Ilọsile Sitokasitik Gradient pẹlu adanwo akoko

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

1. Kini ọrọ ipaniyan n ṣajọpọ lakoko ikẹkọ?
   • Lapapọ nọmba ti ikẹkọ epochs
   • Apapọ ṣiṣiṣẹ (ti o ni iwọn pupọ) ti awọn gradients ti o kọja
   • Awọn iwọn ti awọn dataset
   • Awọn awoṣe ká išedede itan
2. Ni gigun gigun, afonifoji dín, iṣoro wo ni SGD itele ti jiya pe ipa naa ṣe iranlọwọ lati ṣatunṣe?
   • O kún àkúnwọ́sílẹ̀ ní iye
   • O da awọn gradients iširo duro
   • O mu ki eko oṣuwọn laifọwọyi
   • O zig-zags kọja awọn odi giga lakoko ti o nlọsiwaju laiyara lẹgbẹẹ ilẹ
3. Iye aiyipada ti o wọpọ fun olusọdipúpọ ipa ni:
   • 0.9
   • 0.0
   • 0.5
   • 2.0
4. Apejuwe ti ara wo ni a maa n lo julọ lati ṣe apejuwe ipa?
   • A pendulum ti n yipada lailai
   • A orisun omi snapping pada
   • Bọọlu ti o wuwo ti n yi lọ si isalẹ
   • Omi didi
5. Ni aijọju melo ni ipa ti o pọ si igbesẹ ti o munadoko ni itọsọna deede nigbati beta = 0.9?
   • Nipa awọn akoko 2
   • Nipa awọn akoko 0.1
   • Ko si ampilifaya
   • Nipa awọn akoko 10
6. Iru iṣapeye ode oni wo ni o ṣafikun iṣesi-ara-iṣiro-akoko akọkọ?
   • Newton ká ọna
   • Adamu
   • K-tumo si
   • wiwa laileto
7. In the update v = beta times v plus gradient, what happens when beta is zero?
   • The old velocity is doubled
   • Every gradient becomes zero
   • The previous velocity is discarded and the current gradient determines the step
   • The parameters stop depending on the learning rate
8. Why can repeated gradients with opposing sideways directions produce less zig-zagging under momentum?
   • Their opposing contributions can partially cancel in accumulated velocity
   • All past gradients are stored as separate parameters
   • The loss becomes convex automatically
   • Momentum ignores the current gradient entirely