For one trained RNN, what’s the purpose of sampling a sequence by randomly selecting one y as input of x<t+1>? for validation?
Thanks a lot!
Hello, Leiming.
Welcome to the community.
If you go through the slide https://www.coursera.org/learn/nlp-sequence-models/lecture/MACos/sampling-novel-sequences, you will get the answer in just 1.30 minutes of the video on why are we using the sampling in a trained RNN.
Also, there’s a bit-by-bit explanation given in the instructions in W1 A2 that clarifies all your doubts.
I am picking it up for you:
