Isifinyezo: Abacwaningi benze uhlaka lokufunda oluhlanganyelwe lokuqala oluqondiswa izinzwa lwama-BCI angavaleli. Ngokwenza iluphu ehlangene, enezinhlangothi ezimbili eqondanisa ukufunda nokuhlola kobuchopho bomuntu ngama-algorithms ezibalo omshini, ithimba lizuze ukunemba okungakaze kubonwe ngaphambili kubasebenzisi abangaqeqeshiwe ngokuphelele, laphula imikhawulo yokulinganiswa kwendabuko nokuvula indlela yobuchwepheshe obusizayo bansuku zonke.
Amaqiniso Abalulekile
- Ukwephula I-Learning Synchronization Deadlock: Isithiyo esiyinhloko kubunjiniyela be-BCI ukuthi abantu namakhompyutha bafunda ngezindlela ezihluke kakhulu. Ubuchopho bomuntu buzivumelanisa nezimo ngokuzama nangephutha, buhlanganisa ama-synapses abo ngokusekelwe kumpendulo yezinzwa. Ngaleso sikhathi, ikhodi ye-AI iyazibuyekeza isebenzisa amafomula ezibalo aqinile. Ama-BCI ajwayelekile ayaphuma ekuvumelaneni ngoba lezi zinhlelo ezimbili zidonsa ngezindlela ezihlukene.
- Isixazululo Sokufunda Ngokuhlanganyela Esiqondiswa Izinzwa: Uhlaka lukaDkt. Bin He luhlanganisa ngokusobala lezi zinhlelo ezimbili. Ngokwethula isiqondiso esihlelekile sokuthinta (izindlela zezinzwa) ukubumba amasu enhloso yomsebenzisi, abhangqwe nama-algorithms aguquguqukayo akhetha amasignali e-neural ahlanzekile, i-BCI iqinisekisa kokubili ukuzivumelanisa komuntu nomshini. ndawonye kuqhinga elilodwa, elihlanganyelwe lokulawula.
- Ukunemba Okukhethekile Okuzuziwe: Eqeqebeni labasebenzisi abangu-31 abangaqeqeshiwe ngokuphelele, uhlaka lokufunda oluhlanganyelwe luveze amamethrikhi okusebenza asheshayo, aphezulu angavamile ukubonwa ngaphandle kwamaviki okuphrakthiza:
- Ukunemba Okuhlukile: Kufinyelelwe 86% lawula ukunemba kwesiqondiso sekhesa sohlangothi olulodwa (1D) kanye 77.5% ngokulandelela okuyinkimbinkimbi kwegridi ye-two-dimensional (2D).
- Ukunemba Okuqhubekayo Kokulawula: Uketshezi olugciniwe, amanani okulandelela ngesikhathi sangempela we 77.5% (1D) futhi 66.9% (2D).
- Ukweqa i-Invasive Monopoly: Nakuba izimila zobuchopho ezihlaselayo sekuyisikhathi eside zibambe iqhaza ekusebenzeni okunembayo okuphezulu, lolu hlaka olusha olungahlaseli luletha izinzwa zezinga lesikhumba eduze kunangaphambili ukuze zifanise ukunemba kwebanga lokuhlinzwa ngaphandle kwezingozi zezokwelapha ezihambisanayo noma izindleko eziphezulu.
- Ukunqoba i-Calibration Bottleneck: Ama-BCI endabuko angahlaseli adinga amahora okulinganisa okukhathazayo, okungenzi lutho komsebenzisi ngaphambi kweseshini ngayinye. Indlela yokufunda ehlangene eqondiswa izinzwa iqeda lo mgoqo wokusebenza, isungula uhlelo oluguquguqukayo, olugxile kumsebenzisi olulungele ukuthunyelwa ngokushesha.
- Amandla aphezulu Wokuhumusha Umhlaba Wangempela: Ngokuyeka izimfuno zokuqeqeshwa kuyilapho ngesikhathi esifanayo kukhuphula ukusebenzelana kwemizwa yomsebenzisi, lolu hlaka lunikeza indlela eyingozi kakhulu yokufaka ama-BCI angavaleli ekusetshenzisweni kwansuku zonke komtholampilo, ikakhulukazi emitholampilo ye-neurorehabilitation, ukuxhumana okusizayo ezigulini ezikhiyelwe ngaphakathi, kanye nokulungiswa kwezitho zokufakelwa zerobhothi.
