forked from conservation-decisions/K-MDP
-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathexperiments_fisheries.m
207 lines (144 loc) · 5.53 KB
/
experiments_fisheries.m
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
%Add dependencies
addpath('kmeans')
addpath('utils')
addpath('algorithms')
addpath('MDP_graph_representation')
%precision
p = 0.00000000000001;
%Discount factor
discount = 0.96;
%Choose the MDP model from the ones available on problems/fisheries.
NS = 1001;
NA = 11;
P_mat = strcat("problems/fisheries/fisheries_P_",num2str(NS),"_",num2str(NA),".mat");
R_mat = strcat("problems/fisheries/fisheries_R_",num2str(NS),"_",num2str(NA),".mat");
load(P_mat);
load(R_mat);
%Solve the MDP
tic;
%Perform value iteration
[Pol]=mdp_value_iteration(P, R, discount);%Get a policy using value iteration
%Evaluate the policy
[V,Q]= mdp_eval_policy_iterative_q(P, R, discount, Pol); % check that this is doing what it is supposed to; Get the value and Q values
Q(isnan(Q))=0;
time_MDP = toc;
%K
K = [15:10:NS];
K = flip(K);
K = [NS, K, 13];
K = 13;
%Initialize structures for statistics extraction
err_transitive = zeros(length(K),1);
t_transitive = zeros(length(K),1);
t_KMDP_transitive = zeros(length(K),1);
err_astar = zeros(length(K),1);
t_astar = zeros(length(K),1);
t_KMDP_astar = zeros(length(K),1);
err_kmeans = zeros(length(K),1);
t_kmeans = zeros(length(K),1);
t_KMDP_kmeans = zeros(length(K),1);
S2K_Qd = cell(length(K),1);
K2S_Qd = cell(length(K),1);
S2K_astar = cell(length(K),1);
K2S_astar = cell(length(K),1);
S2K_kmeans = cell(length(K),1);
K2S_kmeans = cell(length(K),1);
for i = 1:length(K)
k = K(i);
fprintf('Qd\n');
[PK_Qd, RK_Qd, S2K_Qd{i}, K2S_Qd{i}, PolicyK_Qd, PolKs_Qd, err_transitive(i), t_transitive(i), t_KMDP_transitive(i)] = QdKMDP(k, p, P, R, discount, Q, V);
fprintf('a*_d\n');
[PK_astar, RK_astar, S2K_astar{i}, K2S_astar{i}, PolicyK_astar, PolKs_astar, err_astar(i), t_astar(i), t_KMDP_astar(i)] = aStarKMDP(k, p, P, R, discount, V, Pol);
fprintf('K-means \n');
[PK_kmeans, RK_Kmeans, S2K_kmeans{i}, K2S_kmeans{i}, PolicyK_kmeans, PolKs_kmeans, err_kmeans(i), t_kmeans(i), t_KMDP_kmeans(i)] = kmeansKMDP(k, P, R, discount, Q, V);
end
%DATA to plot and extract
K
disp('Q_d')
err_transitive'
t_transitive'
t_KMDP_transitive'
disp('astar')
err_astar'
t_astar'
t_KMDP_astar'
disp('kmeans++')
err_kmeans
t_kmeans
t_KMDP_kmeans
%Store relevant data
save('problems/fisheries/results/S2K_Qd.mat', 'S2K_Qd');
save('problems/fisheries/results/K2S_Qd.mat', 'K2S_Qd');
save('problems/fisheries/results/S2K_astar.mat', 'S2K_astar');
save('problems/fisheries/results/K2S_astar.mat', 'K2S_astar');
save('problems/fisheries/results/S2K_kmeans.mat', 'S2K_kmeans');
save('problems/fisheries/results/K2S_kmeans_250_251.mat', 'K2S_kmeans');
err_transitive = err_transitive * 100;
err_astar = err_astar * 100;
err_kmeans = err_kmeans * 100;
%Plot releveant data
%Error
figure;
plot(K, err_transitive, 'r-o', 'LineWidth', 1)
hold on;
plot(K, err_astar, 'b-x', 'LineWidth',1)
hold on;
plot(K, err_kmeans, 'g-*', 'LineWidth',1)
hold off;
xlabel('K')
ylabel('gap(%)');
title('Rebuilding global fisheries');
legend('Q^*_d K-MDP', 'Q^*_a K-MDP', 'k-means++ K-MDP');
plot_name_fig = strcat('problems/fisheries/results/fisheries_gap.fig');
plot_name_png = strcat('problems/fisheries/results/fisheries_gap.png');
saveas(gcf, plot_name_fig);
saveas(gcf, plot_name_png);
%Time compute KDMP
figure;
plot(K, t_transitive, 'r-o', 'LineWidth', 1)
hold on;
plot(K, t_astar, 'b-x', 'LineWidth',1)
hold on;
plot(K, t_kmeans, 'g-*', 'LineWidth',1)
hold off;
xlabel('K')
ylabel('time(sec.)');
title('Sea Otter and Northern Abalone time to compute the K-MDP');
legend('Q^*_d K-MDP', 'Q^*_a K-MDP', 'k-means++ K-MDP');
plot_name_fig = strcat('problems/fisheries/results/time-compute.fig');
plot_name_png = strcat('problems/fisheries/results/time-compute.png');
saveas(gcf, plot_name_fig);
saveas(gcf, plot_name_png);
%Time to solve KDMP
figure;
plot(K, t_KMDP_transitive, 'r-o', 'LineWidth', 1)
hold on;
plot(K, t_KMDP_astar, 'b-x', 'LineWidth',1)
hold on;
plot(K, t_KMDP_kmeans, 'b-x', 'LineWidth',1)
hold off;
xlabel('K')
ylabel('time(sec.)');
title('Sea Otter and Northern Abalone time to solve the K-MDP');
legend('Q^*_d K-MDP', 'Q^*_a K-MDP', 'k-means++ K-MDP');
plot_name_fig = strcat('problems/fisheries/results/time-solve.fig');
plot_name_png = strcat('problems/fisheries/results/time-solve.png');
saveas(gcf, plot_name_fig);
saveas(gcf, plot_name_png);
%Plot reduced model and policy graphs.
actions = strings(size(P,3), 1);
actions(1) = "0_H";
actions(2) = "100_H";
actions(3) = "200_H";
actions(4) = "300_H";
actions(5) = "400_H";
actions(6) = "500_H";
actions(7) = "600_H";
actions(8) = "700_H";
actions(9) = "800_H";
actions(10) = "900_H";
actions(11) = "1000_H";
min_prob = 0.001;
%TODO: Check why is crashing here
%[G_reduced_model, p_reduced_model] = model_representation(PK_astar, RK_astar, actions, min_prob);
%[G_reduced_policy, p_reduced_policy] = policy_representation(PK_astar, RK_astar, PolicyK_astar, actions, 0.08);