def __init__(self,
filename,
eps,
point_reader,
dist,
edge_selector,
graph=None,
lower=None,
upper=None):
self.filename = filename
self.path = filename
self.epsilon = eps
self.selector = edge_selector
self.d = dist
if filename.endswith('.pkl'):
file = open(self.path, 'rb')
self.points = pickle.load(file)
else:
self.points = point_reader(filename)
printer("starting approx with eps=" + str(self.epsilon))
self.init_graph()
if graph is not None and lower is not None and upper is not None:
self.lower = lower
self.upper = upper
self.G = graph
self.compute_spanner()
self.build_matrix()
printer("ending approx with eps=" + str(self.epsilon))
def build_matrix(self):
n = len(self.points)
matrix = np.zeros(shape=(n, n)) # initialize matrix
for i in range(n):
printer("Progress: " + str(i / n))
for j in range(i, n):
dist = self.d(self.points[i], self.points[j])
# printer("finished frechet dist computation for ", sep="")
matrix[i][j] = dist
matrix[j][i] = dist
print(str(j), end=", ", flush=True)
print()
self.matrix = matrix
def hausdorff_cloud_r(file):
points = []
file.readline()
for line in file:
cluster = []
line = line.rstrip().split(";")
for pt in line:
if pt == '':
continue
pt = pt.split(",")
pt = [float(p) for p in pt]
pt = np.array(pt)
cluster.append(pt)
points.append(np.array(cluster))
printer("Done loading points")
return points
def __init__(self, filename, point_reader, dist):
self.filename = filename
self.path = filename
self.d = dist
if filename.endswith('.pkl'):
file = open(self.path, 'rb')
self.points = pickle.load(file)
else:
self.points = point_reader(
filename
) # create list of points using point_reader (assume point_reader returns list of points)
printer("starting complete")
self.build_matrix() # compute n choose 2 different distances
printer("ending complete")
def compute_spanner(self):
indx = self.selector(self.lower, self.upper, self.epsilon)
iter = 0
while (indx[0] != -1):
if iter % 100 == 0:
printer("Iter: " + str(iter))
iter += 1
i = indx[0]
j = indx[1]
s = time.time()
dist = self.d(self.points[i], self.points[j])
# print("Dist: " + str(time.time() - s))
s = time.time()
self.G.add_edge(i, j, weight=dist)
# print("Add Edge: " + str(time.time() - s))
s = time.time()
self.update_bounds(dist, i, j)
# print("Update Bounds: " + str(time.time() - s))
s = time.time()
indx = self.selector(self.lower, self.upper, self.epsilon)
def shape_r(path, subdir='/'):
points = []
path = path
dirs = next(os.walk(path))[1]
printer(str(dirs))
for dir in dirs:
subpath = path + dir + subdir
files = next(os.walk(subpath))[2]
printer(str(dir))
for f in files:
file_printer("LABEL: {:} FILE: {:}".format(str(dir), str(f)), prio=0)
points.append(read_model(subpath + f))
printer_class.PRIORITY = 1
printer("Done loading points")
return points
for i, number in enumerate(numbers):
if number > numbers[highest_index]:
highest_index = i
return highest_index
# Images are 28x28 pixels. Reshapen, they are 784x1
training = [(mux(label), image.reshape(784) / 255.0)
for label, image in mnist.read(path=path, dataset="training")]
testing = [(mux(label), image.reshape(784) / 255.0)
for label, image in mnist.read(path=path, dataset="testing")]
nnet = neuralnetwork.NeuralNetwork([784, 30, 10])
ITR = 30
BTC = 3
p = printer.printer(1)
for i in range(ITR):
error = 0
for t, x in training:
for j in range(BTC):
error += nnet.learn(x, t, 1)
message = 'Iteration ' + str(i) + ', error ' + str(error)
p.overwrite(message)
p.clear()
success = 0
for t, x in training:
y = demux(nnet.evaluate(x))
if y == demux(t):
success += 1