week 6, ch 6.3

[p-mishra1]

The equation and the image do not support each other, in the image the input x[t-1] should be x[t] as mentioned in the equation @Atcold .

[Atcold]

Central column of the right image: h[t] = f(h[t-1], x[t]). Which means, the current hidden state is function of the previous hidden state & the current input variable. What does not look correct to you?

[p-mishra1]

yes exactly the equation is correct but in the model diagram both the hidden state as well as the input are of same time stamp h[t] = g(h[t-1], x[t-1]) this is what is mentioned in the model diagram. am i missing something here? i am unable to associate hidden states with the arrows should it be associated with upper arrow(what you are mentioning i suppose) or arrow coming from left(what i am referring to). So i think the diagram is ambiguous.

[Atcold]

I see. The h[t-1], h[t], and h[t+1] should be ‘inside’ the circle, but they don't fit, so they are placed outside. Last year I drew a better version of this diagram.

You can see the full tweet here.