Abaqaphi
Imithetho, amasheke, nezilawuli ezikhawulela ukuziphatha kwemodeli okungaphephile noma okungafuneki.
Free AI glossary
I-AIGlossary
Plain-language definitions of the AI terms you keep hearing — for students, professionals, and anyone curious about how AI works.
213imigomoA–ZSesha + isihlungi sezinhlamvu100%Izincazelo zolimi olulula
A1 term
B1 term
Bala
Izinsiza zokucubungula ezidingekayo ukuze uqeqeshe futhi usebenzise amamodeli, ngokuvamile akalwa ngamahora e-FLOPS noma e-GPU.
D1 term
Dlulisa Ukufunda
Ukusebenzisa ulwazi olufundiwe kumsebenzi owodwa noma esizindeni ukuze kuthuthukiswe omunye umsebenzi.
E1 term
Eval Harness
Uhlaka lokuhlola oluphindaphindayo olusebenzisa ukwaziswa, amasethi edatha, kanye nengqondo yokunikeza amaphuzu kuzo zonke izinguqulo zemodeli.
H2 imigomo
Hlanganisa
Ukuhlanganisa izibikezelo ezivela kumamodeli amaningi ukuthuthukisa ukuqina noma ukunemba.
HumanEval
Ibhentshimakhi yezinkinga zohlelo lwePython esetshenziselwa ukukala ukulunga kokukhiqiza ikhodi ngokuhlolwa kweyunithi.
I116 imigomo
I-Agent Loop
Umjikelezo ophindaphindayo lapho umenzeli we-AI ebuka, ahlele, enze, futhi abonise aze aqedele umgomo noma afinyelele isimo sokuma.
I-AGI (Artificial General Intelligence)
Isistimu ye-AI ecatshangelwayo engakwazi ukwenza imisebenzi eminingi yobuhlakani ezingeni lomuntu ezizindeni eziningi.
I-AI ekhiqizayo
Amasistimu e-AI akhiqiza okuqukethwe okusha okunjengombhalo, izithombe, umsindo, ividiyo, noma ikhodi.
I-AI yomthethosisekelo
Indlela yokuqeqesha nokubumba indlela yokuziphatha lapho okuphumayo okuyimodeli kuqondiswa isethi engaguquki yezimiso ezibhaliwe.
I-algorithm
Isethi echaziwe yemithetho noma izinyathelo ezilandelwa ikhompuyutha ukuxazulula inkinga noma ukuqedela umsebenzi.
I-algorithmic Bias
Ukungalungi okuhleliwe emiphumeleni yamamodeli okubangelwa idatha esontekile, okucatshangwayo, noma ukukhetha kokumodela.
I-algorithmic Transparency
Yeka ukuthi kucace kangakanani ukuqonda kwesistimu ye-AI, imithombo yedatha, kanye nemikhawulo kubhalwe phansi futhi kuyaqondakala.
I-AliBi
Indlela yokuchema ejezisa izikolo zokunaka ngokusekelwe ebangeni lethokheni, isiza amamodeli ukuthi adlulele kunomongo omude.
I-ANN (Cishe Umakhelwane Oseduze)
Ama-algorithms athola ama-vector aseduze nombuzo ngaphandle kokuqhathanisa okuphelele, ukuhweba ngokunemba ngesivinini.
I-API (I-Application Programming Interface)
Indlela ehlelekile yesistimu yesofthiwe eyodwa ukuthumela izicelo futhi yamukele izimpendulo ezivela kwenye isistimu.
I-Artificial Intelligence (AI)
Inkambu ebanzi yezinhlelo zokwakha ezenza imisebenzi edinga ukunakwa kwephethini, ukucabanga, ulimi, noma ukwenza izinqumo.
I-Attention Mechanism
Ingxenye yemodeli egxila kakhulu ezingxenyeni ezifanele zokufakwayo lapho ikhiqiza okukhiphayo.
I-Backpropagation
I-algorithm yokuqeqesha eyinhloko ebuyekeza izisindo zemodeli ngokusabalalisa amaphutha okuqagela emuva ngenethiwekhi.
I-Beam Search
I-algorithm yokuqopha egcina ukulandelana kwekhandidethi eminingana ephezulu esinyathelweni ngasinye ukuze kutholwe imiphumela ephezulu yamathuba.
I-Bi-Encoder
Imodeli ehlanganisa imibuzo namadokhumenti abe amavekhtha ahlukene ukuze aqhathaniswe ngokushesha esikalini.
I-BPE (Byte Pair Encoding)
I-algorithm yethokheni yegama elingaphansi ehlanganisa ukupheya kwezinhlamvu ezivame kakhulu zibe amathokheni asebenziseka kabusha.
I-Causal Inference
Izindlela zokulinganisa ubudlelwano bembangela-nomphumela kunokuxhumana okulula.
