Idanwo ti o sopọ mọ itọsọna · Alabọde Ipele

Nesterov Accelerated Gradient Quiz

Test current versus look-ahead gradients, momentum correction, convex-rate scope, comparable cost, Nadam, and a numeric look-ahead step.

Jẹmọ itọsọna awọn ọnaNesterov Accelerated Gradient
Ibeere 1 ti 8

Kini imọran asọye ti Nesterov Accelerated Gradient ni akawe si ipa kilasika?

Awotẹlẹ gbogbo ibeere8 ibeere

Awọn idahun ati awọn alaye han bi o ṣe n mu ibeere naa.

  1. Kini imọran asọye ti Nesterov Accelerated Gradient ni akawe si ipa kilasika?
    • O mu ipa kuro patapata
    • O nlo oṣuwọn ẹkọ ti o wa titi ti 1.0
    • O ṣe iṣiro gradient ni aaye wiwo-iwaju ju ipo lọwọlọwọ lọ
    • O ṣiṣẹ nikan laisi awọn ipele kekere
  2. Oṣuwọn isọdọkan ti o ṣeeṣe wo ni ọna Nesterov ṣaṣeyọri lori awọn iṣoro rubutu ti didan?
    • Paṣẹ 1/k^2
    • Paṣẹ 1/k
    • Bere fun 1/log k
    • Ko si oṣuwọn iṣeduro
  3. Tani ni akọkọ ṣe afihan ọna isare isare yii?
    • Geoffrey Hinton ni ọdun 2012
    • Isaac Newton
    • John von Neumann
    • Yuri Nesterov ni ọdun 1983
  4. Kini idi ti iwo-iwaju ṣe iranlọwọ ni itosi minima ti o tẹ?
    • O mu ki eko oṣuwọn
    • O ni ifojusọna overshoot ati pe o kan atunse ṣaaju ki o to ṣẹlẹ
    • O npa ọrọ iyara rẹ kuro
    • O randomizes itọsọna igbese
  5. Ni iṣe, bawo ni iye owo iširo ti ipa Nesterov ṣe afiwe si ipa ti kilasika?
    • Ni aijọju kanna, o ṣeun si imudojuiwọn ti a ṣe atunto algebra kan
    • Nipa mẹwa ni igba ti o ga
    • O nilo awọn itọsẹ keji ni gbogbo igbesẹ
    • Ko le ṣe iṣiro lori awọn GPU
  6. Iru iṣapeye adaṣe wo ni o ṣajọpọ iwo-iwaju Nesterov pẹlu igbelowọn ara-ara Adam?
    • RMSprop
    • Adagrad
    • Nadamu
    • LBFGS
  7. Using the stated look-ahead form, theta = 10, learning rate = 0.1, beta = 0.9 and velocity = 2 give which evaluation point?
    • 10.18
    • 9.82
    • 8.20
    • 10.90
  8. If beta is zero in the guide’s Nesterov look-ahead formula, where is the gradient evaluated?
    • At a point displaced by the full old velocity
    • At the previous validation score
    • At an arbitrary future minibatch
    • At the current parameter position