
Enhanced neural activity in anticipation of a sensory event
256:
559:
should be capable of detecting coincident events on thin
523:"Exception cells" should propagate unanticipated events
472:"Name cells" should remain ON during a learned sequence
254:
242:
228:
216:
208:
198:
190:
180:
170:
160:
152:
138:
49:. Unsourced material may be challenged and removed.
642:
402:", cells that fire in anticipation of a sensory
925:Non-fiction books about Artificial intelligence
535:"Aha! cells" should trigger predictive activity
583:10. Hawkins posits, for example, that if the
8:
121:
595:"Name cells" exist in all regions of cortex
439:the actual objects are seen in the scene.
398:, Hawkins (2004) predicts "we should find
311:with special properties that enable it to
127:
120:
816:"Machine Intelligence Meets Neuroscience"
764:
636:
634:
632:
630:
109:Learn how and when to remove this message
289:and describes some of its consequences.
885:"On Intelligence, People and Computers"
626:
900:On Biological and Digital Intelligence
783:
772:
649:(1st ed.). Times Books. pp.
517:only if a local prediction is failing
7:
547:over a learned scene, for example).
505:which perform this type of function.
47:adding citations to reliable sources
795:Saulius Garalevicius' research page
883:Kling, Arnold (22 November 2004).
531:, the repository of new memories.
14:
870:Dill, Franz (October 30, 2004).
23:
872:"Jeff Hawkins: On Intelligence"
381:( Predictions 1, 3, 4, and 7).
34:needs additional citations for
930:Books about human intelligence
591:will also learn the sequence.
458:temporally constant invariants
1:
814:Colwell, Bob (January 2005).
478:temporally constant invariant
423:Spatially specific prediction
281:. The book explains Hawkins'
609:Hierarchical temporal memory
615:Memory-prediction framework
414:have been observed to fire
299:Memory-prediction framework
283:memory-prediction framework
946:
309:hierarchical state machine
296:
126:
499:Neural ensemble#Encoding
703:10.1126/science.1106138
920:2004 non-fiction books
782:Cite journal requires
641:Hawkins, Jeff (2004).
427:2. In primary sensory
418:an anticipated event.
889:Tech Central Station
476:5. By definition, a
345:'s formulation of a
43:improve this article
16:Book by Jeff Hawkins
695:2005Sci...308..662F
503:grandmother neurons
394:1. In all areas of
123:
861:10.1109/MC.2005.24
836:10.1109/MC.2005.24
575:to remain viable.
567:with thousands of
400:anticipatory cells
343:Vernon Mountcastle
273:is a 2004 book by
801:Project Neocortex
689:(5722): 662–667.
410:Note: As of 2005
331:Cognitive modules
266:
265:
191:Publication place
119:
118:
111:
93:
58:"On Intelligence"
937:
896:
891:. Archived from
879:
874:. Archived from
864:
843:
838:. Archived from
791:
785:
780:
778:
770:
768:
755:
754:
752:Official website
737:
736:
734:
733:
727:
721:. Archived from
680:
671:
665:
664:
648:
638:
337:Hebbian learning
279:Sandra Blakeslee
262:QP376 .H294 2004
258:
232:
182:Publication date
147:Sandra Blakeslee
131:
124:
114:
107:
103:
100:
94:
92:
51:
27:
19:
945:
944:
940:
939:
938:
936:
935:
934:
910:
909:
882:
869:
846:
813:
810:
781:
771:
766:10.1.1.132.6744
758:
750:
749:
746:
741:
740:
731:
729:
725:
678:
673:
672:
668:
661:
645:On Intelligence
640:
639:
628:
623:
605:
597:
581:
557:Pyramidal cells
553:
537:
525:
512:
474:
454:
445:
425:
392:
379:auditory system
367:
347:cortical column
301:
295:
247:
199:Media type
183:
134:
115:
104:
98:
95:
52:
50:
40:
28:
17:
12:
11:
5:
943:
941:
933:
932:
927:
922:
912:
911:
908:
907:
897:
895:on 2012-03-05.
880:
878:on 2012-02-05.
867:
866:
865:
842:on 2005-02-04.