I-Classifier
Imodeli eklanyelwe ngokuqondile imisebenzi yokuhlukanisa.
I-CLIP
Isakhiwo semodeli ye-multimodal efunda izethulo ezabiwe phakathi kombhalo nezithombe.
I-Computer Vision
Igatsha le-AI elikhipha incazelo ezithombeni nakuvidiyo.
I-Convolutional Neural Network (CNN)
I-neural architecture ethuthukiselwe ukucubungula idatha efana negridi njengezithombe.
I-Cross-Encoder
Imodeli efaka amaphuzu kumbuzo futhi idokhumenti ndawonye ngephasi eyodwa ukuze uthole izahlulelo zokunemba okuphezulu.
I-Data Drift
Ushintsho kudatha yokufaka yomhlaba wangempela ngokuhamba kwesikhathi okungase kwehlise izinga lokusebenza kwemodeli.
I-Data Provenance
Umsuka obhaliwe, ubunikazi, nomlando wedathasethi noma imodeli ye-artifact.
I-distillation
Ukucindezela ulwazi kusuka kumodeli kathisha enkulu ibe yimodeli yomfundi encane.
I-DPO (Direct Preference Optimization)
Indlela yokuqeqesha eshuna kahle imodeli ngokuqondile kumapheya athandwayo ngaphandle kokudinga imodeli yomklomelo ehlukile.
I-False Negative
Isibikezelo esingalungile lapho imodeli igeja icala leqiniso eliphozithivu.
I-Flash Attention
I-algorithm yokunaka ethuthukisiwe eyehlisa ukusetshenziswa kwenkumbulo futhi isheshise ukuqeqeshwa kwe-transformer kanye nencazelo.
I-Frequency Penalty
Isilungiselelo sokukhipha amakhodi esinciphisa amathuba amathokheni ngokulingana nokuthi avele kangaki kuze kube manje.
I-GAN (Generative Adversarial Network)
Isethaphu ekhiqizayo lapho ijeneretha nombandlululi baqeqeshana.
I-Gradient
Ivekhtha ebonisa ukuthi ipharamitha ngayinye kufanele ishintshe kangakanani ukuze kwehliswe ukulahlekelwa.
I-Groundedness
Izinga impendulo ye-AI esekelwa ngayo idatha yomthombo noma ubufakazi obubuyisiwe.
I-GSM8K
Ibhentshimakhi yezinkinga zamagama ezibalo zebanga lesikole ezisetshenziselwa ukuhlola ukucabanga kwesinyathelo nesinyathelo kumamodeli olimi.
I-hallucination
Uma imodeli ikhiqiza ulwazi olushelelayo kodwa olungamanga noma olungasekelwe.
I-HNSW
Isakhiwo senkomba esisekelwe kugrafu sosesho olusheshayo lomakhelwane oseduze ngaphezu kwamavekhtha anobukhulu obuphezulu.
I-Hybrid Search
Indlela yokubuyisa ehlanganisa ukusesha kwegama elingukhiye (i-lexical) ne-vector (semantic) yokusesha ukuze uthole ukukhumbula okungcono nokunemba.
I-Hyperparameter
Inani lokumisa elibekwe ngaphambi kokuqeqeshwa, njengezinga lokufunda, usayizi weqoqo, noma ukujula.
I-Inference Endpoint
Isixhumi esibonakalayo se-API esisetshenziswayo esithola izicelo zemodeli futhi sibuyisele izibikezelo ekukhiqizeni.
I-Inference-Time Compute
Inani lamandla okucubungula asetshenzisiwe ngenkathi kukhiqizwa impendulo ngayinye.
I-Instruction Tuning
Ukulungisa kahle imodeli kumapheya emiyalelo nezimpendulo ukuze uthuthukise umsebenzi olandelayo.
I-Jailbreak
Indlela esheshayo ehloselwe ukweqa imikhawulo yokuphepha yemodeli.
I-LoRA (Ukujwayela Kwezinga Eliphansi)
Indlela yokushuna esebenza kahle ngepharamitha eyengeza amatrices e-adaptha esezingeni eliphansi.
I-MCP (Model Context Protocol)
Iphrothokholi evulekile evumela izinhlelo zokusebenza ze-AI ukuthi zixhume kumathuluzi angaphandle, imithombo yedatha, nabahlinzeki bomongo ngendlela evamile.
I-MMLU
Ibhentshimakhi yokuhlola amamodeli olimi kuzo zonke izifundo ezingama-57 nezifundo zobuchwepheshe kusetshenziswa imibuzo yokukhetha okuningi.
I-Neural Network
Imodeli yekhompyutha enezingqimba egqugquzelwe ama-neuron ebhayoloji nama-synapse.
I-OCR (Ukubonwa Kwezinhlamvu Ezikhanyayo)
Ubuchwepheshe obuguqula umbhalo ezithombeni noma izikena zibe umbhalo ofundeka ngomshini.
