apunkt/GPy
1
0
Fork 0
mirror of https://github.com/SheffieldML/GPy.git synced 2026-05-15 06:52:39 +02:00
Code Issues Projects Releases Packages Wiki Activity Actions

more robust gradient clippinggit stat

Browse source
This commit is contained in:
Max Zwiessele 2013-06-28 10:59:29 +01:00
parent 3b35d6d321
commit 66a6bde715
1 changed files with 6 additions and 2 deletions
Download patch file Download diff file Expand all files Collapse all files

8
GPy/core/model.py
View file

@ -272,12 +272,13 @@ class Model(Parameterized):
"""
try:
self._set_params_transformed(x)
obj_grads = -self._transform_gradients(self._log_likelihood_gradients() + self._log_prior_gradients())
self._fail_count = 0
except (LinAlgError, ZeroDivisionError, ValueError) as e:
if self._fail_count >= self._allowed_failures:
raise e
self._fail_count += 1
obj_grads = -self._transform_gradients(self._log_likelihood_gradients() + self._log_prior_gradients())
obj_grads = np.clip(-self._transform_gradients(self._log_likelihood_gradients() + self._log_prior_gradients()), -1e100, 1e100)
return obj_grads
def objective_and_gradients(self, x):
@ -285,12 +286,13 @@ class Model(Parameterized):
self._set_params_transformed(x)
obj_f = -self.log_likelihood() - self.log_prior()
self._fail_count = 0
obj_grads = -self._transform_gradients(self._log_likelihood_gradients() + self._log_prior_gradients())
except (LinAlgError, ZeroDivisionError, ValueError) as e:
if self._fail_count >= self._allowed_failures:
raise e
self._fail_count += 1
obj_f = np.inf
obj_grads = -self._transform_gradients(self._log_likelihood_gradients() + self._log_prior_gradients())
obj_grads = np.clip(-self._transform_gradients(self._log_likelihood_gradients() + self._log_prior_gradients()), -1e100, 1e100)
return obj_f, obj_grads
def optimize(self, optimizer=None, start=None, **kwargs):
@ -311,7 +313,9 @@ class Model(Parameterized):
optimizer = optimization.get_optimizer(optimizer)
opt = optimizer(start, model=self, **kwargs)
opt.run(f_fp=self.objective_and_gradients, f=self.objective_function, fp=self.objective_function_gradients)
self.optimization_runs.append(opt)
self._set_params_transformed(opt.x_opt)