Umthombo: I-Carnegie Mellon University
Imishini efakelwayo ebuchosheni isetshenziswe cishe iminyaka engama-30 ukusiza abantu abakhubazekile ekuqedeni imisebenzi yezimoto. Kodwa-ke, izisetshenziswa azifinyeleleki kubantu abaningi abadinga usizo. Naphezu kwamashumi eminyaka besebenza kulo mkhakha, bangaphansi kwekhulu abantu emhlabeni wonke abaye bazuza kulobu buchwepheshe. Izindleko ziphakeme futhi ukuhlinzwa kobuchopho kuyingozi ngokwemvelo.
Kungakho abacwaningi bakaCarnegie Mellon, okuhlanganisa no-Bin He, uprofesa wobunjiniyela bezinto eziphilayo, ubunjiniyela bukagesi namakhompiyutha, kanye ne-Neuroscience Institute, kade besebenzela ukusebenzelana kobuchopho namakhompiyutha angavamile (BCIs) ukuthuthukisa ubuchwepheshe obungabizi kakhulu, obuphephile futhi obufinyeleleka kalula kubantu abaningi.
Kule minyaka eyi-10 kuye kweyi-15 edlule, basebenzise ama-BCI angavamisile ukuze bandize i-drone, balawule ingalo yerobhothi, bagcine ukulawulwa kwengalo yerobhothi, futhi muva nje, baqedele imisebenzi emihle yezimoto ezingeni leminwe. Nokho ukunemba kanye nezinga lokulawula kusetshenziswa ubuchwepheshe obungagayeki kuseyinselele.
Ngokokuqala ngqa, ithimba labacwaningi eliholwa nguYe lenze ubuchwepheshe obuyingxube obuhlanganisa ukufunda kwabantu nokufunda ngomshini kuma-BCI angavamisile.
Imiphumela yocwaningo, eshicilelwe ngo Ukuxhumana Kwemveloikhombisa intuthuko ebalulekile yesayensi nezobuchwepheshe kuma-BCI ngohlaka olunoveli olubhekana ngokuqondile nokungasebenzi kahle kokuqeqeshwa osekunesikhathi eside lukhona futhi lusungula indlela eyingozi ebheke ekuxhumaneni kwemizwa okuqinile nokujwayelekile.
Enye yezinselelo okubhekwana nazo ngohlaka lwamanoveli ukuthi abantu namakhompyutha bafunda ngezindlela ezahlukene. Abantu bafunda ngokuzama noma ngamaphutha njengoba ubuchopho bethu buzama izinto ezihlukile, bathola impendulo, futhi baphinde baxhumeke ngokusekelwe kulokho okusebenzayo. I-AI noma ikhodi yokufunda yomshini izibuyekeza yona kusetshenziswa amafomula ezibalo aqinile, abikezelwe ukuze kutholwe indlela esebenza kahle kakhulu. Ngenxa yokuthi ubuchopho bomuntu nohlelo lwekhompuyutha kuzivumelanisa nezimo ngezindlela ezinjalo ezihlukene, zingaphuma ekuvumelaneni noma zidonse izindlela ezihlukene, zenze isivimbamgwaqo se-BCI.
Kulolu cwaningo, abacwaningi bethule uhlaka lokufunda oluhlangene oluqondiswa izinzwa oluhlanganisa ngokusobala lezi zindlela ezimbili zokufunda. Ngokushumeka umhlahlandlela ophathekayo ohlelekile ukuze ulolonge amasu omsebenzisi kanye nokusebenzisa ama-algorithms aguquguqukayo anesisindo esikhethekile samaphethini e-neural afundisayo, indlela yabo yenoveli iqondanisa i-neuroplasticity yomuntu nokusetshenziswa kwedekhoda.
Ocwaningweni lwababambiqhaza abangu-31 abanamandla abangaqeqeshiwe kuma-BCI, babonisa ukuthi uhlaka lokufunda oluhlangene oluqondiswa izinzwa lukhiqiza izinzuzo ezisheshayo neziqhubekayo ekulawuleni izithombe zemoto kuyo yonke imisebenzi ekhulayo eyinkimbinkimbi. Ababambiqhaza bathole ukunemba okujwayelekile okumaphakathi okungama-86% wokulawula ikhesa enohlangothi olulodwa kanye no-77.5% wokulawula ikhesa enezinhlangothi ezimbili, kanye nokunemba okuqhubekayo okungu-77.5% (1D) no-66.9% (2D).
UBin He, umbhali ophezulu wocwaningo, uthe la mazinga okusebenza awavamile ukubonwa kubasebenzisi be-BCI abanokuqeqeshwa okulinganiselwe.
