diff --git a/GPy/__init__.py b/GPy/__init__.py index 6993d5c2..381d6232 100644 --- a/GPy/__init__.py +++ b/GPy/__init__.py @@ -6,5 +6,6 @@ import kern import models import inference import util +import examples #import examples TODO: discuss! from core import priors diff --git a/GPy/core/model.py b/GPy/core/model.py index 063eaf7d..77e66600 100644 --- a/GPy/core/model.py +++ b/GPy/core/model.py @@ -80,19 +80,22 @@ class model(parameterised): for w in which: self.priors[w] = what - def get(self,name): + def get(self,name, return_names=False): """ - get a model parameter by name + Get a model parameter by name. The name is applied as a regular expression and all parameters that match that regular expression are returned. """ matches = self.grep_param_names(name) if len(matches): - return self._get_params()[matches] + if return_names: + return self._get_params()[matches], np.asarray(self._get_param_names())[matches].tolist() + else: + return self._get_params()[matches] else: raise AttributeError, "no parameter matches %s"%name def set(self,name,val): """ - Set a model parameter by name + Set model parameter(s) by name. The name is provided as a regular expression. All parameters matching that regular expression are set to ghe given value. """ matches = self.grep_param_names(name) if len(matches): @@ -102,6 +105,20 @@ class model(parameterised): else: raise AttributeError, "no parameter matches %s"%name + def get_gradient(self,name, return_names=False): + """ + Get model gradient(s) by name. The name is applied as a regular expression and all parameters that match that regular expression are returned. + """ + matches = self.grep_param_names(name) + if len(matches): + if return_names: + return self._log_likelihood_gradients()[matches], np.asarray(self._get_param_names())[matches].tolist() + else: + return self._log_likelihood_gradients()[matches] + else: + raise AttributeError, "no parameter matches %s"%name + + def log_prior(self): diff --git a/GPy/examples/regression.py b/GPy/examples/regression.py index 79763504..9444e899 100644 --- a/GPy/examples/regression.py +++ b/GPy/examples/regression.py @@ -17,10 +17,8 @@ def toy_rbf_1d(): # create simple GP model m = GPy.models.GP_regression(data['X'],data['Y']) - # contrain all parameters to be positive - m.constrain_positive('') - # optimize + m.ensure_default_constraints() m.optimize() # plot @@ -35,10 +33,8 @@ def rogers_girolami_olympics(): # create simple GP model m = GPy.models.GP_regression(data['X'],data['Y']) - # contrain all parameters to be positive - m.constrain_positive('') - # optimize + m.ensure_default_constraints() m.optimize() # plot @@ -57,10 +53,8 @@ def toy_rbf_1d_50(): # create simple GP model m = GPy.models.GP_regression(data['X'],data['Y']) - # contrain all parameters to be positive - m.constrain_positive('') - # optimize + m.ensure_default_constraints() m.optimize() # plot @@ -75,10 +69,8 @@ def silhouette(): # create simple GP model m = GPy.models.GP_regression(data['X'],data['Y']) - # contrain all parameters to be positive - m.constrain_positive('') - # optimize + m.ensure_default_constraints() m.optimize() print(m) @@ -118,20 +110,15 @@ def multiple_optima(gene_number=937,resolution=80, model_restarts=10, seed=10000 kern = GPy.kern.rbf(1, variance=np.random.exponential(1.), lengthscale=np.random.exponential(50.)) + GPy.kern.white(1,variance=np.random.exponential(1.)) m = GPy.models.GP_regression(data['X'],data['Y'], kernel=kern) - params = m._get_params() - optim_point_x[0] = params[1] - optim_point_y[0] = np.log10(params[0]) - np.log10(params[2]); - - # contrain all parameters to be positive - m.constrain_positive('') + optim_point_x[0] = m.get('rbf_lengthscale') + optim_point_y[0] = np.log10(m.get('rbf_variance')) - np.log10(m.get('white_variance')); # optimize + m.ensure_default_constraints() m.optimize(xtol=1e-6,ftol=1e-6) - params = m._get_params() - optim_point_x[1] = params[1] - optim_point_y[1] = np.log10(params[0]) - np.log10(params[2]); - print(m) + optim_point_x[1] = m.get('rbf_lengthscale') + optim_point_y[1] = np.log10(m.get('rbf_variance')) - np.log10(m.get('white_variance')); pb.arrow(optim_point_x[0], optim_point_y[0], optim_point_x[1]-optim_point_x[0], optim_point_y[1]-optim_point_y[0], label=str(i), head_length=1, head_width=0.5, fc='k', ec='k') models.append(m) diff --git a/GPy/models/GP_regression.py b/GPy/models/GP_regression.py index eee0fb58..72a24307 100644 --- a/GPy/models/GP_regression.py +++ b/GPy/models/GP_regression.py @@ -63,10 +63,10 @@ class GP_regression(model): self._Ystd = np.ones((1,self.Y.shape[1])) if self.D > self.N: - # then it's more efficient to store Youter - self.Youter = np.dot(self.Y, self.Y.T) + # then it's more efficient to store YYT + self.YYT = np.dot(self.Y, self.Y.T) else: - self.Youter = None + self.YYT = None model.__init__(self) @@ -83,23 +83,23 @@ class GP_regression(model): def _model_fit_term(self): """ - Computes the model fit using Youter if it's available + Computes the model fit using YYT if it's available """ - if self.Youter is None: + if self.YYT is None: return -0.5*np.sum(np.square(np.dot(self.Li,self.Y))) else: - return -0.5*np.sum(np.multiply(self.Ki, self.Youter)) + return -0.5*np.sum(np.multiply(self.Ki, self.YYT)) def log_likelihood(self): complexity_term = -0.5*self.N*self.D*np.log(2.*np.pi) - 0.5*self.D*self.K_logdet return complexity_term + self._model_fit_term() def dL_dK(self): - if self.Youter is None: + if self.YYT is None: alpha = np.dot(self.Ki,self.Y) dL_dK = 0.5*(np.dot(alpha,alpha.T)-self.D*self.Ki) else: - dL_dK = 0.5*(mdot(self.Ki, self.Youter, self.Ki) - self.D*self.Ki) + dL_dK = 0.5*(mdot(self.Ki, self.YYT, self.Ki) - self.D*self.Ki) return dL_dK diff --git a/GPy/models/generalized_FITC.py b/GPy/models/generalized_FITC.py index d8e9c23d..a5ed8d0a 100644 --- a/GPy/models/generalized_FITC.py +++ b/GPy/models/generalized_FITC.py @@ -91,9 +91,9 @@ class generalized_FITC(model): def log_likelihood(self): self.posterior_param() - self.Youter = np.dot(self.mu_tilde,self.mu_tilde.T) + self.YYT = np.dot(self.mu_tilde,self.mu_tilde.T) A = -self.hld - B = -.5*np.sum(self.Qi*self.Youter) + B = -.5*np.sum(self.Qi*self.YYT) C = sum(np.log(self.ep_approx.Z_hat)) D = .5*np.sum(np.log(1./self.ep_approx.tau_tilde + 1./self.ep_approx.tau_)) E = .5*np.sum((self.ep_approx.v_/self.ep_approx.tau_ - self.mu_tilde.flatten())**2/(1./self.ep_approx.tau_ + 1./self.ep_approx.tau_tilde)) diff --git a/GPy/models/warped_GP.py b/GPy/models/warped_GP.py index ff6267d2..8ce80c76 100644 --- a/GPy/models/warped_GP.py +++ b/GPy/models/warped_GP.py @@ -48,9 +48,9 @@ class warpedGP(GP_regression): # this supports the 'smart' behaviour in GP_regression if self.D > self.N: - self.Youter = np.dot(self.Y, self.Y.T) + self.YYT = np.dot(self.Y, self.Y.T) else: - self.Youter = None + self.YYT = None return self.Y