def generate_report(l, t_start, t_end):
def residual(x, y):
c = []
for j in xrange(len(x)):
cur_c = []
for i in xrange(len(x[j])):
cur_x = x[j][i]
cur_y = y[j][i]
cur_c.append(math.fabs(g([cur_x, cur_y]) -
resulting_nn([cur_x, cur_y])))
c.append(cur_c)
return c
folder = join(report_dir, str(time()))
mkdir(folder)
resulting_nn = network_from_list(l)
sqrt_err = make_quad_estimate(resulting_nn, g, 0, 0, 1, 1, 0.01, 0.01)
file = open(join(folder, "Report.txt"), "w")
file.write((u"""Solution is complete
%s neurons
Complex Box
f = x + y
%s point in area
%s points on border
delta = %s
J = %s
sqrt_estimate = %s
Time taken: %s
""" % (neurons, len(inner), len(border), delta,
J(f, g, inner, border,resulting_nn),
sqrt_err, t_end - t_start)).encode("utf-8"))
file.write("(")
for i in xrange(neurons):
ns = 2 + dimensions
file.write("(%s)*exp(-(((%s)-x)^2+((%s)-y)^2)/((%s)^2))+" % (l[ns * i],
l[ns * i + 2], l[ns * i + 3], l[ns * i + 1]))
file.write("0)\n")
file.close
neuro.draw.draw(resulting_nn, (-3.0, -3.0, 7.0, 7.0), (0.0, 0.0, 1.0, 1.0),
join(folder, "cfg.png"))
dx, dy = 0.005, 0.005
ax = arange(0, 1.0001, dx)
ay = arange(0.0, 1.0001, dy)
X, Y = meshgrid(ax, ay)
Z = residual(X, Y)
figure(5)
pcolor(X, Y, Z)
colorbar()
axis([0,1,0,1])
savefig(join(folder, "overall.png"))
for c_y in xrange(3):
y = c_y * 0.5
xs = [0.01*x for x in range(100)]
ys = [resulting_nn([x, y]) for x in xs]
zs = [g([x, y]) for x in xs]
figure(c_y)
xlabel('$x$')
ylabel('$y$')
title(ur'Погрешность полученного решения в сечении $y = %.1f$' % y)
plot(xs, ys, label=ur'полученное решение', color='red')
plot(xs, zs, label=ur'точное решение', color='green')
legend(loc='upper right')
savefig(join(folder, "сечение y=%s.png" % y))
for c_x in xrange(3):
x = c_x * 0.5
ys = [0.01*y for y in range(100)]
xs = [resulting_nn([x, y]) for y in ys]
zs = [g([x, y]) for y in ys]
figure(c_x + 100)
xlabel('$x$')
ylabel('$y$')
title(ur'Погрешность полученного решения в сечении $x = %.1f$' % x)
plot(ys, xs, label=ur'полученное решение', color='red')
plot(ys, zs, label=ur'точное решение', color='green')
legend(loc='upper right')
savefig(join(folder, "сечение x=%s.png" % x))
border = ([[0.0, random()] for i in range(5)] +
[[1.0, random()] for i in range(5)] +
[[random(), 0.0] for i in range(5)] +
[[random(), 1.0] for i in range(5)])
def to_minimize(l):
nn = network_from_list(l)
return J(f, g, inner, border, nn)
t_start = time()
l = nelder_mead(to_minimize, neurons * (2 + dimensions))
t_end = time()
resulting_nn = network_from_list(l)
sqrt_err = make_quad_estimate(resulting_nn, g, 0, 0, 1, 1, 0.01, 0.01)
print u"""Solution is complete
%s neurons
Nelder-Mead e-10
f = x + y
%s point in area
%s points on border
delta = %s
J = %s
sqrt_estimate = %s
Time taken: %s""" % (neurons, len(inner), len(border), delta,
J(f, g, inner, border,resulting_nn),
sqrt_err, t_end - t_start)
print "(" ,
