apunkt/GPy
1
0
Fork 0
mirror of https://github.com/SheffieldML/GPy.git synced 2026-07-20 16:51:05 +02:00
Code Issues Projects Releases Packages Wiki Activity Actions

examples corrected

Browse source
This commit is contained in:
Ricardo 2013-06-04 18:54:29 +01:00
parent f167940544
commit d383403c8e
1 changed files with 8 additions and 66 deletions
Download patch file Download diff file Expand all files Collapse all files

74
GPy/examples/classification.py
View file

@ -24,25 +24,11 @@ def crescent_data(seed=default_seed): # FIXME
Y = data['Y'] Y = data['Y']
Y[Y.flatten()==-1] = 0 Y[Y.flatten()==-1] = 0
# Kernel object m = GPy.models.GP_classification(data['X'], Y)
kernel = GPy.kern.rbf(data['X'].shape[1])
# Likelihood object
distribution = GPy.likelihoods.likelihood_functions.binomial()
likelihood = GPy.likelihoods.EP(Y, distribution)
m = GPy.models.GP(data['X'], likelihood, kernel)
m.ensure_default_constraints() m.ensure_default_constraints()
m.update_likelihood_approximation() m.update_likelihood_approximation()
print(m)
# optimize
m.optimize() m.optimize()
print(m) print(m)
# plot
m.plot() m.plot()
return m return m
@ -54,19 +40,12 @@ def oil():
Y = data['Y'][:, 0:1] Y = data['Y'][:, 0:1]
Y[Y.flatten()==-1] = 0 Y[Y.flatten()==-1] = 0
# Kernel object
kernel = GPy.kern.rbf(12)
# Likelihood object
distribution = GPy.likelihoods.likelihood_functions.binomial()
likelihood = GPy.likelihoods.EP(Y, distribution)
# Create GP model # Create GP model
m = GPy.models.GP_classification(data['X'], Y, kernel=kernel) m = GPy.models.GP_classification(data['X'], Y)
# Contrain all parameters to be positive # Contrain all parameters to be positive
m.constrain_positive('') m.constrain_positive('')
m.tie_params('lengthscale') m.tie_params('.*len')
m.update_likelihood_approximation() m.update_likelihood_approximation()
# Optimize # Optimize
@ -86,25 +65,14 @@ def toy_linear_1d_classification(seed=default_seed):
Y = data['Y'][:, 0:1] Y = data['Y'][:, 0:1]
Y[Y.flatten() == -1] = 0 Y[Y.flatten() == -1] = 0
# Kernel object
kernel = GPy.kern.rbf(1)
# Likelihood object
link = GPy.likelihoods.link_functions.probit
distribution = GPy.likelihoods.likelihood_functions.binomial(link)
likelihood = GPy.likelihoods.EP(Y, distribution)
Y[1] = 1
# Model definition # Model definition
#m = GPy.models.GP(data['X'], likelihood=likelihood, kernel=kernel) m = GPy.models.GP_classification(data['X'], Y)
m = GPy.models.GP_classification(data['X'], Y, likelihood=likelihood, kernel=kernel)
m.ensure_default_constraints() m.ensure_default_constraints()
# Optimize # Optimize
m.update_likelihood_approximation() m.update_likelihood_approximation()
# Parameters optimization: # Parameters optimization:
m.optimize() m.optimize()
# m.pseudo_EM() #FIXME
# Plot # Plot
pb.subplot(211) pb.subplot(211)
@ -126,25 +94,15 @@ def sparse_toy_linear_1d_classification(seed=default_seed):
Y = data['Y'][:, 0:1] Y = data['Y'][:, 0:1]
Y[Y.flatten() == -1] = 0 Y[Y.flatten() == -1] = 0
# Kernel object
kernel = GPy.kern.rbf(1) + GPy.kern.white(1)
# Likelihood object
distribution = GPy.likelihoods.likelihood_functions.binomial()
likelihood = GPy.likelihoods.EP(Y, distribution)
Z = np.random.uniform(data['X'].min(), data['X'].max(), (10, 1))
# Model definition # Model definition
m = GPy.models.sparse_GP(data['X'], likelihood=likelihood, kernel=kernel, Z=Z, normalize_X=False) m = GPy.models.sparse_GP_classification(data['X'], Y)
m.set('len', 2.) m['.*len']= 2.
m.ensure_default_constraints() m.ensure_default_constraints()
# Optimize # Optimize
m.update_likelihood_approximation() m.update_likelihood_approximation()
# Parameters optimization: # Parameters optimization:
m.optimize() m.optimize()
# m.EPEM() #FIXME
# Plot # Plot
pb.subplot(211) pb.subplot(211)
@ -169,27 +127,11 @@ def sparse_crescent_data(inducing=10, seed=default_seed):
Y = data['Y'] Y = data['Y']
Y[Y.flatten()==-1]=0 Y[Y.flatten()==-1]=0
# Kernel object m = GPy.models.sparse_GP_classification(data['X'], Y)
kernel = GPy.kern.rbf(data['X'].shape[1]) + GPy.kern.white(data['X'].shape[1])
# Likelihood object
distribution = GPy.likelihoods.likelihood_functions.binomial()
likelihood = GPy.likelihoods.EP(Y, distribution)
sample = np.random.randint(0, data['X'].shape[0], inducing)
Z = data['X'][sample, :]
# create sparse GP EP model
m = GPy.models.sparse_GP(data['X'], likelihood=likelihood, kernel=kernel, Z=Z)
m.ensure_default_constraints() m.ensure_default_constraints()
m.set('len', 10.) m['.*len'] = 10.
m.update_likelihood_approximation() m.update_likelihood_approximation()
# optimize
m.optimize() m.optimize()
print(m) print(m)
# plot
m.plot() m.plot()
return m return m