I-Output Parser
I-logic eqinisekisa futhi iguqule okukhiphayo kwemodeli kube izakhiwo ezithayiphiwe ngokuqinile, ezisebenziseka umshini.
I-Parameter-Efficient Fine-Tuning (PEFT)
Izindlela ezijwayelanisa amamodeli ngokuqeqesha isethi encane yamapharamitha angeziwe.
I-RAG (Isizukulwane Sokubuyisa-Esithuthukisiwe)
Indlela ebuyisa ulwazi lwangaphandle futhi iluphakele ekukhiqizeni ngesikhathi sokunquma.
I-RAG-Fusion
Indlela yokubuyisa ekhiqiza ukuhlukahluka kwemibuzo eminingi, ibuyisa imiphumela ngayinye, futhi ihlanganise amazinga.
I-React
Iphethini yokwazisa ehlanganisa izinyathelo zokucabanga nezenzo zokusebenzisa ithuluzi ukuxazulula imisebenzi ngokwethembeka.
I-ROPE (Ukushumekwa kwe-Rotary Positional)
Indlela yokubhala ikhodi ezungezisa umbuzo namavekhtha angukhiye ukuze ibethelwe izindawo zamathokheni ahlobene.
I-Transformer
I-neural architecture esebenzisa ukunaka ebudlelwaneni bemodeli kuwo wonke amalandelano ngokuhambisana.
I-Watermarking
Ishumeka isignali ebonakalayo embhalweni okhiqizwe yi-AI noma imidiya ukuze kamuva ibonakale njengekhiqizwe ngomshini.
I-XAI (I-AI echazekayo)
Amasu nezinqubo zokwenza izibikezelo ze-AI zibe sobala futhi ziqondakale.
Ibhentshimakhi
Ukuhlolwa okujwayelekile noma isethi yedatha esetshenziselwa ukukala nokuqhathanisa ukusebenza kwemodeli.
Idatha Enkulu
Amasethi wedatha amakhulu kakhulu futhi ayinkimbinkimbi adinga ukugcinwa nokucutshungulwa kwamasu okukala.
Idatha Yokwenziwa
Idatha ekhiqizwe ngokuzenzisa esetshenziselwa ukwandisa, ukulingisa, noma ukuvikela idatha yokuqeqeshwa ebucayi.
Igrafu yolwazi
Isakhiwo segrafu samabhizinisi nobudlelwano obusetshenziselwa ukucabanga noma ukubuyisa.
Ikesi Lokuphepha
Impikiswano ehleliwe, esekelwa ubufakazi, bokuthi isistimu ye-AI iphephile kumongo ochaziwe wokusetshenziswa.
Iketango-lokucabanga
Isitayela sokucabanga lapho imodeli ye-AI idiliza inkinga ibe yizinyathelo eziphakathi nendawo.
Ikhadi Lemodeli
Amadokhumenti achaza ukusetshenziswa okuhlosiwe kwemodeli, amamethrikhi, imikhawulo, nobungozi.
Ikhompyutha Yesikhathi Sokuhlola
Izibalo ezengeziwe zokucatshangelwa ezisetshenziswa ngesikhathi sokwenziwa kwezimpendulo ukuze kuthuthukiswe ikhwalithi noma ukucabanga.
Ilebula ethi Smoothing
Indlela yokujwayela ethambisa amalebula aqinile ukuze kuthuthukiswe ukwenziwa okuvamile.
Imodeli Drift
Ukwehla kokusebenza ngokuhamba kwesikhathi njengoba izimo zomhlaba wangempela zihluka ekuqageleni kokuqeqeshwa.
Imodeli Encane
Imodeli lapho amapharamitha amaningi enguziro noma engasebenzi ukuze kuncishiswe ukubala.
Imodeli Quantization
Ukunciphisa ukunemba kwezinombolo kwezisindo zemodeli ukuze kwehliswe inkumbulo nezindleko zokusho.
Imodeli ye-Distilled
Imodeli encane eqeqeshelwe ukulingisa impatho yemodeli enkulu kuyilapho isebenzisa ikhompuyutha encane ekuqondeni.
Imodeli ye-Multimodal
Imodeli engacubungula noma ikhiqize izinhlobo eziningi zedatha njengombhalo, isithombe, nomsindo.
Imodeli Yebhokisi Elimnyama
Imodeli okulukhuni ukuchazwa ukucabanga kwayo kwangaphakathi ngabantu.
Imodeli Yesisekelo
Imodeli yereferensi elula esetshenziselwa ukuqhathanisa ukuthi izindlela eziyinkimbinkimbi kakhulu ziyayithuthukisa yini imiphumela.
Imodeli Yesisekelo
Imodeli enkulu eqeqeshwe kusengaphambili engakwazi ukujwayela imisebenzi eminingi engezansi.
