First commit

network/importance_sampling.py

@@ -0,0 +1,92 @@
+from random import Random
+from network.network import create_network, get_number_of_activated_antenna
+
+
+class AntennaGenerator(object):
+
+    def __init__(self, box_size, random):
+        self.box_size = box_size
+        self._antennas = list(range(box_size * box_size))
+        self.random = random
+
+    def get_random_antenna(self):
+        idx, prob = self.get_random_index_and_its_probability()
+        if idx == -1:
+            self._antennas = []
+            return -1, prob
+        choosen_antenna = self._antennas[idx]
+        self._remove_antenna(idx)
+        return choosen_antenna, prob
+
+
+class SuccesiveAntennaGenerator(AntennaGenerator):
+    def _remove_antenna(self, idx):
+        ant = self._antennas[idx]
+        if ant == -1:
+            self._antennas = []
+            return ant
+        if (ant + 1) % self.box_size:
+            self._antennas = self._antennas[idx + 2:]
+        else:
+            self._antennas = self._antennas[idx + 1:]
+
+        if ant + self.box_size < self.box_size * self.box_size:
+            self._antennas.remove(ant + self.box_size)
+
+    def get_random_index_and_its_probability(self):
+        idx = self.random.randint(-1, len(self._antennas)-1)
+        prob = (len(self._antennas) + 1) ** -1
+        return idx, prob
+
+
+class BadAntennaGenerator(SuccesiveAntennaGenerator):
+    '''this antenna generator does not return 'real' probabilities
+    and thus cannot be used in importance_sampling'''
+    def _remove_antenna(self, idx):
+        ant = self._antennas[idx]
+        to_remove = {ant, ant + self.box_size, ant - self.box_size}
+        if (ant + 1) % self.box_size:
+            to_remove.add(ant + 1)
+
+        if ant % self.box_size:
+            to_remove.add(ant - 1)
+        self._antennas = [a for a in self._antennas if a not in to_remove]
+
+
+class BetterSuccesiveAntennaGenerator(SuccesiveAntennaGenerator):
+
+    def _get_random_index(self, max_number_inclusive):
+        if max_number_inclusive < 1:
+            return self.random.randint(-1, max_number_inclusive)
+        mean_ = max_number_inclusive/2
+        x = int(self.random.expovariate(mean_**-1))
+        if x > max_number_inclusive:
+            return -1
+        return x
+
+    def get_random_index_and_its_probability(self):
+        idx = self._get_random_index(len(self._antennas)-1)
+        prob = (len(self._antennas) + 1) ** -1
+        return idx, prob
+
+
+def get_random_network(box_size, random, Generator):
+    antenna_generator = Generator(box_size, random)
+    antennas = []
+    antenna, network_prob = antenna_generator.get_random_antenna()
+
+    while antenna != -1:
+        antennas.append(antenna)
+        antenna, p = antenna_generator.get_random_antenna()
+        network_prob *= p
+
+    return create_network(antennas), network_prob
+
+
+def estimate_number_of_activated_antenna(n, box_size, random, Generator):
+    ret, sum_of_probs = 0, 0
+    for i in range(n):
+        network, p = get_random_network(box_size, random, Generator)
+        ret += get_number_of_activated_antenna(network) / p
+        sum_of_probs += p**-1
+    return ret / sum_of_probs