“Ngokuhlanganisa i-neuroscience nokufunda ngomshini sisondela eduze nokunemba kwezixhumanisi zekhompiyutha yobuchopho,” kusho uHe. “Ngokuqondanisa i-neural plasticity eqhutshwa ukuqinisa kanye nokwenza kahle kwe-decoder-based decoder, indlela yethu yeqa imikhawulo yezinqubo zokuqeqesha ezijwayelekile ze-BCI ezincike ekulinganisweni okungenzi lutho noma impendulo yendlela eyodwa.”
Umphumela awukona nje kuphela ukuthuthuka okuphawulekayo kokunemba, kodwa futhi ukusungulwa kwemodi entsha ngokuyisisekelo yokujwayela ngokubambisana komshini womuntu lapho bobabili ozakwethu behlangana bebheke ekwabelaneni, amasu okulawula asekelwe ngokomzimba.
Ngaphandle kokuqinisekiswa kwelabhorethri, Uthe le noveli indlela edidiyelwe inamandla acacile okuhumusha. Amandla okuthola ukulawulwa kwe-BCI okusheshayo nokuthembekile kubasebenzisi abangaqeqeshiwe kubhekana nesithiyo esimaphakathi ekufakweni komtholampilo, ikakhulukazi ekuvuseleleni izinzwa, ukuxhumana okusizayo, nokulawulwa kwezitho zokufakelwa.
“Ngokunciphisa izimfuno zokuqeqeshwa ngenkathi kuthuthukiswa ukubandakanya kwe-neural, uhlaka lokufunda oluhlangene oluqondiswa izinzwa luletha ama-BCI angavamisile eduze nokusetshenziswa okuyingozi, kwansuku zonke,” esho. “Ngokwenza kanjalo, kuphawula ukuguquguquka kwepharadigm kusuka ezinhlelweni eziqinile zokulinganisa kuya ezindaweni eziguquguqukayo, ezigxile kumsebenzisi ze-neural ezisebenza emhlabeni wangempela.
“Uma umsebenzi omningi esiwenza kule ndawo, maningi amathuba okuthi ngelinye ilanga sifinyelele i-BCI engahlaseli enembile njengethuluzi elifakwe ebuchosheni,” kusho yena. “Leli yithemba lami, iphupho lami.”
Uxhaso: Lo msebenzi usekelwe ngokwengxenye yi-National Institute of Neurological Disorders and Stroke kanye ne-BRAIN Initiative ye-National Institutes of Health, kanye ne-National Institute of Biomedical Imaging and Bioengineering training grant.
Abanye abahlanganyeli ku Ukuxhumana Kwemvelo iphepha lihlanganisa umbhali wokuqala u-Hanwen Wang, i-biomedical engineering postdoctoral associate, u-Yisha Zhang, owayenguchwepheshe zelebhu yobunjiniyela be-biomedical, uMaxim Karrenbach, i-Ph.D yobunjiniyela bukagesi kanye nekhompyutha. umfundi, kanye no-Yidan Ding, i-Ph.D yobunjiniyela bezinto eziphilayo. umfundi.
Imibuzo Ebalulekile Iyaphendulwa:
A: Noma amadivaysi obuchopho ahlaselayo angaletha ukunemba okumangalisayo, awafinyeleleki ngokuphelele kubantu abaningi abadinga usizo. Ukufaka ama-microelectrode ngokuhlinzwa ngqo esicutshini sobuchopho kudinga ukuhlinzwa kwezinzwa okuyinkimbinkimbi, okuyingozi enkulu okuthwala izingozi ezingokwemvelo ezifana nezibazi zezicubu, ukopha, noma izifo zobuchopho eziyingozi. Ngaphezu kwalokho, ingqalasizinda yezokwelapha ekhethekile, i-hardware eyenziwe ngokwezifiso, kanye nokwesekwa kokuhlinzwa kudala izindleko ezinqamulayo umshwalense ojwayelekile kanye nemindeni engakwazi ukuzikhokhela.
A: Ubuchopho bomuntu nezinhlelo ze-AI zikhuluma izilimi ezithuthukayo ezihluke kakhulu. Umuntu ufunda umsebenzi ofana nokuqondisa ikhesa ngomzwelo wemvelo kanye nokuzama kanye nephutha—ukuzama iphethini yokucabanga, ukuzwa umphumela, nokuhlanganisa kabusha ama-synapse ngokususelwe kulokho azizwa kulungile. Ikhompyutha ye-AI ibuyekeza ikhodi yayo isebenzisa amafomula aqinile, abandayo ezibalo. Ngenxa yokuthi izinhlangothi zombili zishintsha imithetho ngesikhathi esisodwa ngaphandle kokuxhumanisa, zigcina zilwa zodwa, okudale ukuthi iBCI ibambezeleke. Uhlaka lukaDkt. Bin He lusebenza njengomfundisi womdanso onguchwepheshe, lisebenzisa isiqondiso esithintekayo ukuze sisize umuntu akhethe amasu acacile kuyilapho esebenzisa ama-algorithms aguquguqukayo ukuze aqinisekise ukuthi ikhompuyutha ikala lawo maza obuchopho athile, okuvumela womabili ukuthi ahlangane aye esu lokulawula okwabiwe.
