From 73a122362f422f468c996efa550c309947a1a8ca Mon Sep 17 00:00:00 2001 From: Nicolas Date: Wed, 5 Jun 2013 17:29:46 +0100 Subject: [PATCH 1/4] bugs fixed in tutorial's tests --- GPy/examples/tutorials.py | 112 ++++++++++--------------------------- doc/tuto_GP_regression.rst | 4 +- 2 files changed, 30 insertions(+), 86 deletions(-) diff --git a/GPy/examples/tutorials.py b/GPy/examples/tutorials.py index 5d2dd41c..bb5192d8 100644 --- a/GPy/examples/tutorials.py +++ b/GPy/examples/tutorials.py @@ -17,28 +17,27 @@ def tuto_GP_regression(): X = np.random.uniform(-3.,3.,(20,1)) Y = np.sin(X) + np.random.randn(20,1)*0.05 - kernel = GPy.kern.rbf(D=1, variance=1., lengthscale=1.) + kernel = GPy.kern.rbf(input_dim=1, variance=1., lengthscale=1.) m = GPy.models.GP_regression(X,Y,kernel) print m m.plot() + m.ensure_default_constraints() m.constrain_positive('') - m.unconstrain('') # Required to remove the previous constrains - m.constrain_positive('rbf_variance') - m.constrain_bounded('lengthscale',1.,10. ) - m.constrain_fixed('noise',0.0025) + m.unconstrain('') # may be used to remove the previous constrains + m.constrain_positive('.*rbf_variance') + m.constrain_bounded('.*lengthscale',1.,10. ) + m.constrain_fixed('.*noise',0.0025) m.optimize() - m.optimize_restarts(Nrestarts = 10) - - ########################### - # 2-dimensional example # - ########################### + m.optimize_restarts(num_restarts = 10) + ####################################################### + ####################################################### # sample inputs and outputs X = np.random.uniform(-3.,3.,(50,2)) Y = np.sin(X[:,0:1]) * np.sin(X[:,1:2])+np.random.randn(50,1)*0.05 @@ -53,22 +52,19 @@ def tuto_GP_regression(): m.constrain_positive('') # optimize and plot - pb.figure() m.optimize('tnc', max_f_eval = 1000) - m.plot() print(m) - + return(m) def tuto_kernel_overview(): """The detailed explanations of the commands used in this file can be found in the tutorial section""" - pb.ion() - - ker1 = GPy.kern.rbf(1) # Equivalent to ker1 = GPy.kern.rbf(D=1, variance=1., lengthscale=1.) - ker2 = GPy.kern.rbf(D=1, variance = .75, lengthscale=2.) + ker1 = GPy.kern.rbf(1) # Equivalent to ker1 = GPy.kern.rbf(input_dim=1, variance=1., lengthscale=1.) + ker2 = GPy.kern.rbf(input_dim=1, variance = .75, lengthscale=2.) ker3 = GPy.kern.rbf(1, .5, .5) - + print ker2 + ker1.plot() ker2.plot() ker3.plot() @@ -77,28 +73,13 @@ def tuto_kernel_overview(): k2 = GPy.kern.Matern32(1, 0.5, 0.2) # Product of kernels - k_prod = k1.prod(k2) - k_prodorth = k1.prod_orthogonal(k2) + k_prod = k1.prod(k2) # By default, tensor=False + k_prodtens = k1.prod(k2,tensor=True) # Sum of kernels - k_add = k1.add(k2) - k_addorth = k1.add_orthogonal(k2) - - pb.figure(figsize=(8,8)) - pb.subplot(2,2,1) - k_prod.plot() - pb.title('prod') - pb.subplot(2,2,2) - k_prodorth.plot() - pb.title('prod_orthogonal') - pb.subplot(2,2,3) - k_add.plot() - pb.title('add') - pb.subplot(2,2,4) - k_addorth.plot() - pb.title('add_orthogonal') - pb.subplots_adjust(wspace=0.3, hspace=0.3) - + k_add = k1.add(k2) # By default, tensor=False + k_addtens = k1.add(k2,tensor=True) + k1 = GPy.kern.rbf(1,1.,2) k2 = GPy.kern.periodic_Matern52(1,variance=1e3, lengthscale=1, period = 1.5, lower=-5., upper = 5) @@ -109,18 +90,6 @@ def tuto_kernel_overview(): X = np.linspace(-5,5,501)[:,None] Y = np.random.multivariate_normal(np.zeros(501),k.K(X),1) - # plot - pb.figure(figsize=(10,4)) - pb.subplot(1,2,1) - k.plot() - pb.subplot(1,2,2) - pb.plot(X,Y.T) - pb.ylabel("Sample path") - pb.subplots_adjust(wspace=0.3) - - k = (k1+k2)*(k1+k2) - print k.parts[0].name, '\n', k.parts[1].name, '\n', k.parts[2].name, '\n', k.parts[3].name - k1 = GPy.kern.rbf(1) k2 = GPy.kern.Matern32(1) k3 = GPy.kern.white(1) @@ -128,16 +97,16 @@ def tuto_kernel_overview(): k = k1 + k2 + k3 print k - k.constrain_positive('var') + k.constrain_positive('.