Algorithm 1: Environmental actions through the global feedback

Input: the observables $Y_n^k$

Output:$c_{k+1}$ as the final action in environmental actions mode

Initialization:

Pre-adaptive actions:

Reduce data rate till BER<FEC-threshold

Make steady state mode off, prediction mode on

Probability box% = 96, threshold = 0.01, $PF=2$, j = 1

$c_{k^{\prime }+1}\leftarrow firstactionfromC$

Apply $c_{k^{\prime }+1}$ to the environment (fiber optic channel)

1: for $k=(k^{\prime }+1)toK$ (see Table 3)

2:   Take the observable $Y_n^k$

3:    if the model is not available then

4:    if BER > FEC-threshold then

5:    Estimate $P({\widehat{X}}_n^k|{\stackrel{⌢}{Y}}_n^k)$ using $P({\widehat{Y}}_n^{k-1}|{\widehat{X}}_n^{k-1})$

6:    Estimate ${\overline{X}}_n^k$ by decision making

7:   else

8:     Extract the $P({\widehat{Y}}_n^k|{\widehat{X}}_n^k)\mathrm{and}P({\widehat{Y}}_n^k)$

9:      Calculate the posterior $P({\widehat{X}}_n^k|{\widehat{Y}}_n^k)$

10:   Estimate ${\overline{X}}_n^k$ by decision making

11:  end if

12:  else if model is available then

13:  Load model, evidence and posterior from preceptor library

14:   Estimate ${\overline{X}}_n^k$ by decision making

15:  end if

16:   Send PF1, $P({\widehat{Y}}_n^k|{\stackrel{⌢}{X}}_n^k)\mathrm{and}P({\widehat{Y}}_n^k)$ to the executive

Internal reward

17:  Calculate $BE{R}_k,andr{w}_j{}^k$ and send it to executive

Planning

18:  Localize the set of all close actions to ck

Learning

19:  Apply $c_{k+1}$ virtually ($c_{k+1}\in C$)

20:  Calculate $P(X_n^{(k+1)}|{\widehat{Y}}_n^k)$

21:  Predict ${\overline{X}}_n^{(k+1)}$

22:  Calculate $BE{R}_{k+1}andr{w}_{j+1}^{k+1}$

23:  if $BE{R}^{(k+1)}\le threshold$

and $r{w}_{j+1}^{k+1}\ge r{w}_j^k$ then

24:    apply $c_{k+1}$ to the fiber link

25:    $j\leftarrow j+1$

26:    $k^{\prime }\leftarrow k+1$

27:  else

28:    Turn steady state on (Stay on $c_k$)

29:  end if

30: end for