rnn is confusing. i have watched 4 videos so far. is this correct diagram? please correct me if not
what that box represents? a layer in a networks , neural network or ? sorry if its silly question
Its a hidden neural layer or a layer of neurons. This video does a pretty good explanation I think:
1 Like
That box is the “cell” of the RNN. There is just one and it gets used repeatedly. That’s the key architectural difference between RNNs and all the previous networks like FCNs and CNNs that we have learned about in DLS C1 - C4, all of which have multiple distinct layers. In an RNN, there is just one “cell” or layer, but it gets reused at every timestep. It learns weights that it uses to maintain its cell state that controls how it processes the inputs into the outputs at each timestep.
If you missed that aspect of the architecture, it might be a good idea to go back and listen to the lectures in C5 W1 from the beginning. This is a pretty fundamental point and Prof Ng did explain it in the lectures, including the one from which Gent showed that slide. Things should make a lot more sense listening again with what I said above in mind.
2 Likes