*var') k.constrain_fixed(np.array([1]),1.75) - k.tie_params('len') + k.tie_params('.*len') k.unconstrain('white') k.constrain_bounded('white',lower=1e-5,upper=.5) print k - + k_cst = GPy.kern.bias(1,variance=1.) k_mat = GPy.kern.Matern52(1,variance=1., lengthscale=3) - Kanova = (k_cst + k_mat).prod_orthogonal(k_cst + k_mat) + Kanova = (k_cst + k_mat).prod(k_cst + k_mat,tensor=True) print Kanova # sample inputs and outputs @@ -146,9 +115,8 @@ def tuto_kernel_overview(): # Create GP regression model m = GPy.models.GP_regression(X,Y,Kanova) - pb.figure(figsize=(5,5)) m.plot() - + pb.figure(figsize=(20,3)) pb.subplots_adjust(wspace=0.5) pb.subplot(1,5,1) @@ -156,41 +124,17 @@ def tuto_kernel_overview(): pb.subplot(1,5,2) pb.ylabel("= ",rotation='horizontal',fontsize='30') pb.subplot(1,5,3) - m.plot(which_functions=[False,True,False,False]) + m.plot(which_parts=[False,True,False,False]) pb.ylabel("cst +",rotation='horizontal',fontsize='30') pb.subplot(1,5,4) - m.plot(which_functions=[False,False,True,False]) + m.plot(which_parts=[False,False,True,False]) pb.ylabel("+ ",rotation='horizontal',fontsize='30') pb.subplot(1,5,5) pb.ylabel("+ ",rotation='horizontal',fontsize='30') - m.plot(which_functions=[False,False,False,True]) + m.plot(which_parts=[False,False,False,True]) - ker1 = GPy.kern.rbf(D=1) # Equivalent to ker1 = GPy.kern.rbf(D=1, variance=1., lengthscale=1.) - ker2 = GPy.kern.rbf(D=1, variance = .75, lengthscale=3.) - ker3 = GPy.kern.rbf(1, .5, .25) + return(m) - ker1.plot() - ker2.plot() - ker3.plot() - #pb.savefig("Figures/tuto_kern_overview_basicdef.png") - - kernels = [GPy.kern.rbf(1), GPy.kern.exponential(1), GPy.kern.Matern32(1), GPy.kern.Matern52(1), GPy.kern.Brownian(1), GPy.kern.bias(1), GPy.kern.linear(1), GPy.kern.spline(1), GPy.kern.periodic_exponential(1), GPy.kern.periodic_Matern32(1), GPy.kern.periodic_Matern52(1), GPy.kern.white(1)] - kernel_names = ["GPy.kern.rbf", "GPy.kern.exponential", "GPy.kern.Matern32", "GPy.kern.Matern52", "GPy.kern.Brownian", "GPy.kern.bias", "GPy.kern.linear", "GPy.kern.spline", "GPy.kern.periodic_exponential", "GPy.kern.periodic_Matern32", "GPy.kern.periodic_Matern52", "GPy.kern.white"] - - pb.figure(figsize=(16,12)) - pb.subplots_adjust(wspace=.5, hspace=.5) - for i, kern in enumerate(kernels): - pb.subplot(3,4,i+1) - kern.plot(x=7.5,plot_limits=[0.00001,15.]) - pb.title(kernel_names[i]+ '\n') - - # actual plot for the noise - i = 11 - X = np.linspace(0.,15.,201) - WN = 0*X - WN[100] = 1. - pb.subplot(3,4,i+1) - pb.plot(X,WN,'b') def model_interaction(): X = np.random.randn(20,1) diff --git a/doc/tuto_GP_regression.rst b/doc/tuto_GP_regression.rst index 87744c85..9f01de93 100644 --- a/doc/tuto_GP_regression.rst +++ b/doc/tuto_GP_regression.rst @@ -25,7 +25,7 @@ The first step is to define the covariance kernel we want to use for the model. kernel = GPy.kern.rbf(input_dim=1, variance=1., lengthscale=1.) -The parameter ``D`` stands for the dimension of the input space. The parameters ``variance`` and ``lengthscale`` are optional. Many other kernels are implemented such as: +The parameter ``input_dim`` stands for the dimension of the input space. The parameters ``variance`` and ``lengthscale`` are optional. Many other kernels are implemented such as: * linear (``GPy.kern.linear``) * exponential kernel (``GPy.kern.exponential``) @@ -69,7 +69,7 @@ There are various ways to constrain the parameters of the kernel. The most basic but it is also possible to set a range on to constrain one parameter to be fixed. The parameter of ``m.constrain_positive`` is a regular expression that matches the name of the parameters to be constrained (as seen in ``print m``). For example, if we want the variance to be positive, the lengthscale to be in [1,10] and the noise variance to be fixed we can write:: - m.unconstrain('') # Required to remove the previous constrains + m.unconstrain('') # may be used to remove the previous constrains m.constrain_positive('.