apunkt/GPy
1
0
Fork 0
mirror of https://github.com/SheffieldML/GPy.git synced 2026-05-21 14:05:14 +02:00
Code Issues Projects Releases Packages Wiki Activity Actions

Merge branch 'params' of https://github.com/SheffieldML/GPy into params

Browse source
This commit is contained in:
Neil Lawrence 2014-03-03 17:05:20 +00:00
commit d2c4c10573
9 changed files with 149 additions and 143 deletions
Download patch file Download diff file Expand all files Collapse all files

6
GPy/core/gp.py
View file

@ -31,7 +31,7 @@ class GP(Model):
super(GP, self).__init__(name) super(GP, self).__init__(name)
assert X.ndim == 2 assert X.ndim == 2
if isinstance(X, ObservableArray) or isinstance(X, VariationalPosterior): if isinstance(X, (ObservableArray, VariationalPosterior)):
self.X = X self.X = X
else: self.X = ObservableArray(X) else: self.X = ObservableArray(X)
@ -224,13 +224,9 @@ class GP(Model):
self.kern, self.kern,
self.likelihood, self.likelihood,
self.output_dim, self.output_dim,
self._Xoffset,
self._Xscale,
] ]
def _setstate(self, state): def _setstate(self, state):
self._Xscale = state.pop()
self._Xoffset = state.pop()
self.output_dim = state.pop() self.output_dim = state.pop()
self.likelihood = state.pop() self.likelihood = state.pop()
self.kern = state.pop() self.kern = state.pop()

3
GPy/core/parameterization/param.py
View file

@ -247,7 +247,8 @@ class Param(OptimizationHandlable, ObservableArray, Gradcheckable):
#=========================================================================== #===========================================================================
@property @property
def _description_str(self): def _description_str(self):
if self.size <= 1: return ["%f" % self] if self.size <= 1:
return [str(numpy.take(self, 0))]
else: return [str(self.shape)] else: return [str(self.shape)]
def parameter_names(self, add_self=False, adjust_for_printing=False): def parameter_names(self, add_self=False, adjust_for_printing=False):
if adjust_for_printing: if adjust_for_printing:

