apunkt/GPy
1
0
Fork 0
mirror of https://github.com/SheffieldML/GPy.git synced 2026-05-03 08:42:39 +02:00
Code Issues Projects Releases Packages Wiki Activity Actions

simulation data changes

Browse source
This commit is contained in:
Max Zwiessele 2013-04-16 12:37:31 +01:00
parent 350497c726
commit aaf51f0e19
2 changed files with 29 additions and 38 deletions
Download patch file Download diff file Expand all files Collapse all files

34
GPy/examples/dimensionality_reduction.py
View file

@ -118,13 +118,13 @@ def mrd_simulation(plot_sim=False):
# Y2 -= Y2.mean(0) # Y2 -= Y2.mean(0)
# make_params = lambda ard: np.hstack([[1], ard, [1, .3]]) # make_params = lambda ard: np.hstack([[1], ard, [1, .3]])
D1, D2, D3, N, M, Q = 50, 100, 8, 200, 2, 5 D1, D2, D3, N, M, Q = 50, 100, 8, 300, 2, 6
x = np.linspace(0, 8 * np.pi, N)[:, None] x = np.linspace(0, 4 * np.pi, N)[:, None]
s1 = np.vectorize(lambda x: np.sin(x)) s1 = np.vectorize(lambda x: np.sin(x))
s2 = np.vectorize(lambda x: np.cos(x)) s2 = np.vectorize(lambda x: x * np.cos(x))
s3 = np.vectorize(lambda x:-np.exp(-np.cos(2 * x))) sS = np.vectorize(lambda x:-np.exp(-np.cos(2 * x)))
sS = np.vectorize(lambda x: x * np.sin(2 * x)) s3 = np.vectorize(lambda x: np.sin(2 * x))
s1 = s1(x) s1 = s1(x)
s2 = s2(x) s2 = s2(x)
@ -161,16 +161,16 @@ def mrd_simulation(plot_sim=False):
Y2 += .5 * np.random.randn(*Y2.shape) Y2 += .5 * np.random.randn(*Y2.shape)
Y3 += .5 * np.random.randn(*Y3.shape) Y3 += .5 * np.random.randn(*Y3.shape)
# Y1 -= Y1.mean(0) Y1 -= Y1.mean(0)
# Y2 -= Y2.mean(0) Y2 -= Y2.mean(0)
# Y3 -= Y3.mean(0) Y3 -= Y3.mean(0)
# Y1 /= Y1.std(0) Y1 /= Y1.std(0)
# Y2 /= Y2.std(0) Y2 /= Y2.std(0)
# Y3 /= Y3.std(0) Y3 /= Y3.std(0)
Slist = [s1, s2, sS] Slist = [s1, s2, sS]
Ylist = [Y1, Y2] Ylist = [Y1]
if plot_sim: if plot_sim:
import pylab import pylab
@ -190,20 +190,22 @@ def mrd_simulation(plot_sim=False):
pylab.tight_layout() pylab.tight_layout()
# k = kern.rbf(Q, ARD=True) + kern.bias(Q) + kern.white(Q) # k = kern.rbf(Q, ARD=True) + kern.bias(Q) + kern.white(Q)
k = kern.linear(Q, ARD=True) + kern.bias(Q) + kern.white(Q)
m = mrd.MRD(*Ylist, Q=Q, M=M, kernel=k, initx="concat", _debug=False) k = kern.linear(Q, ARD=True) + kern.bias(Q, .01) + kern.white(Q, .1)
m = mrd.MRD(*Ylist, Q=Q, M=M, kernel=k, initx="concat", initz='permute', _debug=False)
m.ensure_default_constraints() m.ensure_default_constraints()
ardvar = 5. / (m.X.max(axis=0) - m.X.min(axis=0))
for i, Y in enumerate(Ylist): for i, Y in enumerate(Ylist):
m.set('{}_noise'.format(i + 1), Y.var() / 100.) m.set('{}_noise'.format(i + 1), Y.var() / 100.)
# import ipdb;ipdb.set_trace()
cstr = "variance" cstr = "variance"
m.unconstrain(cstr); m.constrain_bounded(cstr, 1e-15, 1.) m.unconstrain(cstr); m.constrain_bounded(cstr, 1e-12, 1.)
# print "initializing beta" # print "initializing beta"
# cstr = "noise" # cstr = "noise"
# m.unconstrain(cstr); m.constrain_fixed(cstr) # m.unconstrain(cstr); m.constrain_fixed(cstr)
# import ipdb;ipdb.set_trace()
# m.optimize('scg', messages=1, max_f_eval=200) # m.optimize('scg', messages=1, max_f_eval=200)
# #
# print "releasing beta" # print "releasing beta"