A: Ngokomlando, iphupho eliwujuqu lama-BCI angavaleli, ukufaka ikepisi elilula ukuze ulawule ingalo yokufakelwa yobuchwepheshe obuphezulu noma isihlalo sabakhubazekile, lahlehliswa ngenxa yokuthi amasistimu ayenensa kakhulu, engalungile, futhi ekhungathekisa ukusetshenziswa ngaphandle kwezinyanga zokuphrakthiza elebhu. Ngokunika amandla abantu abangaqeqeshiwe ngokuphelele ukuthi bathole ngokushesha u-86% wokunemba kokuqondisa okuhlukile emzamweni wabo wokuqala, lolu hlaka luwubhidliza ngokuphelele lowo mgoqo wokuqeqesha. Kuphawula uguquko olukhulu ekufinyeleleni, kubuchwepheshe bokusiza be-plug-and-play, kusisondeza ekusaseni lapho iziguli ezikhubazekile zingakwazi ukulawula izinto zangaphandle ezindlini zazo ngaphandle kokudinga ukuhlinzwa kobuchopho.
Amanothi Omhleli:
- Lesi sihloko sihlelwe umhleli weNeuroscience News.
- Iphepha lejenali libuyekezwe ngokugcwele.
- Ingqikithi eyengeziwe yengezwe abasebenzi bethu.
Mayelana nalezi zindaba zocwaningo lwe-neurotech
Umbhali: Erin Gazica
Umthombo: UCarnegie Mellon
Othintana naye: U-Erin Gazica – uCarnegie Mellon
Isithombe: Isithombe sifakwe ku-Neuroscience News
Ucwaningo lwangempela: Vula ukufinyelela.
“Ukufunda okuhlangene komshini womuntu okuqondiswa izinzwa kusheshisa ukutholwa kokulawula isixhumi esibonakalayo sekhompuyutha yobuchopho” ngo-Hanwen Wang, u-Yisha Zhang, uMaxim Karrenbach, u-Yidan Ding no-Bin He. Ukuxhumana Kwemvelo
I-DOI:10.1038/s41467-026-75435-5
Abstract
Ukufunda ngokuhlanganyela komshini womuntu okuqondiswa izinzwa kusheshisa ukutholwa kokulawula isixhumi esibonakalayo sekhompuyutha yesithombe sobuchopho
I-Brain-computer interfaces (BCIs) inikeza amandla okubuyisela ukusebenza nokukhulisa amakhono abantu. Kodwa-ke, ama-BCI asekelwe ku-electroencephalography (EEG)-based non-invasive electroencephalography (EEG) asabhekene nezinselele ekufundeni ukusebenza kahle nokulawula ukunemba, ikakhulukazi kubasebenzisi abangenalwazi abenza imisebenzi eyinkimbinkimbi.
Lapha, sethula uhlaka lokufunda oluhlanganyelwe oluqondiswa izinzwa oluhlanganisa ukufunda kwemoto yomuntu nokufunda komshini okuguquguqukayo ukuze kuthuthukiswe ukuqeqeshwa nokusebenza kwe-BCI.
Kubahlanganyeli abangu-31 be-BCI-naïve, uhlaka lunikeze amandla ukutholwa kwamakhono okusheshayo, ukufeza ukunemba okulinganiselwe okumaphakathi kwe-inthanethi okungu-86.0% kohlangothi olulodwa (1D) kanye no-77.5% wemisebenzi yesithombe esinezinhlangothi ezimbili (2D), kanye nokunemba okuqhubekayo kokulawula kwe-5% (617) no-5% (617) Ngokwemishini, isiqondiso esithintekayo sinciphise ukuhlola komsebenzisi futhi kwasheshisa ukuzivumelanisa ne-neural, kuyilapho isampula yokuvuselela i-decoder eqondaniswe nezibuyekezo ezinemigudu yokufunda yomuntu.
Ngokuhlanganisa ipulasitiki ye-neural eqhutshwa ukuqinisa ne-adaptive algorithmic optimization, lolu hlaka luthuthukisa ukuqeqeshwa kwe-BCI kusukela ekulinganiseni okungenzi lutho kuya ekufundeni okuhlangene komshini womuntu nomshini, okuvumela ukuxhumana okusebenzayo nokunokwehla kwe-neural ukuxhumana nokuvuselela.
