Two Starting Points, Two Destinations — easy Gradient Descent problem | Incognition
Two Starting Points, Two Destinations
10 pts · 10 coins
Gradient DescentEasy
Two Starting Points, Two Destinations
A one-parameter loss is
L(w)=4w4−3w3−3w2
Gradient descent is run twice with a very small learning rate and left to run until the gradient is essentially zero. The first run starts at w0=−0.5; the second starts at w0=+0.5.