void operator() (const parallel::range_t& range) const {
assert(n.size() == pop.size());
assert(s.size() == pop.size());
for (size_t p = range.begin(); p != range.end(); ++p) {
assert(s[p].empty());
n[p] = 0;
for (size_t q = 0; q < pop.size(); ++q) {
int flag = fit::dominate_flag(pop[p], pop[q]);
if (flag > 0)
s[p].push_back(q);
else if (flag < 0)
++n[p];
}
if (n[p] == 0)
r[p] = 0;
}
}
void operator() (const parallel::range_t& r) const
{
for (size_t i = r.begin(); i != r.end(); ++i)
{
float d = 0.0f;
if (_pop[i]->fit().objs()[0] <= -10000)
d=-1000;
else
{
for (size_t j = 0; j < _pop.size(); ++j)
d += euclidean_dist(_pop[i]->data(),_pop[j]->data());
d /= (float)_pop.size();
}
int l = _pop[i]->fit().objs().size()-1;
_pop[i]->fit().set_obj(l, d);
}
}
void operator() (const parallel::range_t& r) const {
for (size_t i = r.begin(); i != r.end(); ++i) {
for (size_t j = 0; j < _archive.size(); ++j)
{
float d = 0;
// Check if they are not the same image
if ((_pop[i] != _archive[j]) && (_pop[i]->id() != _archive[j]->id()))
{
#ifdef EUCLIDEAN_DISTANCE
// Euclidean distance
d = _pop[i]->dist(*_archive[j]);
#else
// Regular distance
d = _pop[i]->fit().dist(*_archive[j]);
#endif
}
distances(i, j) = d;
}
}
}
void operator() (const parallel::range_t& r) const {
for (size_t i = r.begin(); i != r.end(); ++i)
_pop[i]->mutate();
}
void operator() (const parallel::range_t& r) const {
for (size_t i = r.begin(); i != r.end(); ++i) {
_pop[i] = boost::shared_ptr<Phen>(new Phen());
_pop[i]->random();
}
}