GPy/GPy/core/model.py
2014-08-05 08:28:03 -07:00

371 lines
15 KiB
Python

# Copyright (c) 2012, 2013, GPy authors (see AUTHORS.txt).
# Licensed under the BSD 3-clause license (see LICENSE.txt)
from .. import likelihoods
from ..inference import optimization
from ..util.misc import opt_wrapper
from parameterization import Parameterized
import multiprocessing as mp
import numpy as np
from numpy.linalg.linalg import LinAlgError
import itertools
# import numdifftools as ndt
class Model(Parameterized):
_fail_count = 0 # Count of failed optimization steps (see objective)
_allowed_failures = 10 # number of allowed failures
def __init__(self, name):
super(Model, self).__init__(name) # Parameterized.__init__(self)
self.optimization_runs = []
self.sampling_runs = []
self.preferred_optimizer = 'bfgs'
def log_likelihood(self):
raise NotImplementedError, "this needs to be implemented to use the model class"
def _log_likelihood_gradients(self):
return self.gradient
def optimize_restarts(self, num_restarts=10, robust=False, verbose=True, parallel=False, num_processes=None, **kwargs):
"""
Perform random restarts of the model, and set the model to the best
seen solution.
If the robust flag is set, exceptions raised during optimizations will
be handled silently. If _all_ runs fail, the model is reset to the
existing parameter values.
**Notes**
:param num_restarts: number of restarts to use (default 10)
:type num_restarts: int
:param robust: whether to handle exceptions silently or not (default False)
:type robust: bool
:param parallel: whether to run each restart as a separate process. It relies on the multiprocessing module.
:type parallel: bool
:param num_processes: number of workers in the multiprocessing pool
:type numprocesses: int
\*\*kwargs are passed to the optimizer. They can be:
:param max_f_eval: maximum number of function evaluations
:type max_f_eval: int
:param max_iters: maximum number of iterations
:type max_iters: int
:param messages: whether to display during optimisation
:type messages: bool
.. note:: If num_processes is None, the number of workes in the
multiprocessing pool is automatically set to the number of processors
on the current machine.
"""
initial_parameters = self.optimizer_array.copy()
if parallel:
try:
jobs = []
pool = mp.Pool(processes=num_processes)
for i in range(num_restarts):
self.randomize()
job = pool.apply_async(opt_wrapper, args=(self,), kwds=kwargs)
jobs.append(job)
pool.close() # signal that no more data coming in
pool.join() # wait for all the tasks to complete
except KeyboardInterrupt:
print "Ctrl+c received, terminating and joining pool."
pool.terminate()
pool.join()
for i in range(num_restarts):
try:
if not parallel:
self.randomize()
self.optimize(**kwargs)
else:
self.optimization_runs.append(jobs[i].get())
if verbose:
print("Optimization restart {0}/{1}, f = {2}".format(i + 1, num_restarts, self.optimization_runs[-1].f_opt))
except Exception as e:
if robust:
print("Warning - optimization restart {0}/{1} failed".format(i + 1, num_restarts))
else:
raise e
if len(self.optimization_runs):
i = np.argmin([o.f_opt for o in self.optimization_runs])
self.optimizer_array = self.optimization_runs[i].x_opt
else:
self.optimizer_array = initial_parameters
def ensure_default_constraints(self, warning=True):
"""
Ensure that any variables which should clearly be positive
have been constrained somehow. The method performs a regular
expression search on parameter names looking for the terms
'variance', 'lengthscale', 'precision' and 'kappa'. If any of
these terms are present in the name the parameter is
constrained positive.
DEPRECATED.
"""
raise DeprecationWarning, 'parameters now have default constraints'
def objective_function(self):
"""
The objective function for the given algorithm.
This function is the true objective, which wants to be minimized.
Note that all parameters are already set and in place, so you just need
to return the objective function here.
For probabilistic models this is the negative log_likelihood
(including the MAP prior), so we return it here. If your model is not
probabilistic, just return your objective to minimize here!
"""
return -float(self.log_likelihood()) - self.log_prior()
def objective_function_gradients(self):
"""
The gradients for the objective function for the given algorithm.
