def get_current_scaled_txouts(scale, trunc=0):
txouts = node.get_current_txouts()
total = sum(txouts.itervalues())
results = dict((script, value*scale//total) for script, value in txouts.iteritems())
if trunc > 0:
total_random = 0
random_set = set()
for s in sorted(results, key=results.__getitem__):
if results[s] >= trunc:
break
total_random += results[s]
random_set.add(s)
if total_random:
winner = math.weighted_choice((script, results[script]) for script in random_set)
for script in random_set:
del results[script]
results[winner] = total_random
if sum(results.itervalues()) < int(scale):
results[math.weighted_choice(results.iteritems())] += int(scale) - sum(results.itervalues())
return results
def act(self, state, actions) :
choice = weighted_choice([1.0-self.eps, self.eps])
if choice == 0 : return actions[randint(0, len(actions)-1)]
return self.argmax(state, actions)