wolffd@0: function [policy, EU] = upot_to_opt_policy(pot)
wolffd@0: % UPOT_TO_OPT_POLICY Compute an optimal deterministic policy given a utility potential
wolffd@0: % [policy, EU] = upot_to_opt_policy(pot)
wolffd@0: %
wolffd@0: % policy(a,b, ..., z) = P(do z | a, b, ..), which will be a delta function
wolffd@0: % EU is the contraction of this potential, i.e., P .* U
wolffd@0: 
wolffd@0: sz = pot.sizes; % mysize(pot.p);
wolffd@0: if isempty(sz)
wolffd@0:   EU = pot.u;
wolffd@0:   policy = [];
wolffd@0:   return;
wolffd@0: end
wolffd@0: 
wolffd@0: parent_size = prod(sz(1:end-1));
wolffd@0: self_size = sz(end); 
wolffd@0: C = pot.p .* pot.u; % contraction
wolffd@0: C = reshape(C, parent_size, self_size);
wolffd@0: policy = zeros(parent_size, self_size);
wolffd@0: for i=1:parent_size
wolffd@0:   act = argmax(C(i,:));
wolffd@0:   policy(i, act) = 1;
wolffd@0: end
wolffd@0: policy = myreshape(policy, sz);
wolffd@0: EU = sum(C(:));