138
GPy/core/parameterization/parameter_core.py
View file

@ -2,7 +2,7 @@
# Licensed under the BSD 3-clause license (see LICENSE.txt) # Licensed under the BSD 3-clause license (see LICENSE.txt)
from transformations import Transformation, Logexp, NegativeLogexp, Logistic, __fixed__, FIXED, UNFIXED from transformations import Transformation, Logexp, NegativeLogexp, Logistic, __fixed__, FIXED, UNFIXED
import heapq import numpy as np
__updated__ = '2013-12-16' __updated__ = '2013-12-16'
@ -22,7 +22,7 @@ class Observable(object):
self._observer_callables_ = [] self._observer_callables_ = []
def add_observer(self, observer, callble, priority=0): def add_observer(self, observer, callble, priority=0):
heapq.heappush(self._observer_callables_, (priority, observer, callble)) self._insert_sorted(priority, observer, callble)
def remove_observer(self, observer, callble=None): def remove_observer(self, observer, callble=None):
to_remove = [] to_remove = []
@ -51,9 +51,20 @@ class Observable(object):
if which is None: if which is None:
which = self which = self
if min_priority is None: if min_priority is None:
[callble(which) for _, _, callble in heapq.nlargest(len(self._observer_callables_), self._observer_callables_)] [callble(which) for _, _, callble in self._observer_callables_]
else: else:
[callble(which) for p, _, callble in heapq.nlargest(len(self._observer_callables_), self._observer_callables_) if p > min_priority] for p, _, callble in self._observer_callables_:
if p <= min_priority:
break
callble(which)
def _insert_sorted(self, p, o, c):
ins = 0
for pr, _, _ in self._observer_callables_:
if p > pr:
break
ins += 1
self._observer_callables_.insert(ins, (p, o, c))
class Pickleable(object): class Pickleable(object):
def _getstate(self): def _getstate(self):
@ -202,20 +213,17 @@ class Constrainable(Nameable, Indexable):
unfix = unconstrain_fixed unfix = unconstrain_fixed
def _set_fixed(self, index): def _set_fixed(self, index):
import numpy as np
if not self._has_fixes(): self._fixes_ = np.ones(self.size, dtype=bool) if not self._has_fixes(): self._fixes_ = np.ones(self.size, dtype=bool)
self._fixes_[index] = FIXED self._fixes_[index] = FIXED
if np.all(self._fixes_): self._fixes_ = None # ==UNFIXED if np.all(self._fixes_): self._fixes_ = None # ==UNFIXED
def _set_unfixed(self, index): def _set_unfixed(self, index):
import numpy as np
if not self._has_fixes(): self._fixes_ = np.ones(self.size, dtype=bool) if not self._has_fixes(): self._fixes_ = np.ones(self.size, dtype=bool)
# rav_i = self._raveled_index_for(param)[index] # rav_i = self._raveled_index_for(param)[index]
self._fixes_[index] = UNFIXED self._fixes_[index] = UNFIXED
if np.all(self._fixes_): self._fixes_ = None # ==UNFIXED if np.all(self._fixes_): self._fixes_ = None # ==UNFIXED
def _connect_fixes(self): def _connect_fixes(self):
import numpy as np
fixed_indices = self.constraints[__fixed__] fixed_indices = self.constraints[__fixed__]
if fixed_indices.size > 0: if fixed_indices.size > 0:
self._fixes_ = np.ones(self.size, dtype=bool) * UNFIXED self._fixes_ = np.ones(self.size, dtype=bool) * UNFIXED
@ -245,7 +253,6 @@ class Constrainable(Nameable, Indexable):
def _log_prior_gradients(self): def _log_prior_gradients(self):
"""evaluate the gradients of the priors""" """evaluate the gradients of the priors"""
import numpy as np
if self.priors.size > 0: if self.priors.size > 0:
x = self._get_params() x = self._get_params()
ret = np.zeros(x.size) ret = np.zeros(x.size)
@ -342,7 +349,6 @@ class Constrainable(Nameable, Indexable):
def _remove_from_index_operations(self, which, transforms): def _remove_from_index_operations(self, which, transforms):
if len(transforms) == 0: if len(transforms) == 0:
transforms = which.properties() transforms = which.properties()
import numpy as np
removed = np.empty((0,), dtype=int) removed = np.empty((0,), dtype=int)
for t in transforms: for t in transforms:
unconstrained = which.remove(t, self._raveled_index()) unconstrained = which.remove(t, self._raveled_index())
@ -404,7 +410,6 @@ class OptimizationHandlable(Constrainable, Observable):
Randomize the model. Randomize the model.
Make this draw from the prior if one exists, else draw from N(0,1) Make this draw from the prior if one exists, else draw from N(0,1)
""" """
import numpy as np
# first take care of all parameters (from N(0,1)) # first take care of all parameters (from N(0,1))
# x = self._get_params_transformed() # x = self._get_params_transformed()
x = np.random.randn(self._size_transformed()) x = np.random.randn(self._size_transformed())
@ -414,9 +419,6 @@ class OptimizationHandlable(Constrainable, Observable):
self._set_params(x) self._set_params(x)
# self._set_params_transformed(self._get_params_transformed()) # makes sure all of the tied parameters get the same init (since there's only one prior object...) # self._set_params_transformed(self._get_params_transformed()) # makes sure all of the tied parameters get the same init (since there's only one prior object...)
import numpy as np
class Parameterizable(OptimizationHandlable): class Parameterizable(OptimizationHandlable):
def __init__(self, *args, **kwargs): def __init__(self, *args, **kwargs):
super(Parameterizable, self).__init__(*args, **kwargs) super(Parameterizable, self).__init__(*args, **kwargs)
