def knn_classify(data_file, K=5):
# f_train, f_test, l_train, l_test = get_data(data_file)
f_train, f_test, l_train, l_test = get_data_optimized(data_file)
l_predict = []
for t in range(f_test.shape[0]):
N = []
l = 0
for d in range(f_train.shape[0]):
if len(N) <= K:
N.append(d)
else:
for u in N:
if dist(f_test[t], f_train[d]) < dist(f_test[t], f_train[u]):
N.remove(u)
N.append(d)
break
for n in N:
if l_train[n] == 'A':
l += 1
else:
l -= 1
if l >= 0:
l_predict.append('A')
else:
l_predict.append('B')
acc = (float(sum(l_predict == l_test))/len(l_predict))
plot(f_train, l_train, f_test, l_test, l_predict, acc)
def get_data_optimized(data_file, test_size=0.9):
data = pd.read_csv(data_file)
data = np.array(data)
train_data = []
seed_point = random.randint(0, data.shape[0] - 1)
train_data.append(data[seed_point])
test_data = np.delete(data, seed_point, axis=0)
while len(train_data) <= (int((1 - test_size) * data.shape[0])):
dist_buf = []
for i in range(test_data.shape[0]):
min_dist = float('inf')
for j in range(len(train_data)):
d = dist(train_data[j], test_data[i])
if d < min_dist:
min_dist = d
dist_buf.append(min_dist)
max_point = dist_buf.index(max(dist_buf))
train_data.append(test_data[max_point])
test_data = np.delete(test_data, max_point, axis=0)
train_data = np.array(train_data)
f_train = train_data[:, :-1]
l_train = train_data[:, -1]
f_test = test_data[:, :-1]
l_test = test_data[:, -1]
return f_train, f_test, l_train, l_test
def eval_routing_table(self) : sensor_range = sqrt(5) * self.yard.grid_size # evaluating all the alive neighbours in the communication range for node in self.yard.nodes : self.neighbours[node.id] = [] for adj_node in self.yard.nodes : if adj_node.id != node.id and adj_node.energy > 0 and dist(node.x, node.y, adj_node.x, adj_node.y) <= sensor_range : self.neighbours[node.id].append(adj_node)
def send_data_non_ch(self, energy, packet): d0 = sqrt(energy.free_space / energy.multi_path) # distance of sensor to its CH d = dist(self.x, self.y, self.cell.head.x, self.cell.head.y) if d > d0 : self.energy = self.energy - (packet.ctr_packet_length*energy.trans + energy.multi_path*packet.packet_length*(d ** 4)) else : self.energy = self.energy - (packet.ctr_packet_length*energy.trans + energy.free_space*packet.packet_length*(d ** 2))
def sense(self, sensors, panda_x, panda_y): sensor_range = sqrt(5)*self.yard.grid_size for sensor_node in self.yard.nodes: if (sensor_node.energy > 0 and dist(sensor_node.x, sensor_node.y, panda_x, panda_y) <= sensor_range ): sensors.append(sensor_node) print "len", len(sensors) return len(sensors) > 0
def do_compute_path(r, cx, cy, tx, ty, n, stack, visited, escape_center,
escape):
if n > 700:
return False
# print "A"
stack.append((cx, cy))
# print "AA", len(r), cy
# print "AA", len(r[cy]), cx
# if cy < r[cy][cx] == 'T':
# print "AAA"
# return True
# print "B"
candidates = []
for y in [-1, 0, 1]:
for x in [-1, 0, 1]:
if x != 0 or y != 0:
if big_enough(r, cx + x, cy + y, escape_center, escape):
candidates.append((dist(cx + x, cy + y, tx,
ty), cx + x, cy + y))
# print "C"
for d, x, y in sorted(candidates):
if (x, y) not in visited:
visited.add((x, y))
if r[y][x] == 'T':
stack.append((x, y))
return True
if do_compute_path(r, x, y, tx, ty, n + 1, stack, visited,
escape_center, escape):
return True
# print "D"
stack.pop()
# print "E"
return False
def do_compute_path(r, cx, cy, tx, ty, n, stack, visited, escape_center, escape):
# stop if recursion depth exceeds 700
if n > 700:
return False
stack.append((cx, cy))
candidates = []
for y in [-1, 0, 1]:
for x in [-1, 0, 1]:
if x != 0 or y != 0:
if big_enough(r, cx + x, cy + y, escape_center, escape):
candidates.append((dist(cx + x, cy + y, tx, ty), cx + x, cy + y))
for d, x, y in sorted(candidates):
if (x, y) not in visited:
visited.add((x, y))
if r[y][x] == 'T':
stack.append((x, y))
return True
if do_compute_path(r, x, y, tx, ty, n + 1, stack, visited, escape_center, escape):
return True
stack.pop()
return False