Imodeli Yokusabalalisa
Isakhiwo esikhiqizayo esifunda ukuhlehlisa umsindo ukuze sihlanganise izithombe, umsindo, noma okunye okuqukethwe.
Imodeli yokushumeka
Imodeli ekhethekile yokuguqula idatha ibe ama-vector asetshenziselwa ukusesha kwe-semantic, ukuhlanganisa, nokubuyiswa.
Imodeli Yokuthandwa Kwabantu
Imodeli eqeqeshelwe amazinga abantu ukubikezela ukuthi yiziphi izimpendulo abasebenzisi abangazithanda.
Imodeli Yolimi Lombono (VLM)
Imodeli ye-multimodal ecubungula ngokuhlanganyela ulwazi olubonakalayo nolubhaliwe.
Imodeli Yolimi Olukhulu (LLM)
Imodeli yolimi eqeqeshwe ku-massive text corpora ukuze ikhiqize futhi ihlaziye umbhalo.
Imodeli Yolimi Oluncane (SLM)
Imodeli yolimi oluhlangene ethuthukiselwe ukubambezeleka okuphansi, izindleko, noma ukusetshenziswa kudivayisi.
Imodeli Yomthombo Ovulekile
Imodeli ekhishwe enesisindo somphakathi noma ikhodi ukuze ihlolwe, ilungiswe, futhi isetshenziswe kabusha.
Imodeli Yomvuzo
Imodeli ethola imiphumela ngokusekelwe kumasiginali athandwayo, ngokuvamile asetshenziswa kumapayipi e-RLHF.
Incazelo
Isigaba sesikhathi sokusebenza lapho imodeli eqeqeshiwe ikhiqiza ukuqagela noma okuphumayo.
Ingxube Yochwepheshe (MoE)
Isakhiwo esinamanethiwekhi angaphansi akhethekile lapho ochwepheshe abakhethiwe kuphela besebenza ngokokufaka ngakunye.
Inkokhelo Esekelwe Ekusetshenzisweni
Intengo lapho izindleko zikala khona ngamakholi we-API, amathokheni, isikhathi sokuqagela, noma ikhompuyutha esetshenzisiwe.
Inkumbulo (Inkumbulo yomenzeli)
Ingqikithi egciniwe umenzeli we-AI usebenzisa ezinyathelweni zonke noma izikhathi ukuze athuthukise ukuqhubeka.
Inqolobane ye-KV
Ukhiye ogciniwe kanye nezilinganisi zenani ezivela kumathokheni wangaphambilini avumela ama-transformer ukuthi akhiqize amathokheni amasha ngaphandle kokubala kabusha ukunaka kwesikhathi esidlule.
Intela Yokuqondanisa
Izindleko ezengeziwe ngesikhathi, ukubala, noma isivinini somkhiqizo esidingekayo ukwenza amasistimu aphephe futhi alawuleke kakhudlwana.
Ipayipi
Ukugeleza komsebenzi oku-oda kokucutshungulwa kwangaphambili, izinyathelo zemodeli, nezigaba zokucubungula ngemuva.
Ipharamitha
Isisindo esifundiwe ngaphakathi kwemodeli enomthelela ekuphumeni kwayo.
Iqembu Elibomvu
Ukuhlola ingcindezi isistimu ye-AI eneziyalezo eziphikisanayo ukuze kudalule ukwehluleka nobungozi.
Iqiniso
Ukunemba okungakanani izimangalo zemodeli zifana nolwazi lomhlaba wangempela olungaqinisekiswa.
Iqiniso Eliphansi
Amalebula ayizethenjwa athenjiwe asetshenziselwa ukuqeqesha noma ukuhlola okuphumayo kwemodeli.
Irejista yemodeli
Ikhathalogi emaphakathi yokuguqula, ukugunyaza, namamodeli okulandelela kuzo zonke izindawo.
Isenzo se-AI
Uhlaka lokulawula olusekelwe engcupheni lwe-European Union lwezinhlelo ze-AI nabahlinzeki.
Isethi yedatha
Iqoqo lezibonelo ezihlelekile noma ezingahlelekile ezisetshenziselwa ukuqeqeshwa, ukuqinisekiswa, noma ukuhlola.
Isethi Yokuhlola
Idathasethi ebanjiwe esetshenziselwa ukukala ikhwalithi yemodeli ngemva kokuqeqeshwa.
Isichasiselo
Amalebula engeziwe ngabantu noma imethadatha esetshenziselwa ukuqeqesha noma ukuhlola amamodeli okufunda omshini.
Isici
Okuhluka okokufaka okusetshenziswe imodeli ukwenza izibikezelo.
Isifaki khodi
Ingxenye yemodeli eguqula okokufaka kube izethulo ezifihlekile.
Isifanekiso Esisheshayo
Iphethini yokwaziswa esebenziseka kabusha enezinto eziguquguqukayo, imithetho yokufometha, nemiyalo eqondene nomsebenzi othize.