@ -475,6 +477,112 @@ class Parameterizable(OptimizationHandlable):
import itertools import itertools
[p._set_gradient(g[s]) for p, s in itertools.izip(self._parameters_, self._param_slices_)] [p._set_gradient(g[s]) for p, s in itertools.izip(self._parameters_, self._param_slices_)]
def add_parameter(self, param, index=None):
"""
:param parameters: the parameters to add
:type parameters: list of or one :py:class:`GPy.core.param.Param`
:param [index]: index of where to put parameters
Add all parameters to this param class, you can insert parameters
at any given index using the :func:`list.insert` syntax
"""
# if param.has_parent():
# raise AttributeError, "parameter {} already in another model, create new object (or copy) for adding".format(param._short())
if param in self._parameters_ and index is not None:
self.remove_parameter(param)
self.add_parameter(param, index)
elif param not in self._parameters_:
if param.has_parent():
parent = param._direct_parent_
while parent is not None:
if parent is self:
raise HierarchyError, "You cannot add a parameter twice into the hirarchy"
parent = parent._direct_parent_
param._direct_parent_.remove_parameter(param)
# make sure the size is set
if index is None:
self.constraints.update(param.constraints, self.size)
self.priors.update(param.priors, self.size)
self._parameters_.append(param)
else:
start = sum(p.size for p in self._parameters_[:index])
self.constraints.shift_right(start, param.size)
self.priors.shift_right(start, param.size)
self.constraints.update(param.constraints, start)
self.priors.update(param.priors, start)
self._parameters_.insert(index, param)
param.add_observer(self, self._pass_through_notify_observers, -np.inf)
self.size += param.size
self._connect_parameters()
self._notify_parent_change()
self._connect_fixes()
else:
raise RuntimeError, """Parameter exists already added and no copy made"""
def add_parameters(self, *parameters):
"""
convenience method for adding several
parameters without gradient specification
"""
[self.add_parameter(p) for p in parameters]
def remove_parameter(self, param):
"""
:param param: param object to remove from being a parameter of this parameterized object.
"""
if not param in self._parameters_:
raise RuntimeError, "Parameter {} does not belong to this object, remove parameters directly from their respective parents".format(param._short())
start = sum([p.size for p in self._parameters_[:param._parent_index_]])
self._remove_parameter_name(param)
self.size -= param.size
del self._parameters_[param._parent_index_]
param._disconnect_parent()
param.remove_observer(self, self._pass_through_notify_observers)
self.constraints.shift_left(start, param.size)
self._connect_fixes()
self._connect_parameters()
self._notify_parent_change()
parent = self._direct_parent_
while parent is not None:
parent._connect_fixes()
parent._connect_parameters()
parent._notify_parent_change()
parent = parent._direct_parent_
def _connect_parameters(self):
# connect parameterlist to this parameterized object
# This just sets up the right connection for the params objects
# to be used as parameters
# it also sets the constraints for each parameter to the constraints
# of their respective parents
if not hasattr(self, "_parameters_") or len(self._parameters_) < 1:
# no parameters for this class
return
sizes = [0]
self._param_slices_ = []
for i, p in enumerate(self._parameters_):
p._direct_parent_ = self
p._parent_index_ = i
sizes.append(p.size + sizes[-1])
self._param_slices_.append(slice(sizes[-2], sizes[-1]))
self._add_parameter_name(p)
#===========================================================================
# notification system
#===========================================================================
def _parameters_changed_notification(self, which):
self.parameters_changed()
def _pass_through_notify_observers(self, which):
self._notify_observers(which)
#=========================================================================== #===========================================================================
# TODO: not working yet # TODO: not working yet
@ -487,7 +595,7 @@ class Parameterizable(OptimizationHandlable):
dc = dict() dc = dict()
for k, v in self.__dict__.iteritems(): for k, v in self.__dict__.iteritems():
if k not in ['_direct_parent_', '_parameters_', '_parent_index_'] + self.parameter_names(): if k not in ['_direct_parent_', '_parameters_', '_parent_index_', '_observer_callables_'] + self.parameter_names():
if isinstance(v, (Constrainable, ParameterIndexOperations, ParameterIndexOperationsView)): if isinstance(v, (Constrainable, ParameterIndexOperations, ParameterIndexOperationsView)):
dc[k] = v.copy() dc[k] = v.copy()
else: else:
@ -497,6 +605,7 @@ class Parameterizable(OptimizationHandlable):
dc['_direct_parent_'] = None dc['_direct_parent_'] = None
dc['_parent_index_'] = None dc['_parent_index_'] = None
dc['_observer_callables_'] = []
dc['_parameters_'] = ArrayList() dc['_parameters_'] = ArrayList()
dc['constraints'].clear() dc['constraints'].clear()
dc['priors'].clear() dc['priors'].clear()
@ -506,6 +615,7 @@ class Parameterizable(OptimizationHandlable):
s.__dict__ = dc s.__dict__ = dc
for p in params: for p in params:
import ipdb;ipdb.set_trace()
s.add_parameter(p) s.add_parameter(p)
return s return s