33
GPy/models/mrd.py
View file

@ -273,39 +273,28 @@ class MRD(model):
def plot_X_1d(self, colors=None): def plot_X_1d(self, colors=None):
fig = pylab.figure(num="MRD X 1d", figsize=(min(8, (3 * len(self.bgplvms))), min(12, (2 * self.X.shape[1])))) fig = pylab.figure(num="MRD X 1d", figsize=(min(8, (3 * len(self.bgplvms))), min(12, (2 * self.X.shape[1]))))
fig.clf()
ax1 = fig.add_subplot(self.X.shape[1], 1, 1)
if colors is None: if colors is None:
colors = ax1._get_lines.color_cycle colors = pylab.gca()._get_lines.color_cycle
ax1.plot(self.X, c='k', alpha=.3) pylab.clf()
plots = ax1.plot(self.X.T[0], c=colors.next()) plots = []
ax1.fill_between(numpy.arange(self.X.shape[0]), for i in range(self.X.shape[1]):
self.X.T[0] - 2 * numpy.sqrt(self.gref.X_variance.T[0]), ax = fig.add_subplot(self.X.shape[1], 1, i + 1)
self.X.T[0] + 2 * numpy.sqrt(self.gref.X_variance.T[0]),
facecolor=plots[-1].get_color(),
alpha=.3)
ax1.text(1, 1, r"$\mathbf{{X_{}}}".format(1),
horizontalalignment='right',
verticalalignment='top',
transform=ax1.transAxes)
for i in range(self.X.shape[1] - 1):
ax = fig.add_subplot(self.X.shape[1], 1, i + 2)
ax.plot(self.X, c='k', alpha=.3) ax.plot(self.X, c='k', alpha=.3)
plots.extend(ax.plot(self.X.T[i + 1], c=colors.next())) plots.extend(ax.plot(self.X.T[i], c=colors.next(), label=r"$\mathbf{{X_{}}}$".format(i)))
ax.fill_between(numpy.arange(self.X.shape[0]), ax.fill_between(numpy.arange(self.X.shape[0]),
self.X.T[i + 1] - 2 * numpy.sqrt(self.gref.X_variance.T[i + 1]), self.X.T[i] - 2 * numpy.sqrt(self.gref.X_variance.T[i]),
self.X.T[i + 1] + 2 * numpy.sqrt(self.gref.X_variance.T[i + 1]), self.X.T[i] + 2 * numpy.sqrt(self.gref.X_variance.T[i]),
facecolor=plots[-1].get_color(), facecolor=plots[-1].get_color(),
alpha=.3) alpha=.3)
if i < self.X.shape[1] - 2: ax.legend(borderaxespad=0.)
if i < self.X.shape[1] - 1:
ax.set_xticklabels('') ax.set_xticklabels('')
ax1.set_xticklabels('')
# ax1.legend(plots, [r"$\mathbf{{X_{}}}$".format(i + 1) for i in range(self.X.shape[1])], # ax1.legend(plots, [r"$\mathbf{{X_{}}}$".format(i + 1) for i in range(self.X.shape[1])],
# bbox_to_anchor=(0., 1 + .01 * self.X.shape[1], # bbox_to_anchor=(0., 1 + .01 * self.X.shape[1],
# 1., 1. + .01 * self.X.shape[1]), loc=3, # 1., 1. + .01 * self.X.shape[1]), loc=3,
# ncol=self.X.shape[1], mode="expand", borderaxespad=0.) # ncol=self.X.shape[1], mode="expand", borderaxespad=0.)
pylab.draw() pylab.draw()
fig.tight_layout(h_pad=.01, rect=(0, 0, 1, .95)) fig.tight_layout(h_pad=.01) # , rect=(0, 0, 1, .95))
return fig return fig
def plot_X(self): def plot_X(self):