Dear Mentor,
Could you please guide me why my answer is wrong in this question?
In my opinion,
at the time step 0, the RNN estimates P(y<1>)
at the time step 1, the RNN estimates P(y<2> | y<1>)
…
at the time step t, the RNN estimates P(y<t+1> | y<1>…y< t > )
Thank you.
No it a recurrent neural network, the y-th element depends on all previous elements, its a chain can you see!
I think you’re confusing yourself a little bit with the wording. The process starts at step 1, not step 0. Step 1 would be the first step, but with indexing you index step 1 as 0. P(y | y<1> … y) is read as “Possibility of y given y<1> up to y” Because you’re predicting y given everything you’ve seen before the current step (t). You’re not predicting the next step like you put it as y<t+1>.