*rbf_variance') m.constrain_bounded('.*lengthscale',1.,10. ) m.constrain_fixed('.*noise',0.0025) From e511a7ba03de1ac14c1b6fc4df08e901f489496e Mon Sep 17 00:00:00 2001 From: Nicolas Date: Wed, 5 Jun 2013 17:32:26 +0100 Subject: [PATCH 2/4] merged conflict in tutorial's tests (again) --- GPy/examples/tutorials.py | 4 ---- 1 file changed, 4 deletions(-) diff --git a/GPy/examples/tutorials.py b/GPy/examples/tutorials.py index fc33d2bc..4371d7a8 100644 --- a/GPy/examples/tutorials.py +++ b/GPy/examples/tutorials.py @@ -114,12 +114,8 @@ def tuto_kernel_overview(): Y = 0.5*X[:,:1] + 0.5*X[:,1:] + 2*np.sin(X[:,:1]) * np.sin(X[:,1:]) # Create GP regression model -<<<<<<< HEAD - m = GPy.models.GP_regression(X,Y,Kanova) -======= m = GPy.models.GPRegression(X, Y, Kanova) pb.figure(figsize=(5,5)) ->>>>>>> efbf169a6a17d824234d538553ffcbe0c4bddc40 m.plot() pb.figure(figsize=(20,3)) From a6ed0031942b24bc4e46c29b9e284c52d78876e0 Mon Sep 17 00:00:00 2001 From: Max Zwiessele Date: Wed, 5 Jun 2013 17:33:22 +0100 Subject: [PATCH 3/4] num_data refactoring --- GPy/core/gp_base.py | 3 +-- GPy/util/plot_latent.py | 28 ++++++++++++++-------------- GPy/util/visualize.py | 12 ++++++------ 3 files changed, 21 insertions(+), 22 deletions(-) diff --git a/GPy/core/gp_base.py b/GPy/core/gp_base.py index 9188fe6f..9c7e4a9e 100644 --- a/GPy/core/gp_base.py +++ b/GPy/core/gp_base.py @@ -28,8 +28,7 @@ class GPBase(Model): self._Xmean = np.zeros((1, self.input_dim)) self._Xstd = np.ones((1, self.input_dim)) - Model.__init__(self) - + super(GPBase, self).__init__() # All leaf nodes should call self._set_params(self._get_params()) at # the end diff --git a/GPy/util/plot_latent.py b/GPy/util/plot_latent.py index e147d840..c36c5e34 100644 --- a/GPy/util/plot_latent.py +++ b/GPy/util/plot_latent.py @@ -2,9 +2,9 @@ import pylab as pb import numpy as np from .. import util -def plot_latent(Model, labels=None, which_indices=None, resolution=50, ax=None, marker='o', s=40): +def plot_latent(model, labels=None, which_indices=None, resolution=50, ax=None, marker='o', s=40): """ - :param labels: a np.array of size Model.N containing labels for the points (can be number, strings, etc) + :param labels: a np.array of size model.num_data containing labels for the points (can be number, strings, etc) :param resolution: the resolution of the grid on which to evaluate the predictive variance """ if ax is None: @@ -12,26 +12,26 @@ def plot_latent(Model, labels=None, which_indices=None, resolution=50, ax=None, util.plot.Tango.reset() if labels is None: - labels = np.ones(Model.N) + labels = np.ones(model.num_data) if which_indices is None: - if Model.input_dim==1: + if model.input_dim==1: input_1 = 0 input_2 = None - if Model.input_dim==2: + if model.input_dim==2: input_1, input_2 = 0,1 else: try: - input_1, input_2 = np.argsort(Model.input_sensitivity())[:2] + input_1, input_2 = np.argsort(model.input_sensitivity())[:2] except: raise ValueError, "cannot Atomatically determine which dimensions to plot, please pass 'which_indices'" else: input_1, input_2 = which_indices #first, plot the output variance