def executeSystematicSearchAlgorithm(options):
"""
Execute the specified algorthim and prints the result
"""
try:
num_vars, clauses = datautil.parseCNF(options.file)
comments = ''
res = None
if DAVIS_PUTNAM == options.algorithm.lower():
comments += 'Using DP algorithm (The orignal DP not DPLL)\n'
comments += 'The DP algorithm do not give a prove of satisfiablity\n'
comments += 'If the formula is unsatisfiable, the algorithm gives ' \
'you a core'
res = dp.solve(num_vars, clauses,
var_selection_heuristics[options.vselection])
elif DPLL == options.algorithm.lower():
comments += 'Using DPLL algorithm\n'
res = dpll.solve(num_vars, clauses,
var_selection_heuristics[options.vselection])
printComments(comments)
print formatSystematicSearchResult(res)
except Exception, e:
traceback.print_exc()
print '%s: %s' % (e.__class__.__name__, str(e))
def executeLocalSearchAlgorithm(options):
"""
Execute the specified algorithm and prints the result
"""
try:
num_vars, clauses = datautil.parseCNF(options.file)
comments = ''
# Chose and run algorithm
res = None
if GSAT == options.algorithm.lower():
if options.weighted:
comments += 'Solved With: Weighted GSAT'
res = wgsat.solve(num_vars, clauses, len(clauses)//2)
else:
comments += 'Solved With: GSAT'
res = gsat.solve(num_vars, clauses, len(clauses)//2)
elif GWSAT == options.algorithm.lower():
if options.weighted:
comments += 'Solved With: Weighted GWSAT'
res = wgwsat.solve(num_vars, clauses, len(clauses)//2, 0.4)
else:
comments += 'Solved With: GWSAT'
res = gwsat.solve(num_vars, clauses, len(clauses)//2, 0.35)
else:
raise Exception('Unspecified algorithm')
# If the formula it is not satisfiable this lines are never executed
del res[0]
printComments(comments)
print formatLocalSearchResult(res)
except Exception, e:
traceback.print_exc()
print '%s: %s' % (e.__class__.__name__, str(e))
def main(options):
global state_list
state_list = []
n_restarts = 30
for restart in range(n_restarts):
np.random.seed(restart)
global replay_buf
global q_l_agent
global epsilon
replay_buf = ReplayBuf(30000, 13, n_actions=4)
q_l_agent = Estimator(replay_buf)
run_stats = RunStats()
n_episodes = 100
epsilon = 1
for i in range(n_episodes):
epsilon = epsilon * 0.97
q_l_agent.train(discount_factor=0.999, replay_buf=replay_buf)
num_vars, clauses = datautil.parseCNF(options.file)
res = None
res = dpll.solve(num_vars, clauses, automatic_heuristic, run_stats)
print("Ep {} done in {} splits".format(i, run_stats.n_splits))
replay_buf.game_over()
run_stats.finish_episode()
np.save("run_stats/run_stats" + str(restart),
np.asarray(run_stats.episode_stats),
allow_pickle=True,
fix_imports=True)