809:
806:
805:
804:
798:
792:
784:|journal=
756:
745:
744:External links
742:
739:
738:
666:
660:978-0805074567
659:
625:
624:
622:
619:
618:
617:
612:
604:
601:
596:
593:
585:inferotemporal
580:
577:
552:
549:
536:
533:
524:
521:
511:
508:
507:
506:
473:
470:
453:
450:
444:
441:
424:
421:
420:
419:
412:mirror neurons
391:
388:
387:
386:
366:
363:
297:Main article:
294:
291:
285:theory of the
264:
263:
260:
252:
251:
248:
243:
240:
239:
234:
226:
225:
220:
214:
213:
210:
206:
205:
200:
196:
195:
192:
188:
187:
184:
181:
178:
177:
172:
168:
167:
162:
158:
157:
154:
150:
149:
140:
136:
135:
132:
117:
116:
31:
29:
22:
15:
13:
10:
9:
6:
4:
3:
2:
942:
931:
928:
926:
923:
921:
918:
917:
915:
905:
901:
898:
894:
890:
886:
881:
877:
873:
868:
862:
858:
854:
850:
845:
844:
841:
837:
833:
829:
825:
821:
817:
812:
811:
807:
802:
799:
796:
793:
789:
776:
767:
762:
757:
753:
748:
747:
743:
728:on 2017-08-09
724:
720:
716:
712:
708:
704:
700:
696:
692:
688:
684:
677:
670:
667:
662:
656:
652:
647:
646:
637:
635:
633:
631:
627:
620:
616:
613:
610:
607:
606:
602:
600:
594:
592:
590:
586:
578:
576:
574:
570:
566:
563:, even for a
562:
558:
550:
548:
546:
542:
534:
532:
530:
522:
520:
518:
509:
504:
500:
496:
495:
494:
492:
488:
484:
479:
471:
469:
467:
463:
459:
451:
449:
442:
440:
438:
434:
430:
422:
417:
413:
409:
408:
407:
405:
401:
397:
389:
384:
383:
382:
380:
376:
375:visual system
372:
364:
362:
360:
356:
351:
348:
344:
340:
338:
334:
332:
327:
325:
321:
320:state machine
316:
314:
310:
307:
300:
292:
290:
288:
284:
280:
276:
272:
271:
261:
259:
257:LC Class
253:
249:
246:
245:Dewey Decimal
241:
238:
235:
233:
227:
224:
223:0-8050-7456-2
221:
219:
215:
211:
207:
204:
201:
197:
194:United States
193:
189:
185:
179:
176:
173:
169:
166:
163:
159:
155:
151:
148:
144:
141:
137:
130:
125:
113:
110:
102:
91:
88:
84:
81:
77:
74:
70:
67:
63:
60: –
59:
55:
54:Find sources:
48:
44:
38:
37:
32:This article
30:
26:
21:
20:
906:(7 Oct 2004)
904:Ben Goertzel
902:A review by
893:the original
888:
876:the original
852:
848:
840:the original
823:
819:
775:cite journal
730:. Retrieved
723:the original
686:
682:
669:
644:
598:
582:
554:
538:
526:
516:
513:
502:
486:
482:
477:
475:
465:
461:
457:
455:
446:
436:
426:
415:
399:
393:
368:
354:
352:
341:
335:
328:
324:feed forward
317:
306:feed forward
302:
275:Jeff Hawkins
269:
268:
267:
143:Jeff Hawkins
105:
96:
86:
79:
72:
65:
53:
41:Please help
36:verification
33:
529:hippocampus
371:predictions
175:Times Books
133:Front cover
914:Categories
732:2006-11-18
621:References
466:name cells
462:name cells
293:The theory
250:612.8/2 22
165:Psychology
99:April 2013
69:newspapers
855:: 12–15.
830:: 12–15.
761:CiteSeerX
561:dendrites
203:Paperback
171:Publisher
849:Computer
820:Computer
711:15860620
603:See also
569:synapses
373:use the
359:machines
237:55510125
153:Language
808:Reviews
719:5720234
691:Bibcode
683:Science
543:of the
541:saccade
161:Subject
156:English
83:scholar
763:
717:
709:
657:
573:theory
565:neuron
487:active
437:before
429:cortex
416:before
396:cortex
355:per se
145:&
139:Author
85:
78:
71:
64:
56:
826:(1).
726:(PDF)
715:S2CID
679:(PDF)
404:event
313:learn
287:brain
209:Pages
90:JSTOR
76:books
828:IEEE
788:help
707:PMID
655:ISBN
501:for
497:See
369:His
318:The
277:and
231:OCLC
218:ISBN
186:2004
62:news
857:doi
832:doi
699:doi
687:308
651:272
555:9.
545:eye
491:AND
485:or
212:272
45:by
916::
887:.
853:38
851:.
824:38
822:.
818:.
779::
777:}}
773:{{
713:.
705:.
697:.
685:.
681:.
653:.
629:^
589:V4
483:ON
468:.
433:V1
406:.
315:.
863:.
859::
834::
790:)
786:(
769:.
735:.
701::
693::
663:.
112:)
106:(
101:)
97:(
87:·
80:·
73:·
66:·
39:.
Text is available under the Creative Commons Attribution-ShareAlike License. Additional terms may apply.