as a function of the latent space - Xtest, xx,yy,xmin,xmax = util.plot.x_frame2D(Model.X[:,[input_1, input_2]],resolution=resolution) - Xtest_full = np.zeros((Xtest.shape[0], Model.X.shape[1])) + Xtest, xx,yy,xmin,xmax = util.plot.x_frame2D(model.X[:,[input_1, input_2]],resolution=resolution) + Xtest_full = np.zeros((Xtest.shape[0], model.X.shape[1])) Xtest_full[:, :2] = Xtest - mu, var, low, up = Model.predict(Xtest_full) + mu, var, low, up = model.predict(Xtest_full) var = var[:, :1] ax.imshow(var.reshape(resolution, resolution).T, extent=[xmin[0], xmax[0], xmin[1], xmax[1]], cmap=pb.cm.binary,interpolation='bilinear',origin='lower') @@ -55,12 +55,12 @@ def plot_latent(Model, labels=None, which_indices=None, resolution=50, ax=None, m = marker index = np.nonzero(labels==ul)[0] - if Model.input_dim==1: - x = Model.X[index,input_1] + if model.input_dim==1: + x = model.X[index,input_1] y = np.zeros(index.size) else: - x = Model.X[index,input_1] - y = Model.X[index,input_2] + x = model.X[index,input_1] + y = model.X[index,input_2] ax.scatter(x, y, marker=m, s=s, color=util.plot.Tango.nextMedium(), label=this_label) ax.set_xlabel('latent dimension %i'%input_1) @@ -88,4 +88,4 @@ def plot_latent_indices(Model, which_indices=None, *args, **kwargs): ax = plot_latent(Model, which_indices=[input_1, input_2], *args, **kwargs) # TODO: Here test if there are inducing points... ax.plot(Model.Z[:, input_1], Model.Z[:, input_2], '^w') - return ax \ No newline at end of file + return ax diff --git a/GPy/util/visualize.py b/GPy/util/visualize.py index 66322c15..e13335f9 100644 --- a/GPy/util/visualize.py +++ b/GPy/util/visualize.py @@ -43,16 +43,16 @@ class vector_show(data_show): class lvm(data_show): - def __init__(self, vals, Model, data_visualize, latent_axes=None, sense_axes=None, latent_index=[0,1]): - """Visualize a latent variable Model + def __init__(self, vals, model, data_visualize, latent_axes=None, sense_axes=None, latent_index=[0,1]): + """Visualize a latent variable model - :param Model: the latent variable Model to visualize. + :param model: the latent variable model to visualize. :param data_visualize: the object used to visualize the data which has been modelled. :type data_visualize: visualize.data_show type. :param latent_axes: the axes where the latent visualization should be plotted. """ if vals == None: - vals = Model.X[0] + vals = model.X[0] data_show.__init__(self, vals, axes=latent_axes) @@ -68,13 +68,13 @@ class lvm(data_show): self.cid = latent_axes[0].figure.canvas.mpl_connect('axes_enter_event', self.on_enter) self.data_visualize = data_visualize - self.Model = Model + self.Model = model self.latent_axes = latent_axes self.sense_axes = sense_axes self.called = False self.move_on = False self.latent_index = latent_index - self.latent_dim = Model.input_dim + self.latent_dim = model.input_dim # The red cross which shows current latent point. self.latent_values = vals From 71462a1347b3e7075d0025698c04bf5bcc276492 Mon Sep 17 00:00:00 2001 From: Max Zwiessele Date: Wed, 5 Jun 2013 17:33:37 +0100 Subject: [PATCH 4/4] dim reduction adaption --- GPy/examples/dimensionality_reduction.py | 13 ++++++------- GPy/inference/sgd.py | 10 +++++----- GPy/models/mrd.py | 2 +- 3 files changed, 12 insertions(+), 13 deletions(-) diff --git a/GPy/examples/dimensionality_reduction.py b/GPy/examples/dimensionality_reduction.py index ec6d2ca6..8b2b7a78 100644 --- a/GPy/examples/dimensionality_reduction.py +++ b/GPy/examples/dimensionality_reduction.py @@ -27,7 +27,7 @@ def BGPLVM(seed=default_seed): # k = GPy.kern.rbf(Q, ARD = False) + GPy.kern.white(Q, 0.00001) m = GPy.models.BayesianGPLVM(Y, Q, kernel=k, num_inducing=num_inducing) - m.constrain_positive('(rbf|bias|noise|white|S)') + # m.constrain_positive('(rbf|bias|noise|white|S)') # m.constrain_fixed('S', 1) # pb.figure() @@ -117,10 +117,9 @@ def swiss_roll(optimize=True, N=1000, num_inducing=15, Q=4, sigma=.2, plot=False m.optimize('scg', messages=1) return m -def BGPLVM_oil(optimize=True, N=100, Q=5, num_inducing=25, max_f_eval=4e3, plot=False, **k): +def BGPLVM_oil(optimize=True, N=200, Q=10, num_inducing=15, max_f_eval=4e3, plot=False, **k): np.random.seed(0) data = GPy.util.datasets.oil() - from GPy.core.transformations import logexp_clipped # create simple GP model kernel = GPy.kern.rbf(Q, ARD=True) + GPy.kern.bias(Q, np.exp(-2)) + GPy.kern.white(Q, np.exp(-2)) @@ -132,14 +131,14 @@ def BGPLVM_oil(optimize=True, N=100, Q=5, num_inducing=25, max_f_eval=4e3, plot= m.data_labels = data['Y'][:N].argmax(axis=1) # m.constrain('variance|leng', logexp_clipped()) - m['lengt'] = m.X.var(0).max() / m.X.var(0) + m['.*lengt'] = 1. # m.X.var(0).max() / m.X.var(0) m['noise'] = Yn.var() / 100. m.ensure_default_constraints() # optimize if optimize: - m.optimize('scg', messages=1, max_f_eval=max_f_eval) + m.optimize('scg', messages=1, max_f_eval=max_f_eval, gtol=.05) if plot: y = m.likelihood.Y[0, :] @@ -266,9 +265,9 @@ def bgplvm_simulation(optimize='scg', if optimize: print "Optimizing model:" - m.optimize('scg', max_iters=max_f_eval, + m.optimize(optimize, max_iters=max_f_eval, max_f_eval=max_f_eval, - messages=True, gtol=1e-6) + messages=True, gtol=.05) if plot: m.plot_X_1d("BGPLVM Latent Space 1D") m.kern.plot_ARD('BGPLVM Simulation ARD Parameters') diff --git a/GPy/inference/sgd.py b/GPy/inference/sgd.py index 0002bb22..e443f45a 100644 --- a/GPy/inference/sgd.py +++ b/GPy/inference/sgd.py @@ -18,10 +18,10 @@ class opt_SGD(Optimizer): """ - def __init__(self, start, iterations = 10, learning_rate = 1e-4, momentum = 0.9, Model = None, messages = False, batch_size = 1, self_paced = False, center = True, iteration_file = None, learning_rate_adaptation=None, actual_iter=None, schedule=None, **kwargs): + def __init__(self, start, iterations = 10, learning_rate = 1e-4, momentum = 0.9, model = None, messages = False, batch_size = 1, self_paced = False, center = True, iteration_file = None, learning_rate_adaptation=None, actual_iter=None, schedule=None, **kwargs): self.opt_name = "Stochastic Gradient Descent" - self.Model = Model + self.Model = model self.iterations = iterations self.momentum = momentum self.learning_rate = learning_rate @@ -42,11 +42,11 @@ class opt_SGD(Optimizer): self.learning_rate_0 = self.learning_rate.mean() self.schedule = schedule - # if len([p for p in self.Model.kern.parts if p.name == 'bias']) == 1: + # if len([p for p in self.model.kern.parts if p.name == 'bias']) == 1: # self.param_traces.append(('bias',[])) - # if len([p for p in self.Model.kern.parts if p.name == 'linear']) == 1: + # if len([p for p in self.model.kern.parts if p.name == 'linear']) == 1: # self.param_traces.append(('linear',[])) - # if len([p for p in self.Model.kern.parts if p.name == 'rbf']) == 1: + # if len([p for p in self.model.kern.parts if p.name == 'rbf']) == 1: # self.param_traces.append(('rbf_var',[])) self.param_traces = dict(self.param_traces) diff --git a/GPy/models/mrd.py b/GPy/models/mrd.py index b078fd27..8ebff315 100644 --- a/GPy/models/mrd.py +++ b/GPy/models/mrd.py @@ -78,7 +78,7 @@ class MRD(Model): self.NQ = self.num_data * self.input_dim self.MQ = self.num_inducing * self.input_dim - Model.__init__(self) # @UndefinedVariable + model.__init__(self) # @UndefinedVariable self._set_params(self._get_params()) @property