nav_comparepolicies.m

来自「approximate reinforcement learning」· M 代码 · 共 23 行

eps0 = 1e-7;

load navdata_ct_parallelconv
ptheta = qistats.theta{find(qistats.delta < eps0, 1, 'first')};

find(qistats.delta < eps0, 1, 'first')

load navdata_ct_serialconv
stheta = qistats.theta{find(qistats.delta < eps0, 1, 'first')};
find(qistats.delta < eps0, 1, 'first')

[pQ, ph] = max(ptheta, [], 1);
[sQ, sh] = max(stheta, [], 1);

hdiff = find(ph ~= sh);

if ~isempty(hdiff),
    disp('Policies different at indices:');
    hdiff
else
    disp('Policies identical');
end;