I’ve doubt on this question
Another Doubt I’ve on this :
please help
@Manu-Pillai-1566551720093198 @rachitbansal2500 @Rahul_garg
I have edited the question, it was supposed to ask for the tth time step.
A tth state vector would also depend on the y that that particular cell outputs, that’s what the concept of RNN says, the activation output from a particular cell contains within itself all the previous outputs from the RNN cells and the one from the present time step as well. Hence, option 3 would be correct too.
The 4th option was supposed to be a< 1 >, it was a typing mistake from our side which has been edited now. I hope you understand why that would be correct too.
Hence, the answer would be ‘All of the above’
I still doubt on this @rachitbansal2500
any cell at time t has 2 inputs only a<t-1> , x<t> which outputs a<t> . It means for a<t> we need 2 things only.
But at time t for output y<t> we need 3 things : a<t-1> , x<t> , a<t>.
But here question is asking about a<t> so i guess 2 options are correct.
At any time t, how can a<t> be input to itself, coz it is the activation output at time t , so a<t> cant be dependent on a<t>
@mohituniyal2010 This image might help you get more clarity:
As it’s visible, the y from a RNN cell plays a role in forming the final activation output from that cell, which also logically evident- consider the ‘y’ term to be a term representing an output from one cell while the activation output represents a collective output from that cell and all the previous cells that have been passed by so far.
Yes, as I mentioned in my previous option that option had to be a<1> and not a<t>, it was a typing mistake previously and has been corrected now.
that’s why i had written this earlier
so the first statement is correct right,?? then I got your point…
Yes, the first statement is correct
