Maswali yanayounganishwa na mwongozo · Ngumu Kiwango

Maswali ya Uzushi wa Kushuka Mara Mbili

Kagua asili mbili karibu na ufasiri na jinsi jambo hilo linavyoweza kutofautiana katika ukubwa wa kielelezo, saizi ya seti ya data na muda wa mafunzo.

Njia za mwongozo zinazohusianaUzushi wa Kushuka Mara Mbili
Swali 1 ya 8

Kosa la jaribio kawaida hufikia wapi kwenye safu ya asili mbili?

Hakiki kila swali8 maswali

Majibu na maelezo huonekana unapojibu maswali.

  1. Kosa la jaribio kawaida hufikia wapi kwenye safu ya asili mbili?
    • Katika kizingiti cha ukalimani ambapo modeli inafaa tu data zote za mafunzo
    • Mwanzoni mwa mafunzo na vigezo vichache
    • Tu baada ya mabilioni ya vigezo
    • Hitilafu ya mtihani haifikii kilele; inaanguka monotonically
  2. Ni nini hufanyika kwa makosa ya jaribio kwani modeli inazidishwa sana?
    • Inainuka kwa muda usiojulikana kwa sababu ya kupita kiasi
    • Inaweza kupungua tena, mara nyingi chini ya kiwango cha chini cha classical
    • Inakaa gorofa kabisa
    • Inakuwa nasibu na haitabiriki
  3. In which setting does the cited implicit-bias analysis characterize a minimum-norm solution?
    • Any neural network trained with gradient descent
    • Any model that reaches zero training error
    • A single-hidden-layer linear network under stated gradient-flow and initialization assumptions
    • A model trained without an optimization objective
  4. 'Epoch-wise' asili maradufu inarejelea athari inayoonekana kama kazi ya nini?
    • Idadi ya GPU zilizotumika
    • Ratiba ya viwango vya kujifunza pekee
    • Ukubwa wa seti ya mtihani
    • Muda wa mafunzo (idadi ya nyakati)
  5. Ni wazo gani la kitamaduni linaloleta changamoto ya uzao maradufu?
    • Ubadilishanaji wa upendeleo wa umbo la U
    • Uenezaji wa nyuma
    • Kushuka kwa gradient ya Stochastic
    • Kanuni ya mnyororo
  6. Ni nani aliyesaidia kutangaza na kurasimisha mkondo wa asili maradufu mwaka wa 2019?
    • Newton na Leibniz
    • Belkin, Hsu, Ma, na Mandal
    • Turing na Kanisa
    • Washindani wa Kaggle tu wasiojulikana
  7. Which model families did OpenAI report studying in its deep double-descent experiments?
    • Only decision trees and linear regression
    • Only recurrent speech recognizers
    • Convolutional neural networks, ResNets, and transformers
    • Only graphical game generators
  8. Besides model size and training duration, which axis can also show double-descent behavior?
    • Font size in a report
    • The number of words in a product name
    • The screen resolution of a monitor
    • Training dataset size