Isihlahla Semicabango
Indlela yokucabanga lapho imodeli ihlola izindlela eziningi zezixazululo zegatsha bese ikhetha ethembisa kakhulu.
Isihlahla Sesinqumo
Imodeli eyenza izibikezelo ngochungechunge lokuhlukaniswa kwesici uma-ke.
Isihlungi Sokuphepha
Isendlalelo sokulinganisela esivimba noma sibhale kabusha okokufaka noma okuphumayo kwemodeli okungaphephile.
Isijeziso Sobukhona
Ukulungiselelwa kokukhipha amakhodi okunciphisa amathuba amathokheni avele nhlobo, okukhuthaza izihloko ezintsha.
Isijeziso Sokuphindaphinda
Isilungiselelo sokukhipha amakhodi esehlisa amathuba amathokheni imodeli esiwakhiqizile kakade ukuze kuncishiswe izihibe.
Isikhala Esifihlekile
Isikhala sokumelela esicindezelwe lapho imiqondo efanayo ibekwe eduze kwenye njengama-vector.
Isikhawu Sokuzethemba
Ibanga lezibalo okungenzeka liqukethe inani langempela le-metric yemodeli elinganisiwe.
Isisekelo Solwazi
Iqoqo elikhethiwe lamadokhumenti noma amarekhodi asetshenziselwa ukubuyisa, ukusekela okuzenzakalelayo, noma izimpendulo eziyisisekelo.
Isisindo
Inani lezinombolo elifundiwe elikala amasignali adlula kunethiwekhi ye-neural.
Isistimu Yokuzimela
Uhlelo olukwazi ukwenza izinqumo futhi lusebenze ngokulinganiselwe noma okungenakho ukulawula komuntu okuqondile ngesikhathi sangempela.
Isitolo Sesici
Isistimu ephethwe yokugcina nokuphakela izici ze-ML eziqinisekisiwe ngokungaguquki ukuze kuqeqeshwe futhi kuqondiswe.
Isizinda Sesici
Ukuguqula idatha eluhlaza ibe izici ezifundisayo ezingase zisetshenziswe imodeli.
Isizindalwazi seVector
Isizindalwazi esilungiselelwe ukugcina kanye nemibuzo ngamavekhtha okushumeka anobukhulu obuphezulu.
Iwindi Lokuqukethwe
Inani eliphakeme lamathokheni okufakwayo imodeli yolimi engawacubungula ngesikhathi esisodwa.
Izinga lokufunda
I-hyperparameter yokuqeqesha elawula ukuthi mangakanani amapharamitha ashintsha isinyathelo ngasinye sokubuyekeza.
Izinga lokushisa
Isilungiselelo sesampula esilawula ukungahleliwe emiphumeleni ekhiqiziwe.
Izingcaphuno
Izinkomba zamavesi omthombo noma amadokhumenti afakwe empendulweni yemodeli ukusekela izimangalo zayo.
K2 imigomo
Khumbula
Ingxenye yamaphozithivu angempela imodeli ewakhomba kahle.
Kudivayisi i-AI
Ukuchazwa kwe-AI kwenziwa endaweni ku-hardware yomsebenzisi esikhundleni sensizakalo yamafu ekude.
L1 term
LLM-njengomahluleli
Ukusebenzisa imodeli yolimi ukuze uthole amaphuzu noma uqhathanise imiphumela evela kwamanye amamodeli ngesikhathi sokuhlola.
N2 imigomo
Ngemva kokuqeqeshwa
Izinyathelo zokuqeqesha zisetshenziswa ngemva kokuqeqeshwa kusengaphambili, njengokushuna iziyalezo, ukwenza kahle okuthandwayo, nokushuna kokuphepha.
Ngokushesha
Imiyalo yokokufaka nomongo onikezwe imodeli ekhiqizayo.
O2 imigomo
Okuhle Kwamanga
Isibikezelo esingalungile lapho imodeli ihlaba umkhosi ngokungalungile icala eliphikisayo njengeliphozithivu.
Okukhiphayo Okuhlelekile
Okukhiphayo okuyimodeli kuvinjelwa ku-schema esichaziwe njenge-JSON, izimpikiswano zamathuluzi, noma izinkambu ezithayiphiwe.
P3 imigomo
Pass@k
Imethrikhi yokuhlola ikhodi ekala ithuba lokuthi okungenani isampuli eyodwa kwamasampula akhiqizwe ngu-k aphumelele ukuhlolwa.
Perplexity
Imethrikhi yemodeli yolimi ekala ukuthi imodeli imangele kangakanani ngamathokheni alandelayo wangempela.
Prompt Engineering
Ukuklama ukwaziswa kokuthuthukisa ikhwalithi yokuphumayo, ukwethembeka, nokulawulwa.