The gradients are w.r.t. the *negative* objective function, as
this framework works with *negative* log-likelihoods as a default.
You can find the gradient for the parameters in self.gradient at all times.
This is the place, where gradients get stored for parameters.
This function is the true objective, which wants to be minimized.
Note that all parameters are already set and in place, so you just need
to return the gradient here.
For probabilistic models this is the gradient of the negative log_likelihood
(including the MAP prior), so we return it here. If your model is not
probabilistic, just return your *negative* gradient here!
"""
return -(self._log_likelihood_gradients() + self._log_prior_gradients())
def _grads(self, x):
"""
Gets the gradients from the likelihood and the priors.
Failures are handled robustly. The algorithm will try several times to
return the gradients, and will raise the original exception if
the objective cannot be computed.
:param x: the parameters of the model.
:type x: np.array
"""
try:
# self._set_params_transformed(x)
self.optimizer_array = x
obj_grads = self._transform_gradients(self.objective_function_gradients())
self._fail_count = 0
except (LinAlgError, ZeroDivisionError, ValueError):
if self._fail_count >= self._allowed_failures:
raise
self._fail_count += 1
obj_grads = np.clip(self._transform_gradients(self.objective_function_gradients()), -1e100, 1e100)
return obj_grads
def _objective(self, x):
"""
The objective function passed to the optimizer. It combines
the likelihood and the priors.
Failures are handled robustly. The algorithm will try several times to
return the objective, and will raise the original exception if
the objective cannot be computed.
:param x: the parameters of the model.
:parameter type: np.array
"""
try:
self.optimizer_array = x
obj = self.objective_function()
self._fail_count = 0
except (LinAlgError, ZeroDivisionError, ValueError):
if self._fail_count >= self._allowed_failures:
raise
self._fail_count += 1
return np.inf
return obj
def _objective_grads(self, x):
try:
self.optimizer_array = x
obj_f, obj_grads = self.objective_function(), self._transform_gradients(self.objective_function_gradients())
self._fail_count = 0
except (LinAlgError, ZeroDivisionError, ValueError):
if self._fail_count >= self._allowed_failures:
raise
self._fail_count += 1
obj_f = np.inf
obj_grads = np.clip(self._transform_gradients(self.objective_function_gradients()), -1e100, 1e100)
return obj_f, obj_grads
def optimize(self, optimizer=None, start=None, **kwargs):
"""
Optimize the model using self.log_likelihood and self.log_likelihood_gradient, as well as self.priors.
kwargs are passed to the optimizer. They can be:
:param max_f_eval: maximum number of function evaluations
:type max_f_eval: int
:messages: whether to display during optimisation
:type messages: bool
:param optimizer: which optimizer to use (defaults to self.preferred optimizer)
:type optimizer: string
TODO: valid args
"""
if self.is_fixed:
raise RuntimeError, "Cannot optimize, when everything is fixed"
if self.size == 0:
raise RuntimeError, "Model without parameters cannot be optimized"
if start == None:
start = self.optimizer_array
if optimizer is None:
optimizer = self.preferred_optimizer
if isinstance(optimizer, optimization.Optimizer):
opt = optimizer
opt.model = self
else:
optimizer = optimization.get_optimizer(optimizer)
opt = optimizer(start, model=self, **kwargs)
opt.run(f_fp=self._objective_grads, f=self._objective, fp=self._grads)
self.optimization_runs.append(opt)
self.optimizer_array = opt.x_opt
def optimize_SGD(self, momentum=0.1, learning_rate=0.01, iterations=20, **kwargs):
# assert self.Y.shape[1] > 1, "SGD only works with D > 1"
sgd = SGD.StochasticGD(self, iterations, learning_rate, momentum, **kwargs) # @UndefinedVariable
sgd.run()
self.optimization_runs.append(sgd)
def _checkgrad(self, target_param=None, verbose=False, step=1e-6, tolerance=1e-3):
"""
Check the gradient of the ,odel by comparing to a numerical
estimate. If the verbose flag is passed, individual
components are tested (and printed)
:param verbose: If True, print a "full" checking of each parameter
:type verbose: bool
:param step: The size of the step around which to linearise the objective
:type step: float (default 1e-6)
:param tolerance: the tolerance allowed (see note)
:type tolerance: float (default 1e-3)
Note:-
The gradient is considered correct if the ratio of the analytical
and numerical gradients is within <tolerance> of unity.
