InoteveraGaidhi rinotevera
Tarisa mberi uye Shumba Optimizers
Tekinoroji
Nhungamiro yehunyanzvi
InfoNCE uye SimCLR Zvinangwa
InfoNCE ndiko kurasikirwa kwakasiyana kunodzidzisa modhi kudhonza mapeya anoenderana pamwe chete uye kusundira zvisingaenderane peya nepakati mukumisikidza nzvimbo.
Teerera kugwaro iriAI-yakarondedzerwa · ~2 min
Pa peji ino2 min verenga
Pfupiso
SimCLR inzvimbo yakakosha iyo yakashandisa kurasikirwa uku kudzidza ine simba inomiririra mifananidzo kubva kune isina kunyorwa data, kukwikwidza inotariswa pretraining.
Kudzika Kwakadzika
InfoNCE (Noise-Contrastive Estimation for mutual information) inodzidzisa encoder kuitira kuti mubvunzo nechaiko chawo chechokwadi ive nemhando yepamusoro yezvibodzwa pane mubvunzo nezvakawanda zvakaipa. Iyo inonyanya kuve softmax muchinjiko-entropy pamusoro pekufanana zvibodzwa: kune anchor, iyo yakanaka inofanirwa kuhwina inopesana nezvakashata. SimCLR (2020) yakashandisira izvi kumifananidzo: tora mufananidzo mumwe, shandisa maviri akasarudzika ekuwedzera kuti ugadzire maviri akanaka, mhanyisa zvese kuburikidza neyakagovaniswa encoder pamwe nemusoro wekufungidzira, uye shandisa yakajairwa tembiricha-yakaiswa muchinjiro-entropy (NT-Xent, iyo InfoNCE musiyano) kuti iwo maviri akawedzerwa maonero anokwezva nepo mimwe mifananidzo yese mubatch ichiita seisina kunaka. SimCLR yakaratidza kuti kuwedzera kwedata kwakasimba, musoro wekufungidzira usiri wemutsetse, saizi hombe dzebhechi, uye tembiricha yakarongedzwa pamwechete rega mamodheru anozvitarisira aenderane neanotariswa paImageNet - pasina chero mavara panguva yekudzidzira.
Kuitwa muNyika yeChokwadi
SimCLR kufanodzidzira encoder yemufananidzo pamifananidzo isina kunyorwa, wozogadzirisa padiki diki rakanyorwa seti yekuiswa.
CLIP uchishandisa InfoNCE chinangwa kuenzanisa mifananidzo ine zvinyorwa zvayo, zvichiita kuti zero-shot image classification.
Kuvaka kutsvaga kwekuona/kutsvagisa uko mifananidzo yakafanana inogara padhuze munzvimbo yakadzidzwa yekumisikidza.
Kuzvitarisira wega pretraining yezvekurapa kana sesatellite mifananidzo uko kune mavara mashoma asi data raw rakawanda.
Ramangwana reInfoNCE uye SimCLR Zvinangwa
Zvinangwa zvinopesana zvakapararira kupfuura SimCLR: CLIP inoenderana nemifananidzo nemavara uchishandisa InfoNCE munzira dzese, uye kurasikirwa kwakafanana kunotyaira odhiyo, vhidhiyo, uye yekudzosa modhi. Tsvagiridzo ikozvino inoderedza kutsamira pamabhechi akakura uye akawanda asina kunaka kuburikidza nemabhangi ekurangarira (MoCo), kana kubvisa pachena zvisina kunaka zvachose (BYOL, SimSiam, DINO). Tarisira kuenderera mberi kusanganiswa kwekusiyanisa, distillation, uye masked-modelling pretraining, ine multimodal alignment (zvinyorwa, mufananidzo, odhiyo) semuganho mukuru weiyo nheyo modhi.
Kuzvishandisa
Nei zvichikosha
- Mutengo uye bhajeti. Zvisarudzo zvezvivakwa zvinotyaira kuita uye mutengo wekushandisa kwemakore.