Q2 imigomo
QLoRA
Indlela yokushuna kahle ehlanganisa ukulinganisa kwesisindo esingu-4-bit nama-adaptha e-LoRA ukuze kwehliswe izidingo zenkumbulo.
Quantization
Iguqula izisindo zemodeli zibe amafomethi anemba aphansi njenge-8-bit noma 4-bit.
R1 term
Reranker
Imodeli ehlela kabusha isethi yokuqala yemiphumela ebuyisiwe ukuze ibeke izinto ezifanele kakhulu phezulu.
S1 term
SentencePiece
Ithokheni ye-language-agnostic efunda amayunithi egama elingaphansi ngokuqondile embhalweni ongahluziwe ngaphandle kokuhlukanisa kusengaphambili endaweni emhlophe.
T3 imigomo
Themba Ukulinganisa
Ukuqondanisa ukuzethemba komsebenzisi emiphumeleni ye-AI nokwethembeka kwangempela kwesistimu kumsebenzi ngamunye.
Top-k Isampula
Isu lokukhipha amakhodi amasampula kuphela kusuka kumathokheni alandelayo okungenzeka ukuthi alandelayo.
Top-p (Nucleus) Sampling
Isu lokukhipha amakhodi amasampula asuka kusethi yethokheni encane kakhulu amathuba awo afinyelela ku-p.
U74 imigomo
Ubumfihlo Obuhlukile
Indlela yobumfihlo eyengeza umsindo wezibalo ukuze amarekhodi angawodwana angakwazi ukuchazwa ngokuthembekile kokuphumayo.
Ubunjiniyela besici
Ukuklama noma ukuguqula okokufaka okuguquguqukayo ukwenza ukufunda kube lula futhi kuphumelele kakhudlwana.
Uhlelo lokuncoma
Ipayipi eliyimodeli elibikezela izintandokazi zabasebenzisi zokulinganisa okuqukethwe noma imikhiqizo.
Uhlu Lwedatha
Irekhodi lokuthi idatha ivela kuphi, yaguqulwa kanjani, nokuthi isetshenziswa kuphi.
Ukubambezeleka
Isikhathi phakathi kokuthumela isicelo nokuthola okukhiphayo kwemodeli.
Ukubusa kwe-AI
Izinqubomgomo, amazinga, kanye nezindlela zokwengamela eziqondisa ukuthi i-AI ithuthukiswa futhi isetshenziswe kanjani emphakathini.
Ukubuyisa
Ukuthola imibhalo efanele noma amarekhodi emthonjeni wolwazi wombuzo.
Ukubuyisa Ukunemba
Ingxenye yezinto ezibuyisiwe ezihambisana nombuzo womsebenzisi.
Ukubuyiswa Kwedokhumenti Yomzali
Iphethini yokubuyisa esesha izingcezu ezincane kodwa ebuyisela amadokhumenti awo amakhudlwana ongumzali ukuze athole umongo ocebile.
Ukubuyiswa Kwemibuzo Eminingi
Isu lokubuyisa eliphinda libhale umbuzo womsebenzisi libe izinhlobo eziningana ukuze kuthuthukiswe ukukhumbula.
Ukuchazwa
Izinga lapho ukuziphatha kwemodeli kungachazwa futhi kuchazelwe abantu.
Ukuchema
Iphethini engaguquki yephutha noma engalungile kudatha noma imodeli yokuziphatha.
Ukucubungula Ulimi Lwemvelo (NLP)
Igatsha le-AI ligxile ekuqondeni nasekukhiqizeni ulimi lwabantu.
Ukufaka ngokweqile
Uma imodeli ibamba ngekhanda idatha yokuqeqeshwa futhi ingasebenzi kahle kokufakwayo okungabonwa.
Ukufunda kwe-Zero-Shot
Ukuxazulula imisebenzi ngaphandle kwezibonelo eziqondene nomsebenzi ngokuthembela olwazini olujwayelekile lwangaphambilini.
Ukufunda ngomshini (ML)
Izindlela ezivumela amasistimu ukuthi afunde amaphethini kudatha futhi athuthuke ngokuhamba kwesikhathi.
Ukufunda Okugadiwe
Ukuqeqesha imodeli enezibonelo ezinelebula ezifaka okokufaka kumephu emikhiqizweni eyaziwayo.
Ukufunda Okujulile
Isethi engaphansi yokufunda komshini esebenzisa amanethiwekhi emizwa anezendlalelo eziningi ukuze kufundwe ukumela.
Ukufunda Okumbalwa
Ukufunda noma ukulungisa indlela yokuziphatha kusuka kwizibonelo ezimbalwa kuphela.
Ukufunda Okungagadiwe
Amaphethini okufunda kusuka kudatha engenamalebula ngaphandle kokuphumayo okuqondiwe okucacile.
Ukufunda Okungaphakathi Kokuqukethwe
Ikhono lemodeli lokulandela amaphethini kusukela kuzibonelo ezinikezwe ngokuqondile ekwazisweni.