def execute(self, rings, event_rings, packet) : boundary_nodes = [] outer_ring = len(rings) - 1 # identifying the outermost nodes for counter in rings : if len(rings[outer_ring]) > 0 : for node in rings[outer_ring] : boundary_nodes.append(node.head) break else : outer_ring -= 1 print len(boundary_nodes),"boundary_nodes" starting_node = None # randomly picking up the outermost node to initiate the protocol while starting_node == None : node_id = random.randint(0,len(boundary_nodes) - 1) if boundary_nodes[node_id].energy > 0 : starting_node = boundary_nodes[node_id] print "Starting Greedy Node : ",starting_node interference_node = None maxd = dist(0, 0, self.yard.l, self.yard.b) # executing the protocol until the greedy path intersects the cyclic path while interference_node == None : min_dist = maxd nearest_node = None if len(self.neighbours[starting_node.id]) > 0 : # energy dissipitated while transmiting starting_node.send_data_non_ch(self.yard.energy, packet) print "transmiting",starting_node # identifying which alive neighbour in the communication range is closest to sink for node in self.neighbours[starting_node.id] : if node.energy > 0 : distance_sink = dist(node.x, node.y, self.yard.sink.x, self.yard.sink.y) if min_dist < distance_sink : min_dist = distance_sink nearest_node = node if node.cell.head.ring in event_rings : interference_node = node nearest_node = interference_node print "ring number", node.cell.head.ring print "interference_node",interference_node break if nearest_node is not None : # energy dissipitated while receiving nearest_node.receive_data(self.yard.energy, packet) print "receiving",nearest_node.id starting_node = nearest_node else : # all nodes in the communication range are Dead break
def execute(self, panda_x, panda_y):
packet = Packet()
self.eval_rings()
sensors = []
fig = plt.figure()
plt.ion()
plt.show()
# plt.legend()
plt_node_ch = fig.add_subplot(211)
plt_energy = fig.add_subplot(212)
# initialising the GreedyRouting Protocol
greedyRouting = GreedyRouting(self.yard)
while self.sense(sensors, panda_x, panda_y):
self.plot_nodes_ch(plt_node_ch, panda_x, panda_y)
self.plot_energy(plt_energy)
plt.pause(1e-50)
# time.sleep(0.1)
CyclicRouting.iteration += 1
# constructing the greedy routing table
greedyRouting.eval_routing_table()
cluster_heads = []
for node in sensors:
# A cluster head maintains cache, therefore a cluster would trigger the ring only once
if node.cell.head not in cluster_heads:
cluster_heads.append(node.cell.head)
if node.is_cluster_head:
continue
node.send_data_non_ch(self.yard.energy, packet)
rings = []
for head in cluster_heads:
# Panda detected from different cluster, can belong to same ring or other
rings.append(head.ring)
# executing the Greedy Routing
greedyRouting.execute(self.rings, rings, packet)
for ring in rings:
for cell in self.rings[ring]:
maxd = dist(0, 0, self.yard.l, self.yard.b)
d_nodes = {
1: [maxd, None], # clockwise
-1: [maxd, None] # anti-clockwise
}
for node in self.rings[ring] :
try:
val = ((node.head.x - self.yard.sink.x) / (self.yard.sink.x - cell.head.x)
- (node.head.y - self.yard.sink.y) / (self.yard.sink.y - cell.head.y))
except:
continue
if val is 0:
val = 1
sign = val/abs(val)
d = dist(node.head.x, node.head.y, cell.head.x, cell.head.y)
if d < d_nodes[sign][0]:
d_nodes[sign][0] = d
d_nodes[sign][1] = node
if d_nodes[1][1] is not None:
cell.head.send_data_ch(self.yard.energy, packet, d_nodes[1][0])
cell.head.receive_data(self.yard.energy, packet)
if d_nodes[-1][1] is not None:
cell.head.send_data_ch(self.yard.energy, packet, d_nodes[-1][0])
cell.head.receive_data(self.yard.energy, packet)
#only one node remains in the ring || should be thought of
if d_nodes[1][1] is None and d_nodes[-1][1] is None:
cell.head.energy = 0
self.yard.clusterize()
self.eval_rings()
sensors = []
print "iterations done: ", CyclicRouting.iteration
plt.show()
plt.pause(10)
def fitness(self):
if self.atgoal:
return 2 + 1 / (self.brain.step**2)
else:
return 1 / (dist(self._pos, self._goal.pos)**2)
def _ingoal(self):
return dist(self._pos, self._goal.pos) < 10
cv2.CHAIN_APPROX_SIMPLE)
#get rid of outliers that are far away
centLst = []
avgArea = 0
for cont in contours:
M = cv2.moments(cont)
cX = int(M["m10"] / (M["m00"] + 0.00000001))
cY = int(M["m01"] / (M["m00"] + 0.00000001))
avgArea += cv2.contourArea(cont)
centLst.append((cX, cY))
combs = combinations(centLst, 2)
avgDist = 0
tot = 0
for combo in combs:
avgDist += h.dist(combo[0], combo[1])
tot += 1
avgDist /= tot + 0.00000000000001
avgArea /= tot + 0.00000000000001
contImg = img.copy()
for i in range(0, len(contours) - 1):
if avgDist * avgArea > cv2.contourArea(contours[i]) * h.dist(
centLst[i], h.shortest_dist(centLst[i], centLst)):
contImg = cv2.drawContours(contImg, contours, i, (0, 0, 255))
else:
contImg = cv2.drawContours(contImg, contours, i, (0, 255, 0))
cv2.imshow('thresh', contImg)
key = cv2.waitKey(1) & 0xFF
if key == ord('q'):