$(3)$

$1$ point possible $($graded$,$ results hidden$)$

Fixing $s_0,W^{s,s},W^{s,x}$ and by only learning the linear classifier in the final layer, can this RNN separate the $3$

examples regardless of how they were labeled?

Yes

No