Kedis ku xidhan hagaha · Adag Heerka

Kediska Aragtida Kernel Tangent Neural

Tijaabi sida Kernel-ka Neural Tangent Kernel uu ugu xidho tababbarka shabakada ballac-ballaadhan ee hababka kernel-ka iyo firfircoonida-meel bannaan ee xadka toosan.

Wadooyinka haga ee la xidhiidhaAragtida Kernel Tangent Neural
Su'aal 1 ee 8

Xadka aan dhammaadka lahayn, sidee buu u dhaqmaa Kernel Tangent Neural inta lagu jiro tababarka?

Horudhac su'aal kasta8 su'aalo

Jawaabaha iyo sharraxaadaha waxay soo baxayaan markaad qaadanayso kediska.

  1. Xadka aan dhammaadka lahayn, sidee buu u dhaqmaa Kernel Tangent Neural inta lagu jiro tababarka?
    • Dhab ahaan waa joogto
    • Si degdeg ah ayey u beddeshaa tillaabo kasta
    • Eber ayay noqotaa
    • Aad bay u kortaa
  2. Waa maxay NTK xisaab ahaan?
    • Hessian ee khasaaraha
    • Wax soo saarka gudaha ee jaangooyooyinka halbeegga shabakadda ee laba qalab
    • The softmax ee wax soo saarka
    • Kala duwanaanshaha miisaanka
  3. Yaa keenay Kernel Tangent Neural?
    • Hinton iyo Salakhutdinov
    • LeCun iyo Bengio
    • Jacob, Gabriel, iyo Hongler (2018)
    • Vaswani iyo asxaabtiisa
  4. Waa maxay nidaamka 'tababarka caajisnimada'?
    • Marka tababarka la joojiyo
    • Marka optimizer-ku aanu isticmaalin wax dardar ah
    • Marka xogta si tartiib ah loo quudiyo
    • Marka cabbiradu ay si dhib yar uga guuraan bilawga oo shabakadu u dhaqanto si toos ah
  5. Xaddidaadda muhiimka ah ee aragtida NTK ee shabakadaha dhabta ah waa in ay seegto:
    • Muujinaya barashada shabakadaha balaladhka ah
    • Doorka hoos-u-dhaca
    • Jiritaanka miisaanka
    • Isticmaalka xogta tababarka
  6. Sababtoo ah NTK waxay yaraynaysaa tababarka dib u noqoshada kernel, maxaa suurtogal noqon kara?
    • Baabi'inta baahida xog kasta
    • Saadaasha wax soo saarka la tababaray iyada oo aan si buuxda loo tababarin
    • Moodooyinka socda iyada oo aan la xaddidin
    • Dammaanad qaadida qaladka imtixaanka eber had iyo jeer
  7. In the infinite-width limit, what probabilistic model are neural networks equivalent to at initialization?
    • Qaabka qarsoon ee Markov
    • Go'aan geed go'aan
    • A Gaussian process
    • A finite-state game controller
  8. For least-squares regression in the NTK limit, what mathematical form does the network function's training evolution follow?
    • A discrete rule that randomly changes labels
    • A decision tree that grows one branch per epoch
    • A nonlinear search over audio frames
    • A linear differential equation in function space