Ci bi mu melGis bi ci topp
Wàcci degrade stokastik ak saa
Fondamentaal yi
GUIDE bu am solo
Dëddu ak yamale stokastik
Dropout ab pexe regularisasioŋ la buy dindi ci anam wu mucc benn wàll ci neuron yi ci jéego bu nekk ci tàggat yaram, di forse reso bi mu tabax ay représentation yu bari te dëgër.
Déglul tegtal biiIA-nettali · ~2 min
Ci xët wii2 simili jàng
Résumé
It became one of the most influential techniques for fighting overfitting in deep learning.
Plongeur bu xóot
Groupe Hinton moo ko dugal ci atum 2012, dropout dafay wax ci benn ñakk kattan bu mag ci reso yu mag yi: neuron yi mën nañu ànd ànd, jàng saafara njuumti yi ci seen biir ci anam wudul liggéey ci done yiñ tàggat. Bépp paas bu jëm kanam ci diiru tàggat yaram, dropout dafay def génnug neuron bu nekk ci nul ak yenn probabilite p (dafay faral di nekk 0.5 ci diisaay yu dëgër). Ndax bépp neuron mën na ni mes, reso bi mënul wéeru ci jàppante yu yomba dagg, te dafa wara tasaare xibaar bu am njariñ ci yunit yu bari. Loolu dafa melni tàggat ensemble bu mag bu reso yu sew yu bokk poids. Biñu yeggee ci test bi, dañuy fay dropout ba noppi jëfandikoo reso bi yépp, ak activations yuñ scale ngir génne gi ñuy seentu méngoo ak tàggat bi. Resultaa bi mooy generalisation bi gëna baax ci njëgu tàggat bu gëna gudd tuuti.
Doxal ci àdduna dëgg
Yokk benn couche Dropout ak p ci diggante 0.5 ci digganté couche yu dëgër yi ci benn nataal wala mbind buy xaaj ci PyTorch wala Keras
Modèle transformateur yiy jëfandikoo dropout ci diisaayu bàyyi xel ak joxe feed-forward ci diiru tàggat yaram
Monte Carlo dropout, fu dropout des ci inference ngir génne xayma yu wóorul ci wàllu pajum wala kaaraange-gëna am solo
xóotaayu stokastik (DropPath) dafay sànni ay blok yu des ngir yamale reso yu xóot lool yu melni ResNets ak transformatëri gis-gis
Ëlëgu bàyyi jàngi ak yamale Stochastic
Ci reso yiy xool konvolusionel, normalisasioŋ ci lots dafa toxal bu baax dropout buñ miin, waaye anam yi ñuy jëfandikoo dañuy gëna am doole feneen: transformateur yi dañuy jëfandikoo dropout ci bàyyi xel ak feed-forward, ak DropPath (xootaayu stochastic) dafay daaneel bloc residuel yépp. Monte Carlo dropout, mooy tëye dropout bi ci inference, ñu koy jëfandikoo ngir xayma ñàkka wóoru model bi. Xaarandil ni yamale stochastic nekk jumtukaay bu yomb, buñu méngale ak architecture bu nekk te du benn rëset buñ tëral.
Teg ko ci jëf
Lu tax mu am solo
- dogal yu gëna leer. Daf lay jàppale nga tàqale kàddu yu leer ci wàllu xarala ak làkku fësal njaay.
- Njëgg ak budget. Mën nga laaj laaj yu gëna baax ci samp gi balaa ngay dugal xaalis wala sa jotu liggéey.
- Ekip ak def liggéey. Ekip yi bokk xam-xam ñoo gëna mëna jël yenn dogal ci wàllu produit, politik ak jàng.
Moytul
- Ekip yu bari mën nañu jëfandikoo benn baat ci anam wu wuute, kon teela leeral yaatuwaayam.
- Benchmark yi mën nañu nuru lu am doole waaye performance yi ci àdduna bi duñu tolloo.
- Bëgg kalite done ak palaŋu jàngat dafay faral di jur njariñ yu yomba dagg.
Jéego yi ci topp
- Tàmbaleel ci joxe leeral ci làkk wu leer ci njariñ li nga soxla.
- Tannal benn metric bu baax ak benn anam bu baaxul balaa ngay saytu.
- Doxal ab pilote bu ndaw ak ay done yu representatif, du ab demo bu leer.
- Dokument fu Dropout ak Regularisasioŋ Stochastic di jàppale ak fu pexe yu gëna yomba gëna baax.
Weyal di banneexu
Bulletin bu amul fayda
Jàngal xibaar AI bu bes
Ñatti jaar-jaari IA yuñ firndeel bis bu nekk ci ayu-bis bi, ñu bind leen ci Àngle bu leer. Gratis ba fàww, amul yëgle.
Benn email bis bu nekk ci ayu-bis bi. Benn klik nga dindi abonemaa bi. Dunu musa jaay wala séddoo sa adres.
Nattal sa bopp
Take the Dropout and Stochastic Regularization quiz
Feedback ci saasi ci bépp tontu, ak sertifikaa buñ mëna séddoo ak ID buñ mëna saytu soo jàllee ci njàng mi.
Jàppale njàngum IA te doo fay. AI Understanding mbootaay la buy def te Yàlla tax — amul yëgle, amul paywall, musul am. Defal ab sarax
Laaj yi ñuy faral di laaj
What is Dropout and Stochastic Regularization?
Dropout ab pexe regularisasioŋ la buy dindi ci anam wu mucc benn wàll ci neuron yi ci jéego bu nekk ci tàggat yaram, di forse reso bi mu tabax ay représentation yu bari te dëgër. Nekk na benn ci pexe yi gëna am solo ngir xeex lu ëpp li ñuy jàng ci njàng mu xóot mi.
Lu dropout di def ci diiru tàggat yaram?
Dropout dafay nul neuron bu nekk ak probabilite p ci diiru tàggat yaram, kon ñu jëfandikoo beneen sous-réseau bu sew ci jéego bu nekk.
Lan moo waral bàyyi jàngi gëna yombal généralisasioŋ bi?
Suñu dindie ay yunit ci anam wu mucc, dropout dafay tere neuron yi sos ay lëkkaloo yu yomba dagg yuy liggéey ci done yu tàggat rek, gëna dëgër.
Lan mooy dall ku bàyyi jàngi ci waxtu examen (inference)?
Ci gàttal, reso bi yépp dafay dox ak dropout bu desee, te aktivasioŋ yi dañu leen di eskale ngir génne yi ñuy seentu méngoo ak séddaleb tàggat bi.
Tolluwaayu bàyyi njàng bu p = 0.5 ci benn layer, lu muy tekki ci diiru tàggat yaram?
Ak p = 0.5 neuron bu nekk amna 50 pursaa chance ñu zero, kon ci moyenne lu tollu ci genn-wàll dañuy daanu ci pass forward bu nekk.
Lan mooy dropout luñuy faral di wax ni jegewaale?
Sampling beneen sous-reseau jéego bu nekk mingi jege limu exponentiel ci reso yu bokk poids, xeetu ensembling bu yomb.
Weyal di jàng
Gid yu jëm ci loolu
Tann nañu yeneen njiit ngir topic bii