Ukufunda Okuqhubekayo
Izindlela zokuqeqesha ezivumela imodeli ukuthi iqhubeke nokufunda kudatha entsha ngaphandle kokukhohlwa ulwazi lwangaphambilini.
Ukufunda Okuqiniswayo Kwimpendulo Yomuntu (RLHF)
Indlela yokuqeqesha esebenzisa amasiginali athandwayo womuntu ukuze alolonge ukuziphatha okuyimodeli.
Ukufunda Okuzigadile
Ukufunda izethulo kudatha engenamalebula ngokubikezela izingxenye ezifihliwe noma eziguquliwe.
Ukugada Okubuthakathaka
Ukusebenzisa amalebula anomsindo, i-heuristic, noma ingxenye ukuqeqesha amamodeli lapho amalebula ahlanzekile eyivelakancane.
Ukugcinwa Kwedatha enguZero
Inqubomgomo lapho ukulayishwa kwesicelo/impendulo kungagcinwa khona ngemva kokucubungula ngaphezu kwamawindi okusebenza esikhashana.
Ukuhamba komsebenzi we-Agent
Inqubo enezinyathelo eziningi lapho isistimu ye-AI ihlela khona, yenze, ihlole imiphumela, futhi iphindaphinde iqonde emgomweni.
Ukuhlaziya Imizwa
Umsebenzi we-NLP ohlukanisa ithoni yemizwa noma umbono embhalweni.
Ukuhlukaniswa Kwenhloso
Ukubikezela inhloso yomsebenzisi embuzweni wombhalo ukuwuhambisa ngendlela efanele.
Ukujwayela
Imodeli isebenza kahle kangakanani kudatha entsha, engabonakali ngaphandle kwesethi yokuqeqeshwa.
Ukujwayela
Ukuguqula amanani abe esikalini esingashintshi ukuze kuthuthukiswe ukuzinza kokuthuthukisa.
Ukukhishwa Kwekhodi Okugadiwe
Isu lokukhiqiza elibophezela amathokheni okukhiphayo ezakhiweni ezivumelekile noma ukukhetha okuthobela inqubomgomo.
Ukulahleka Kwe-Cross-Entropy
Umsebenzi ojwayelekile wenhloso osetshenziswa ukuqeqesha amamodeli okuhlukanisa ngokujezisa amathuba angalungile.
Ukulahlekelwa Ukuqeqesha
Inani lephutha lemodeli lenziwe ikhompuyutha phakathi nokuqeqeshwa futhi lathuthukiswa lehla ngokuhamba kwesikhathi.
Ukulebula Idatha
Inqubo yokwabela omaka noma imiphumela eqondiwe kudatha eluhlaza yokufunda okugadiwe.
Ukulinganisa
Ukuthi amaphuzu okuzethemba kwemodeli afanelana kahle kangakanani namathuba okulutha.
Ukunaka Kwewindi Elishelelayo
Iphethini yokunaka lapho ithokheni ngayinye ibheka kuphela iwindi losayizi ongashintshi wamathokheni aseduze ukuze kuncishiswe ukubala.
Ukunakwa kwamakhanda amaningi
Umshini we-transformer osebenzisa ukunakwa okuningana ngokuhambisana ukuze uthwebule izinhlobo ezahlukene zobudlelwano.
Ukunemba
Ingxenye yezinto ezinhle ezibikezelwe ezilungile ngempela.
Ukunqunywa kolwazi
Iphuzu lakamuva ngesikhathi libonakala kudatha yokuqeqeshwa yemodeli.
Ukuphepha kwe-AI
Inkambu egxile ekwehliseni ukuziphatha okulimazayo, ukwehluleka, kanye nezingozi zokusebenzisa kabi ezinhlelweni ze-AI.
Ukuqashelwa Kwebhizinisi Eliqanjiwe (NER)
Umsebenzi we-NLP okhomba amabhizinisi afana nabantu, izindawo, izinsuku, noma izinhlangano.
Ukuqeqeshwa okuphakathi
Isigaba sokuqeqesha esimaphakathi phakathi kokuqeqeshwa kwangaphambili kanye nokuqeqeshwa kwangemuva, okuvame ukusetshenziselwa amandla noma ukulungiswa kwesizinda.
Ukuqina
Ikhono lemodeli lokugcina ukusebenza ngaphansi komsindo, amashifu, noma okokufaka okuphikisayo.
Ukuqinisa Ukufunda
Ukuqeqeshwa ngamasignali omklomelo lapho umenzeli efunda izenzo ezandisa imbuyiselo yesikhathi eside.
Ukuqinisekisa Isethi
Idathasethi esetshenziswa ngesikhathi sokuthuthukiswa ukuze kushune amamodeli futhi kuvinjwe ukucwiliswa ngokweqile.