125
GPy/core/parameterization/parameterized.py
View file

@ -88,114 +88,6 @@ class Parameterized(Parameterizable, Pickleable, Gradcheckable):
return G return G
return node return node
def add_parameter(self, param, index=None):
"""
:param parameters: the parameters to add
:type parameters: list of or one :py:class:`GPy.core.param.Param`
:param [index]: index of where to put parameters
Add all parameters to this param class, you can insert parameters
at any given index using the :func:`list.insert` syntax
"""
# if param.has_parent():
# raise AttributeError, "parameter {} already in another model, create new object (or copy) for adding".format(param._short())
if param in self._parameters_ and index is not None:
self.remove_parameter(param)
self.add_parameter(param, index)
elif param not in self._parameters_:
if param.has_parent():
parent = param._direct_parent_
while parent is not None:
if parent is self:
from GPy.core.parameterization.parameter_core import HierarchyError
raise HierarchyError, "You cannot add a parameter twice into the hirarchy"
parent = parent._direct_parent_
param._direct_parent_.remove_parameter(param)
# make sure the size is set
if index is None:
self.constraints.update(param.constraints, self.size)
self.priors.update(param.priors, self.size)
self._parameters_.append(param)
else:
start = sum(p.size for p in self._parameters_[:index])
self.constraints.shift_right(start, param.size)
self.priors.shift_right(start, param.size)
self.constraints.update(param.constraints, start)
self.priors.update(param.priors, start)
self._parameters_.insert(index, param)
param.add_observer(self, self._pass_through_notify_observers, -np.inf)
self.size += param.size
self._connect_parameters()
self._notify_parent_change()
self._connect_fixes()
else:
raise RuntimeError, """Parameter exists already added and no copy made"""
def add_parameters(self, *parameters):
"""
convenience method for adding several
parameters without gradient specification
"""
[self.add_parameter(p) for p in parameters]
def remove_parameter(self, param):
"""
:param param: param object to remove from being a parameter of this parameterized object.
"""
if not param in self._parameters_:
raise RuntimeError, "Parameter {} does not belong to this object, remove parameters directly from their respective parents".format(param._short())
start = sum([p.size for p in self._parameters_[:param._parent_index_]])
self._remove_parameter_name(param)
self.size -= param.size
del self._parameters_[param._parent_index_]
param._disconnect_parent()
param.remove_observer(self, self._pass_through_notify_observers)
self.constraints.shift_left(start, param.size)
self._connect_fixes()
self._connect_parameters()
self._notify_parent_change()
parent = self._direct_parent_
while parent is not None:
parent._connect_fixes()
parent._connect_parameters()
parent._notify_parent_change()
parent = parent._direct_parent_
def _connect_parameters(self):
# connect parameterlist to this parameterized object
# This just sets up the right connection for the params objects
# to be used as parameters
# it also sets the constraints for each parameter to the constraints
# of their respective parents
if not hasattr(self, "_parameters_") or len(self._parameters_) < 1:
# no parameters for this class
return
sizes = [0]
self._param_slices_ = []
for i, p in enumerate(self._parameters_):
p._direct_parent_ = self
p._parent_index_ = i
sizes.append(p.size + sizes[-1])
self._param_slices_.append(slice(sizes[-2], sizes[-1]))
self._add_parameter_name(p)
#===========================================================================
# notification system
#===========================================================================
def _parameters_changed_notification(self, which):
self.parameters_changed()
def _pass_through_notify_observers(self, which):
self._notify_observers(which)
#=========================================================================== #===========================================================================
# Pickling operations # Pickling operations
#=========================================================================== #===========================================================================
@ -212,6 +104,11 @@ class Parameterized(Parameterizable, Pickleable, Gradcheckable):
else: else:
cPickle.dump(self, f, protocol) cPickle.dump(self, f, protocol)
def copy(self):
c = super(Parameterized, self).copy()
c.add_observer(c, c._parameters_changed_notification, -100)
return c
def __getstate__(self): def __getstate__(self):
if self._has_get_set_state(): if self._has_get_set_state():
return self._getstate() return self._getstate()
@ -332,9 +229,13 @@ class Parameterized(Parameterizable, Pickleable, Gradcheckable):
return ParamConcatenation(paramlist) return ParamConcatenation(paramlist)
def __setitem__(self, name, value, paramlist=None): def __setitem__(self, name, value, paramlist=None):
try: param = self.__getitem__(name, paramlist) if isinstance(name, slice):
except AttributeError as a: raise a self[''][name] = value
param[:] = value else:
try: param = self.__getitem__(name, paramlist)
except AttributeError as a: raise a
param[:] = value
def __setattr__(self, name, val): def __setattr__(self, name, val):
# override the default behaviour, if setting a param, so broadcasting can by used # override the default behaviour, if setting a param, so broadcasting can by used
if hasattr(self, '_parameters_'): if hasattr(self, '_parameters_'):
@ -379,7 +280,7 @@ class Parameterized(Parameterizable, Pickleable, Gradcheckable):
cl = max([len(str(x)) if x else 0 for x in constrs + ["Constraint"]]) cl = max([len(str(x)) if x else 0 for x in constrs + ["Constraint"]])
tl = max([len(str(x)) if x else 0 for x in ts + ["Tied to"]]) tl = max([len(str(x)) if x else 0 for x in ts + ["Tied to"]])
pl = max([len(str(x)) if x else 0 for x in prirs + ["Prior"]]) pl = max([len(str(x)) if x else 0 for x in prirs + ["Prior"]])
format_spec = " \033[1m{{name:<{0}s}}\033[0;0m | {{desc:^{1}s}} | {{const:^{2}s}} | {{pri:^{3}s}} | {{t:^{4}s}}".format(nl, sl, cl, pl, tl) format_spec = " \033[1m{{name:<{0}s}}\033[0;0m | {{desc:>{1}s}} | {{const:^{2}s}} | {{pri:^{3}s}} | {{t:^{4}s}}".format(nl, sl, cl, pl, tl)
to_print = [] to_print = []
for n, d, c, t, p in itertools.izip(names, desc, constrs, ts, prirs): for n, d, c, t, p in itertools.izip(names, desc, constrs, ts, prirs):
to_print.append(format_spec.format(name=n, desc=d, const=c, t=t, pri=p)) to_print.append(format_spec.format(name=n, desc=d, const=c, t=t, pri=p))

