Algorithm 1. Cross-entropy method... | Emerald Publishing

Algorithm 1.

Cross-entropy method applied to noisy optimisation

Input: cross-entropy sampling density P(·\|λ); initial sampling distribution parameter λ; number of CE samples per iteration n_CE; number of elite samples n_E; number of sample settlement trajectories n_MC; maximum number of iterations n_max.
1	$l \leftarrow 0$ ⁠;
2	while l < n_max do
3		for m ← 1 to n_CE do
4			generate random heuristic parameter values w^(m) from sampling density P(·\|λ*);
5			generate n_MC settlement trajectories and measurement following strategy $S (w^{(m)})$ ⁠;
6			evaluate the expected total life-cycle cost q_m with n_MC samples (equation (14));
7		end
8		sort (w⁽¹⁾, …, w^(n_CE⁾) in increasing order of q_m;
9		fit the distribution parameter λ* to the n_E elite samples;
10		l← l+1;
11	end
12	w← mean of P(·\|λ);
13	return w*

Input: cross-entropy sampling density P(·\|λ); initial sampling distribution parameter λ; number of CE samples per iteration n_CE; number of elite samples n_E; number of sample settlement trajectories n_MC; maximum number of iterations n_max.
1	$l \leftarrow 0$ ;
2	while l < n_max do
3		for m ← 1 to n_CE do
4			generate random heuristic parameter values w^(m) from sampling density P(·\|λ*);
5			generate n_MC settlement trajectories and measurement following strategy $S (w^{(m)})$ ;
6			evaluate the expected total life-cycle cost q_m with n_MC samples (equation (14));
7		end
8		sort (w⁽¹⁾, …, w^(n_CE⁾) in increasing order of q_m;
9		fit the distribution parameter λ* to the n_E elite samples;
10		l← l+1;
11	end
12	w← mean of P(·\|λ);
13	return w*