Source code for bayespy.inference.vmp.nodes.stochastic

################################################################################
# Copyright (C) 2013-2014 Jaakko Luttinen
#
# This file is licensed under the MIT License.
################################################################################


import numpy as np

from bayespy.utils import misc

import h5py

from .node import Node

[docs]class Distribution(): """ A base class for the VMP formulas of variables. Sub-classes implement distribution specific computations. If a sub-class maps the plates differently, it needs to overload the following methods: * compute_weights_to_parent * plates_to_parent * plates_from_parent """
[docs] def compute_message_to_parent(self, parent, index, u_self, *u_parents): """ Compute the message to a parent node. """ raise NotImplementedError()
[docs] def compute_weights_to_parent(self, index, weights): """ Maps the mask to the plates of a parent. """ # Sub-classes may need to overwrite this method return weights
[docs] def plates_to_parent(self, index, plates): """ Resolves the plate mapping to a parent. Given the plates of the node's moments, this method returns the plates that the message to a parent has for the parent's distribution. """ return plates
[docs] def plates_from_parent(self, index, plates): """ Resolve the plate mapping from a parent. Given the plates of a parent's moments, this method returns the plates that the moments has for this distribution. """ return plates
[docs] def random(self, *params, plates=None): """ Draw a random sample from the distribution. """ raise NotImplementedError()
[docs] def squeeze(self, axis): """Squeeze a plate axis from the distribution The default implementation does no changes to the distribution. Override if needed. """ return self
[docs]class Stochastic(Node): """ Base class for nodes that are stochastic. u observed Sub-classes must implement: _compute_message_to_parent(parent, index, u_self, *u_parents) _update_distribution_and_lowerbound(self, m, *u) lowerbound(self) _compute_dims initialize_from_prior() If you want to be able to observe the variable: _compute_fixed_moments_and_f Sub-classes may need to re-implement: 1. If they manipulate plates: _compute_weights_to_parent(index, weights) _compute_plates_to_parent(self, index, plates) _compute_plates_from_parent(self, index, plates) """ # Sub-classes must over-write this _distribution = None
[docs] def __init__(self, *args, initialize=True, dims=None, **kwargs): self._id = Node._id_counter Node._id_counter += 1 super().__init__(*args, dims=dims, **kwargs) # Initialize moment array axes = len(self.plates)*(1,) self.u = [misc.nans(axes+dim) for dim in dims] # Not observed self.observed = False self.ndims = [len(dim) for dim in self.dims] if initialize: self.initialize_from_prior()
[docs] def get_pdf_nodes(self): return (self,) + super().get_pdf_nodes()
def _get_pdf_nodes_conditioned_on_parents(self): return (self,) def _get_id_list(self): """ Returns the stochastic ID list. This method is used to check that same stochastic nodes are not direct parents of a node several times. It is only valid if there are intermediate stochastic nodes. To put it another way: each ID corresponds to one factor q(..) in the posterior approximation. Different IDs mean different factors, thus they mean independence. The parents must have independent factors. Stochastic nodes should return their unique ID. Deterministic nodes should return the IDs of their parents. Constant nodes should return empty list of IDs. """ return [self._id] def _compute_plates_to_parent(self, index, plates): return self._distribution.plates_to_parent(index, plates) def _compute_plates_from_parent(self, index, plates): return self._distribution.plates_from_parent(index, plates) def _compute_weights_to_parent(self, index, weights): return self._distribution.compute_weights_to_parent(index, weights)
[docs] def get_moments(self): # Just for safety, do not return a reference to the moment list of this # node but instead create a copy of the list. return [ui for ui in self.u]
