def test2Neuro(self):
return True
def signal(point):
x, y = point
return 20.0 * math.exp(-(math.pow(0.5 - x, 2.0) + math.pow(0.82 - y, 2.0))/(0.5 ** 2)) + \
25.0 * math.exp(-(math.pow(0.34 - x, 2.0) + math.pow(0.16 - y, 2.0))/(0.35 ** 2))
points = [[random.random(), random.random()] for _ in range(300)]
def residual(l):
NN = network_from_list(l)
return sum(map(lambda x: math.fabs(signal(x) - NN(x)), points))
res = complex_box(residual, 8, [0, 0.1, 0.0, 0.0] * 2, [100, 1.0, 1.0, 1.0] * 2)
print res
def test3Neuro(self):
def signal(point):
x, y = point
return 43.0 * math.exp(-(math.pow(0.5 - x, 2.0) + math.pow(0.82 - y, 2.0))/(0.5 ** 2)) + \
20.0 * math.exp(-(math.pow(0.34 - x, 2.0) + math.pow(0.16 - y, 2.0))/(0.35 ** 2)) + \
14.0 * math.exp(-(math.pow(0.2 - x, 2.0) + math.pow(0.7 - y, 2.0))/(0.31 ** 2))
points = [[random.random(), random.random()] for _ in range(100)]
def residual(l):
NN = network_from_list(l)
return sum(map(lambda x: math.fabs(signal(x) - NN(x)), points))
res = complex_box(residual, 12, [0, 0.1, 0.0, 0.0] * 3, [60, 0.7, 1.0, 1.0] * 3)
print res
def test1Neuro(self):
return True
def signal(point):
x, y = point
return 43.0 * math.exp(-(math.pow(0.5 - x, 2.0) + math.pow(0.82 - y, 2.0))/(0.5 ** 2))
points = [[random.random(), random.random()] for _ in range(100)]
def residual(l):
NN = network_from_list(l)
return sum(map(lambda x: math.fabs(signal(x) - NN(x)), points))
res = complex_box(residual, 4, [0, 0.1, 0.0, 0.0], [100, 1.0, 1.0, 1.0])
diffs = map(lambda x, y: math.fabs(x - y), [43.0, 0.5, 0.5, 0.82], res)
print diffs
self.assertTrue(all(map(lambda x: x < 1e-4, diffs)))
# -*- coding: utf-8 -*-
from time import time
from __init__ import to_minimize, neurons, dimensions, J_grad_wrap
from complex_box import complex_box
from first_order_methods.simulated_annealing import simulated_annealing
from report import generate_report
t_start = time()
l = complex_box(to_minimize, neurons * (2 + dimensions),
[-250.0, 1.0, -2.0, -2.0] * neurons,
[250.0, 10.0, 6.0, 6.0] * neurons)
l = simulated_annealing(l, J_grad_wrap, to_minimize)
t_end = time()
generate_report(l, t_start, t_end)
print u"Готово"