- Sarudzo dzakajeka. Dzidzo yehunyanzvi inobatsira zvikwata kusarudza murwi wakakodzera, kwete iwo mutsva chete.
- Kudzora kwemhando yepamusoro. Sarudzo dzeinjiniya dziri nani dzinoderedza zviitiko zvekuvimbika mukugadzira.
Chenjera
- Kugadzirisa imwe bhenji kunogona kuvanza yakafara system kushaya simba.
- Infrastructure uye mari yekugadzirisa inowanzotarisirwa pasi.
- Chengetedzo uye kucherechedzwa mapundu anogona kukura sezvo masisitimu anowedzera kuoma.
Matanho anotevera
- Tsanangura latency, mhando, uye mutengo zvinangwa usati waitwa.
- Benchmark pasi pechokwadi mutoro uye data mamiriro.
- Chishandiso chekutarisa zvikanganiso, kudonha, uye mushandisi maitiro.
- Gadzirira nzira dzekudzosera kumashure uye dzezviitiko usati wawedzera.
Ramba Uchiongorora
Tsamba yemahara
Tora iyo yemazuva ese AI pfupiso
Nhatu dzakasimbiswa dzeAI nyaya mangwanani ega ega pakati pevhiki, dzakanyorwa neChirungu chakajeka. Yemahara zvachose, hapana ads.
Imwe email zuva rega rega revhiki. Bvisa kunyoresa kamwe chete. Hatimbotengese kana kugovera kero yako.
Zviedze iwe pachako
Take the InfoNCE and SimCLR Objectives quiz
Mhinduro yepakarepo pamhinduro yega yega, uye chitupa chinogovaniswa chine ID inobatika kana wapedza kosi.
Tsigira dzidzo yemahara yeAI. AI Understanding ndeye 501(c)(3) isingabatsiri - hapana ads, hapana paywall, zvachose. Ita chipo
Mibvunzo inowanzo bvunzwa
Chii chinonzi InfoNCE uye SimCLR Zvinangwa?
InfoNCE ndiko kurasikirwa kwakasiyana kunodzidzisa modhi kudhonza mapeya anoenderana pamwe chete uye kusundira zvisingaenderane peya nepakati mukumisikidza nzvimbo. SimCLR inzvimbo yakakosha iyo yakashandisa kurasikirwa uku kudzidza ine simba inomiririra mifananidzo kubva kune isina kunyorwa data, kukwikwidza inotariswa pretraining.
Chii chinonzi InfoNCE chinangwa chinokurudzira modhi kuita?
InfoNCE ndeye softmax pamusoro pekufanana kunopa mubayiro kufanana kwepamusoro kwechokwadi chakanaka uye chakaderera kufanana kwezvisina kunaka.
MuSimCLR, vaviri vakanaka vanogadzirwa sei?
SimCLR inogadzira maviri akanaka kubva kune maviri akawedzera maonero eiyo yakafanana sosi mufananidzo; mimwe mifananidzo inoshanda seyakaipa.
Ndeipi basa rinoitwa tembiricha τ muNT-Xent / InfoNCE?
Kugovanisa kufanana ne τ pamberi pe softmax inodzora kuti kurasikirwa kunosiyanisa zvakakomba sei kubva kune dzakaoma dzisina kunaka.
Chii chinoitika kune SimCLR's fungidziro musoro mushure mekudzidzira?
SimCLR yakaona kuti maficha pamberi pekufungidzira musoro kutamiswa zviri nani, saka musoro unokandwa kure mushure mekudzidzira.
Sei SimCLR yaivimba nehukuru hukuru?
MuSimCLR mamwe mapikicha ari mubatch anoita seasina kunaka, saka mabheti akakura anopa zvimwe zvakaipa uye kudzidza kwakasimba.
Ramba uchidzidza
Mazano anoenderana
Mamwe madhairekitori akasarudzirwa nyaya iyi