Skip to content
ml-recall

#gradient-checking

2 questions from Neural Network Basics

Timed test

2 questions · 2 min · score and explanations at the end

Take a test
Q 1 / 2Score 0 / 0

Question 1: Short answer

mediumAI-checkedNeural Network Basics

To check a backpropagated gradient of L(θ)=θ3L(\theta) = \theta^3 at θ=2\theta = 2 (exact value 12), you use the step ε=0.01\varepsilon = 0.01. The central difference [L(θ+ε)−L(θ−ε)]/(2ε)[L(\theta + \varepsilon) - L(\theta - \varepsilon)]/(2\varepsilon) gives , and the forward difference [L(θ+ε)−L(θ)]/ε[L(\theta + \varepsilon) - L(\theta)]/\varepsilon gives . Give 4 decimals.

Fill in every blank (hints in the boxes). Enter moves to the next blank.