Maswali yanayounganishwa na mwongozo · Ngumu Kiwango

Majukumu ya Ushawishi kwa Maswali ya Maelezo ya Data ya Mafunzo

Hukagua viambajengo vya utendaji wa ushawishi, makadirio ya Kihessian, maana ya ishara za ushawishi na maelezo ya data katika miundo ya lugha.

Njia za mwongozo zinazohusianaKazi za Ushawishi kwa Uwasilishaji wa Data ya Mafunzo
Swali 1 ya 8

Ni swali gani la uwongo ambalo huathiri utendakazi takriban?

Hakiki kila swali8 maswali

Majibu na maelezo huonekana unapojibu maswali.

  1. Ni swali gani la uwongo ambalo huathiri utendakazi takriban?
    • Jinsi hasara kwenye sehemu ya jaribio ingebadilika ikiwa mfano wa mafunzo ungeondolewa au kuongezwa uzito
    • Jinsi sahihi ya mfano kwa wastani
    • Jinsi mtindo unavyofanya mazoezi kwa kasi
    • Mfano una vigezo ngapi
  2. Ni kitu gani cha kihesabu hufanya hesabu halisi ya ushawishi kuwa ngumu kwa mifano kubwa?
    • Kazi ya softmax
    • Hessian inverse ya hasara
    • Ratiba ya kiwango cha kujifunza
    • Ukubwa wa kundi
  3. Ni nani aliyebadilisha kazi za ushawishi kwa ujifunzaji wa kina wa kisasa katika karatasi inayojulikana ya 2017?
    • Wema na Bengio
    • Vaswani et al.
    • Koh na Liang
    • Hinton na LeCun
  4. Under Koh and Liang’s standard I_up,loss convention, what does a negative value indicate?
    • The point was not in the training data
    • Upweighting the point is estimated to increase test loss
    • The point has no relationship to the test example
    • Upweighting the point is estimated to lower test loss, so it is helpful
  5. Anthropic ilitumia makadirio gani kuongeza utendakazi kwa miundo mikubwa ya lugha?
    • EK-FAC (Mviringo unaotokana na Kronecker)
    • Kuacha
    • Utafutaji wa boriti
    • Mchanganyiko
  6. Utafiti wa Anthropic ulipata nini kuhusu mifano yenye ushawishi?
    • Daima ni nakala halisi za pato
    • Mara nyingi hushiriki muundo wa kufikirika badala ya maneno yanayofanana
    • Zinasambazwa bila mpangilio
    • Mfano wa mafunzo ya kwanza tu ndio muhimu
  7. In the influence-function approximation, what connects a training-example gradient to the effect on a test loss?
    • The test set’s token count
    • A random forest over training labels
    • The inverse Hessian of the training loss
    • The model’s output vocabulary size
  8. If removing one training example is estimated to increase a test point’s loss, what does that suggest about the example?
    • It was guaranteed to be a duplicate
    • It had no relationship to the model
    • It necessarily caused the model to overfit
    • It had a helpful influence on that test point under the approximation