Updating Meta learner parameters

[smkh1985]

It's mentioned in the original article that update's rule is : Phi <--- Phi - Eps * (W - Phi)
why didn't you use Eps and why you used (Phi-W)

[gabrielhuang]

The learning rate can be tuned through the optimizer, and the minus sign is due to the fact that the optimizer goes in the opposite direction to the gradient (therefore the sign is flipped)

…

On Sun, Jul 18, 2021 at 12:57 PM majid ***@***.***> wrote: It's mentioned in the original article that update's rule is : Phi <--- Phi - Eps * (W - Phi) why didn't you use Eps and why you used (Phi-W) — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#8>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AB3P5RDH4MH27GG5PJZZPLDTYMBZJANCNFSM5ASJ3LIQ> .

[Minhvt34]

Hi author,
I am curious about the "eps" parameter. To perform SGD for the meta step, why don't you use eps?