def greedysearch(bn, scr, iters, cstr=None):
gr_better = operator.__gt__
gr_accept = lambda n,sss : n>sss["curr_score"]
gr_step = bnsearch.stepper_f(bnsearch.acts, gr_better, gr_accept, cstr)
gr_stop = bnsearch.stopper_f(iters)
gr_isss = bnsearch.initial_search_status(bn,scr)
return bnsearch.localsearch(gr_isss, gr_step, gr_stop)
def find_initial_temperature(bn, sc):
def it_accept(n, sss):
d = n - sss["curr_score"]
if d < 0:
sss["dsum"] += d
sss["k"] += 1
return True
it_step = bnsearch.stepper_f(bnsearch.acts, operator.__gt__, it_accept)
it_stop = bnsearch.stopper_f(5000)
isss = {"k": 0, "dsum": 0.0}
isss.update(bnsearch.initial_search_status(bn, sc))
bnsearch.localsearch(isss, it_step, it_stop)
return isss["dsum"] / isss["k"] / math.log(0.8)
isss.update(bnsearch.initial_search_status(bn, sc))
bnsearch.localsearch(isss, it_step, it_stop)
return isss["dsum"] / isss["k"] / math.log(0.8)
if __name__ == '__main__':
import sys
def sa_accept(n, sss):
c = sss["curr_score"]
wp = math.exp((n - c) / sss["t"])
apt = n > c or random.uniform(0, 1) < wp
return apt
sa_step = bnsearch.stepper_f(bnsearch.acts, operator.__gt__, sa_accept)
def sa_stop(sss):
endtime = sss["endtime"]
currtime = time.time()
lefttime = endtime - currtime
sss["t"] = sss["t0"] * lefttime / searchtime # update temperature
return currtime >= endtime
bn, sc = bnsearch.sys2bnscr()
searchtime = int(sys.argv[4])
endtime = time.time() + searchtime
t0 = find_initial_temperature(bn, sc)
isss = {"searchtime": searchtime, "endtime": endtime, "t0": t0, "t": t0}
isss.update(bnsearch.initial_search_status(bn, sc))