"""
x = self.optimizer_array.copy()
if not verbose:
# make sure only to test the selected parameters
if target_param is None:
transformed_index = range(len(x))
else:
transformed_index = self._raveled_index_for(target_param)
if self._has_fixes():
indices = np.r_[:self.size]
which = (transformed_index[:, None] == indices[self._fixes_][None, :]).nonzero()
transformed_index = (indices - (~self._fixes_).cumsum())[transformed_index[which[0]]]
if transformed_index.size == 0:
print "No free parameters to check"
return
# just check the global ratio
dx = np.zeros(x.shape)
dx[transformed_index] = step * (np.sign(np.random.uniform(-1, 1, transformed_index.size)) if transformed_index.size != 2 else 1.)
# evaulate around the point x
f1 = self._objective(x + dx)
f2 = self._objective(x - dx)
gradient = self._grads(x)
dx = dx[transformed_index]
gradient = gradient[transformed_index]
denominator = (2 * np.dot(dx, gradient))
global_ratio = (f1 - f2) / np.where(denominator == 0., 1e-32, denominator)
global_diff = np.abs(f1 - f2) < tolerance and np.allclose(gradient, 0, atol=tolerance)
if global_ratio is np.nan:
global_ratio = 0
return np.abs(1. - global_ratio) < tolerance or global_diff
else:
# check the gradient of each parameter individually, and do some pretty printing
try:
names = self._get_param_names()
except NotImplementedError:
names = ['Variable %i' % i for i in range(len(x))]
# Prepare for pretty-printing
header = ['Name', 'Ratio', 'Difference', 'Analytical', 'Numerical']
max_names = max([len(names[i]) for i in range(len(names))] + [len(header[0])])
float_len = 10
cols = [max_names]
cols.extend([max(float_len, len(header[i])) for i in range(1, len(header))])
cols = np.array(cols) + 5
header_string = ["{h:^{col}}".format(h=header[i], col=cols[i]) for i in range(len(cols))]
header_string = map(lambda x: '|'.join(x), [header_string])
separator = '-' * len(header_string[0])
print '\n'.join([header_string[0], separator])
if target_param is None:
param_index = range(len(x))
transformed_index = param_index
else:
param_index = self._raveled_index_for(target_param)
if self._has_fixes():
indices = np.r_[:self.size]
which = (param_index[:, None] == indices[self._fixes_][None, :]).nonzero()
param_index = param_index[which[0]]
transformed_index = (indices - (~self._fixes_).cumsum())[param_index]
# print param_index, transformed_index
else:
transformed_index = param_index
if param_index.size == 0:
print "No free parameters to check"
return
gradient = self._grads(x).copy()
np.where(gradient == 0, 1e-312, gradient)
ret = True
for nind, xind in itertools.izip(param_index, transformed_index):
xx = x.copy()
xx[xind] += step
f1 = self._objective(xx)
xx[xind] -= 2.*step
f2 = self._objective(xx)
numerical_gradient = (f1 - f2) / (2 * step)
if np.all(gradient[xind] == 0): ratio = (f1 - f2) == gradient[xind]
else: ratio = (f1 - f2) / (2 * step * gradient[xind])
difference = np.abs(numerical_gradient - gradient[xind])
if (np.abs(1. - ratio) < tolerance) or np.abs(difference) < tolerance:
formatted_name = "\033[92m {0} \033[0m".format(names[nind])
ret &= True
else:
formatted_name = "\033[91m {0} \033[0m".format(names[nind])
ret &= False
r = '%.6f' % float(ratio)
d = '%.6f' % float(difference)
g = '%.6f' % gradient[xind]
ng = '%.6f' % float(numerical_gradient)
grad_string = "{0:<{c0}}|{1:^{c1}}|{2:^{c2}}|{3:^{c3}}|{4:^{c4}}".format(formatted_name, r, d, g, ng, c0=cols[0] + 9, c1=cols[1], c2=cols[2], c3=cols[3], c4=cols[4])
print grad_string
self.optimizer_array = x
return ret