Fonction yuy ñàkk
Fonction perte mooy benn nimero biy wax model bi ni prediction yi juum, soppi mébet bu leerul def ko lu math mëna optimiser.
Résumé
Choosing the right loss shapes what the model actually learns.
Plongeur bu xóot
Bépp model buñu tàggat mingi soxla joxe leeral bu leer ci gacce, te loolu mooy li fonction perte di joxe. Dafay méngale li model bi waxoon ak tontu bu dëggu bi ba noppi génne benn lim: lu gëna kawe dafay tekki lu gëna bon. Kon tàggat yaram mooy wàññi limu nit ñi. Tanneef ngir ñàkk du kosmetik. Ci liggéeyu regression, njuumte kaare moyenne dafay yar njuumte yu mag yi bu baax ci kaare différence bi, waaye njuumte absolu moyenne dafay gëna yamale njuumte yépp te baña nangu njuumte yi. Ci wàllu xaaj, entropi cross-loss dafay natt ba fu distribution probabilite bi ñu séentu nekk ci etiketu dëgg bi, di yar tontu yu juum yi wóolu seen bopp bu baax. Tann perte bu méngoowul ak sa mébet mën na tax model bi gëna xarañ ci lu jaarul yoon, kon fonction perte bi dafay encoder li nga bëgg.
Gis-gis xarala
Entropi croix, di fas wi ñuy liggéey ngir xaaj, mingi joge ci theorie information: dafay natt bit yu gëna bari yiñ soxla ngir kode etiketu dëgg yi ci jëfandikoo probabilite yiñ séentu ci model bi. Ndax dafay màgg bu baax su amee wóolu boppam lu jaarul yoon, gradient bi dafay puus model bi bu baax ngir saafara njuumti yu wóolu boppam yu ëpp. Fonction perte war nañu wuute (wala nuru) ndax backpropagation dafa soxla seen gradient. Loolu moo waral ñuy jëfandikoo surrogate yu nooy ci barabu metrics yu ñor, yuñ mënul wuutale lu melni njub.
njeextalu pexe
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.
Ëlëgu liggéey yiy ñàkk
Design fonction loss mingi gëna bari ci barab bi IA bu bees bi di doxalee. Lu weesu cross-entropy buñ miin, pexe yu melni nooy etiketu, ñàkka focal ngir done yu tolloowul, ak ñàkka wuute ngir jàng representation dañuy nekk lu ñuy faral di def leegi. Ci misaali làkk yu mag yi, mébetu tàggat bi ak dooleel-jàng-ci-feedback xeetu neexal yi dañuy gëna bari ay perte yuñ defar bu baax ngir doxal ton, jàppale, ak kaaraange. Xaarandil yokkute bu wéy ci perte yuñ jagleel ak composite yu boole ay mébet yu bari, ndax ñooy benn ci levier yi gëna direct ngir saytu li benn model di valeur.
Doxal ci àdduna dëgg
Jëfandikoo cross-entropy loss ngir tàggat ab xaajkat spam imeel buy daanel xaajkat yu wóolu
Tann moy njuumte absolu ngir xayma njëgu kër suko defee yenn kër yu mag duñu ëpp doole ci tàggat yaram
Jëfandikoo ab ñakk wute ngir ab xeetu xàmmee kanam buy boole nataali benn nit
Defar ab xeetu neexal buy ñàkk ngir yóbbu ab chatbot ci tontu yu gëna am njariñ te gëna jub
Risk yi ak balustrade yi
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.
Roadmap ngir samp gi
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.
Dokumenteer fi Loss Functions di jàppale ak fi pexe yu gëna yomba gëna baax.
Weyal di banneexu
Free newsletter
Get the daily AI briefing
Three verified AI stories every weekday morning, written in plain English. Free forever, no ads.
One email each weekday. Unsubscribe in one click. We never sell or share your address.
Test yourself
Take the Loss Functions quiz
Instant feedback on every answer, and a shareable certificate with a verifiable ID once you pass a course.
Support free AI education. AI Understanding is a 501(c)(3) nonprofit — no ads, no paywall, ever. Make a donation
Gis bi ci topp
Triplet ñàkk ak jàng metric
Laaj yi ñuy faral di laaj
What is Loss Functions?
Fonction perte mooy benn nimero biy wax model bi ni prediction yi juum, soppi mébet bu leerul def ko lu math mëna optimiser. Tann njuumte bu dëggu bi mooy wane li model bi di jàng.
Luy fonction perte natt?
Fonction perte dafay genne benn nimero buy xayma gap bi am ci digganté liñu wax luy waaja am ak tontu yu dëggu yi; lu gëna suufe moo gën.
Ban fonction perte lañu gëna jëfandikoo ci liggéeyi xaaj?
Entropi croix dafay méngale seddaleb probabilite yiñ séentu ak etiketu dëgg yi te mooy tànneef buñ miin ngir xaaj.
Buñu ko méngale ak njuumte kaare moyenne, lu tax nga tànn njuumte absolu moyenne?
Jumtukaay absolu moyen dafay jappee njuumte yépp ci anam wu méngoo, du leen kaare, kon yenn outlier yu tar ñoo gëna néew doole.
Lan moo waral fonction perte dafa wara wuute?
Backpropagation soxla na gradient bu ñàkk bi ngir yeesal poid yi, kon ñàkk bi dafa wara wuute wala am surrogate buñ mëna jëfandikoo.
naka la cross-entropy di doxalee ci wax lu wóor waaye lu jaarul yoon?
Entropie croix dafay màgg bu baax sudee model bi dafa juum ci wóolu boppam, loolu mooy defar gradient bu dëgër ngir saafara njuumte gi.