8
GPy/core/sparse_gp.py
View file

@ -114,14 +114,12 @@ class SparseGP(GP):
def _getstate(self): def _getstate(self):
""" """
Get the current state of the class, Get the current state of the class,
here just all the indices, rest can get recomputed
""" """
return GP._getstate(self) + [self.Z, return GP._getstate(self) + [
self.num_inducing, self.Z,
self.X_variance] self.num_inducing]
def _setstate(self, state): def _setstate(self, state):
self.X_variance = state.pop()
self.num_inducing = state.pop() self.num_inducing = state.pop()
self.Z = state.pop() self.Z = state.pop()
GP._setstate(self, state) GP._setstate(self, state)

2
GPy/kern/_src/kern.py
View file

@ -73,7 +73,7 @@ class Kern(Parameterized):
See GPy.plotting.matplot_dep.plot See GPy.plotting.matplot_dep.plot
""" """
assert "matplotlib" in sys.modules, "matplotlib package has not been imported." assert "matplotlib" in sys.modules, "matplotlib package has not been imported."
from ..plotting.matplot_dep import kernel_plots from ...plotting.matplot_dep import kernel_plots
kernel_plots.plot(self,*args) kernel_plots.plot(self,*args)
def plot_ARD(self, *args, **kw): def plot_ARD(self, *args, **kw):

4
GPy/models/sparse_gp_regression.py
View file

@ -88,7 +88,7 @@ class SparseGPRegressionUncertainInput(SparseGP):
# kern defaults to rbf (plus white for stability) # kern defaults to rbf (plus white for stability)
if kernel is None: if kernel is None:
kernel = kern.rbf(input_dim) + kern.white(input_dim, variance=1e-3) kernel = kern.RBF(input_dim) + kern.White(input_dim, variance=1e-3)
# Z defaults to a subset of the data # Z defaults to a subset of the data
if Z is None: if Z is None:
@ -99,5 +99,5 @@ class SparseGPRegressionUncertainInput(SparseGP):
likelihood = likelihoods.Gaussian() likelihood = likelihoods.Gaussian()
SparseGP.__init__(self, X, Y, Z, kernel, likelihood, X_variance=X_variance) SparseGP.__init__(self, X, Y, Z, kernel, likelihood, X_variance=X_variance, inference_method=VarDTC())
self.ensure_default_constraints() self.ensure_default_constraints()

2
GPy/plotting/matplot_dep/kernel_plots.py
View file

@ -106,7 +106,7 @@ def plot(kernel, x=None, plot_limits=None, which_parts='all', resolution=None, *
raise ValueError, "Bad limits for plotting" raise ValueError, "Bad limits for plotting"
Xnew = np.linspace(xmin, xmax, resolution or 201)[:, None] Xnew = np.linspace(xmin, xmax, resolution or 201)[:, None]
Kx = kernel.K(Xnew, x, which_parts) Kx = kernel.K(Xnew, x)
pb.plot(Xnew, Kx, *args, **kwargs) pb.plot(Xnew, Kx, *args, **kwargs)
pb.xlim(xmin, xmax) pb.xlim(xmin, xmax)
pb.xlabel("x") pb.xlabel("x")