def _get_message_and_mask_to_parent(self, index, u_parent=None): u_parents = self._message_from_parents(exclude=index) u_parents[index] = u_parent m = self._distribution.compute_message_to_parent(self.parents[index], index, self.u, *u_parents) mask = self._distribution.compute_weights_to_parent(index, self.mask) != 0 return (m, mask) def _set_mask(self, mask): self.mask = np.logical_or(mask, self.observed) def _check_shape(self, u, broadcast=True): if len(u) != len(self.dims): raise ValueError("Incorrect number of arrays") for (dimsi, ui) in zip(self.dims, u): sh_true = self.plates + dimsi sh = np.shape(ui) ndim = len(dimsi) errmsg = ( "Shape of the given array not equal to the shape of the node.\n" "Received shape: {0}\n" "Expected shape: {1}\n" "Check plates." .format(sh, sh_true) ) if not broadcast: if sh != sh_true: raise ValueError(errmsg) else: if ndim == 0: if not misc.is_shape_subset(sh, sh_true): raise ValueError(errmsg) else: plates_ok = misc.is_shape_subset(sh[:-ndim], self.plates) dims_ok = (sh[-ndim:] == dimsi) if not (plates_ok and dims_ok): raise ValueError(errmsg) return def _set_moments(self, u, mask=True, broadcast=True): self._check_shape(u, broadcast=broadcast) # Store the computed moments u but do not change moments for # observations, i.e., utilize the mask. for ind in range(len(u)): # Add axes to the mask for the variable dimensions (mask # contains only axes for the plates). u_mask = misc.add_trailing_axes(mask, self.ndims[ind]) # Enlarge self.u[ind] as necessary so that it can store the # broadcasted result. sh = misc.broadcasted_shape_from_arrays(self.u[ind], u[ind], u_mask) self.u[ind] = misc.repeat_to_shape(self.u[ind], sh) # TODO/FIXME/BUG: The mask of observations is not used, observations # may be overwritten!!! ??? # Hah, this function is used to set the observations! The caller # should be careful what mask he uses! If you want to set only # latent variables, then use such a mask. # Use mask to update only unobserved plates and keep the # observed as before np.copyto(self.u[ind], u[ind], where=u_mask) # Make sure u has the correct number of dimensions: shape = self.get_shape(ind) ndim = len(shape) ndim_u = np.ndim(self.u[ind]) if ndim > ndim_u: self.u[ind] = misc.add_leading_axes(u[ind], ndim - ndim_u) elif ndim < ndim_u: # This should not ever happen because we already checked the # shape at the beginning of the function. raise RuntimeError( "This error should not happen. Fix shape checking." "The size of the variable %s's %s-th moment " "array is %s which is larger than it should " "be, that is, %s, based on the plates %s and " "dimension %s. Check that you have provided " "plates properly." % (self.name, ind, np.shape(self.u[ind]), shape, self.plates, self.dims[ind]))
[docs] def update(self, annealing=1.0): if not np.all(self.observed): u_parents = self._message_from_parents() m_children = self._message_from_children() if annealing != 1.0: m_children = [annealing * m for m in m_children] self._update_distribution_and_lowerbound(m_children, *u_parents)
[docs] def observe(self, x, mask=True): """ Fix moments, compute f and propagate mask. """ raise NotImplementedError()
[docs] def unobserve(self): # Update mask self.observed = False self._update_mask()
[docs] def lowerbound(self): # Sub-class should implement this raise NotImplementedError()
def _update_distribution_and_lowerbound(self, m_children, *u_parents): # Sub-classes should implement this raise NotImplementedError()
[docs] def save(self, filename): # Open HDF5 file h5f = h5py.File(filename, 'w') try: # Write each node nodegroup = h5f.create_group('nodes') if self.name == '': raise ValueError("In order to save nodes, they must have " "(unique) names.") self._save(nodegroup.create_group(self.name)) finally: # Close file h5f.close()
def _save(self, group): """ Save the state of the node into a HDF5 file. group can be the root """ for i in range(len(self.u)): misc.write_to_hdf5(group, self.u[i], 'u%d' % i) misc.write_to_hdf5(group, self.observed, 'observed') return
[docs] def load(self, filename): h5f = h5py.File(filename, 'r') try: self._load(h5f['nodes'][self.name]) finally: h5f.close() return
def _load(self, group): """ Load the state of the node from a HDF5 file. """ # TODO/FIXME: Check that the shapes are correct! for i in range(len(self.u)): ui = group['u%d' % i][...] self.u[i] = ui old_observed = self.observed self.observed = group['observed'][...] # Update masks if necessary if np.any(old_observed != self.observed): self._update_mask()
[docs] def random(self): """ Draw a random sample from the distribution. """ raise NotImplementedError()
[docs] def show(self): """ Print the distribution using standard parameterization. """ print(str(self))
def __str__(self): """ """ raise NotImplementedError("String representation not yet implemented for " "node class %s" % (self.__class__.__name__))