Expand NLP Model

[nickumia]

I think the model from my master's thesis is still very powerful. How can I use the concepts from the past and the new AI concepts I'm learning to build a better model.

I also think that the controller method for learning might still have some value. How does this work when combined with reinforcement learning concepts??

[nickumia]

Related to the controller method:

flowchart LR
  T[[Transfer Function]]
  s1[[1/s]]
  s2[[1/s]]
  feedback1[[Feedback]]
  feedback2[[Feedback]]
  feedback3[[Feedback]]
  input(Input)
  output(Output)
  sum((x))
  input -- + -->sum
  sum --> T
  T -- a -->s1
  s1 -- v -->s2
  s2 -- x -->output
  output --> feedback3
  feedback3 -- "-" --> sum
  s2 --> feedback2
  feedback2 -- "-" --> sum
  s1 --> feedback1
  feedback1 -- "-" --> sum

In traditional control:

• S --> constant --> variable
  • "don't consider how it changes. All of the defining points are independent of S
• tau and K define system (are constant)