Ukuqondanisa kwe-AI
Umsebenzi wokwenza amasistimu e-AI aziphathe ngokwezinjongo zomuntu, amanani, kanye nezithiyo zokuphepha.
Ukuqopha Okucatshangelwayo
Indlela yokusheshisa okucatshangwayo lapho imodeli encane esalungiswa iphakamisa amathokheni imodeli enkulu ewaqinisekisa ngokuhambisana.
Ukusesha kweSemantic
Sesha elifana nencazelo kunokuba igama elingukhiye ligqagqene, ngokuvamile lisebenzisa ukushumeka.
Ukusetshenziswa Kwethuluzi
Ikhono lemodeli lokubiza amathuluzi angaphandle afana nosesho, izibali, noma ama-API.
Ukushumeka
Isethulo sevekhtha yezinombolo esithwebula incazelo ye-semantic yombhalo, izithombe, noma enye idatha.
Ukushumeka Amagama
Ukumelwa kwevekhtha eminyene yamagama athwebula ubudlelwano be-semantic.
Ukushuna Kahle
Ukuqhubeka nokuqeqeshwa kudatha eqondene nesizinda ukuze kulungiswe imodeli eqeqeshwe ngaphambilini emsebenzini othile.
Ukuthena
Ukususa imodeli yesisindo engabalulekile noma ama-neurons ukunciphisa usayizi nokubala.
Ukuthunyelwa Kwethunzi
Ukusebenzisa imodeli ngokuhambisana nethrafikhi yokukhiqiza ngaphandle kokuthinta izinqumo ezibhekene nomsebenzisi.
Ukuvuza kweBenchmark
Uma izibonelo zokuhlola ibhentshimakhi noma okuhlukile kukhona kudatha yokuqeqeshwa, ukwenyusa ukusebenza okubikiwe.
Ukuzidlalela
Ukusethwa kokuqeqeshwa lapho imodeli ithuthuka ngokukhiqiza idatha ngokusebenzisana noma imiqhudelwano namakhophi ayo.
Ukuzijwayeza Kwesizinda
Izindlela ezidlulisa imodeli eqeqeshwe esizindeni esisodwa ukuze yenze kangcono kwesinye isizinda.
Ukuziqeqesha
Ukuqeqeshwa kokuqala kwemodeli yezinga elikhulu kudatha ebanzi ngaphambi kokujwayela umfula.
Ukwahlukanisa
Umsebenzi lapho imodeli yabela okokufaka esigabeni esichazwe ngaphambilini esisodwa noma eziningi.
Ukwaziswa Kwesistimu
Umyalelo obaluleke kakhulu osetha ukuziphatha, inqubomgomo, nesitayela sokuphendula semodeli.
Ukwehla kweGradient
Indlela yokuthuthukisa ebuyekeza amapharamitha esiqondisweni esinciphisa iphutha.
Ukwengezwa Kwedatha
Amasu adala izibonelo zokuqeqesha ezilungisiwe ukuze kuthuthukiswe ukwenziwa kwamamodeli.
Ukwenza amathokheni
Inqubo yokuhlukanisa umbhalo ube amathokheni okufaka imodeli.
Ulwazi Distillation
Ukuqeqesha imodeli encane ukulingisa okuphumayo kwemodeli enkulu.
Umbhalo Wekhodi Wendawo
Ulwazi lwengezwe ekushumekeni kwamathokheni ukuze ama-transformer akwazi ukuhlukanisa ukuhleleka kokulandelana.
Umenzeli we-AI
Isistimu yesofthiwe engabuka, icabange, futhi ithathe izinyathelo zokufeza umgomo, ngokuvamile isebenzisa amathuluzi nenkumbulo.
Umjovo osheshayo
Iphethini yokuhlasela lapho kufakwa khona iziqondiso ezinonya kokokufaka kwemodeli noma kokuqukethwe okubuyisiwe.
Umngcele Wesinqumo
Indawo esesikhaleni sesici esihlukanisa amakilasi abikezelwe umuntu ohlukanisa isigaba.
Umsebenzi Ukushaya
Imodeli yekhono lokukhiqiza amakholi ahleliwe acupha amathuluzi angaphandle noma ama-API.
Umsebenzi Wokulahlekelwa
Umgomo wezibalo olinganisela iphutha lokubikezela phakathi nokuqeqeshwa.
Umthetho Wokukala
Ubudlelwano obunobufakazi obubonisa ukuthi ukusebenza kuthuthukisa kanjani ngosayizi wemodeli, idatha, noma ikhompuyutha.
Umuntu-in-the-Loop
Ukugeleza komsebenzi lapho abantu bebuyekeza, beqondisa, noma bebhala ngaphezulu imiphumela ye-AI.
Uphawu
Isiqephu sombhalo esicutshungulwe amamodeli olimi, njengesiqephu segama noma uphawu.