From 6a40d1034c5ca6bad43d2bc9d31170a09925685f Mon Sep 17 00:00:00 2001
From: James Thornton <jtthornton1994@gmail.com>
Date: Thu, 27 Oct 2022 11:16:55 +0200
Subject: [PATCH] Delete apf.py

---
 filterflow/apf.py | 52 -----------------------------------------------
 1 file changed, 52 deletions(-)
 delete mode 100644 filterflow/apf.py

diff --git a/filterflow/apf.py b/filterflow/apf.py
deleted file mode 100644
index 0269000..0000000
--- a/filterflow/apf.py
+++ /dev/null
@@ -1,52 +0,0 @@
-import attr
-import tensorflow as tf
-
-from filterflow.base import State
-from filterflow.constants import MIN_RELATIVE_LOG_WEIGHT
-from filterflow.observation.base import ObservationModelBase
-from filterflow.proposal import ProposalModelBase
-from filterflow.resampling.base import ResamplerBase
-from filterflow.resampling.criterion import ResamplingCriterionBase
-from filterflow.smc import SMC
-from filterflow.transition.base import TransitionModelBase
-from filterflow.utils import normalize
-
-
-class AuxiliaryParticleFilter(SMC):
-    def __init__(self, observation_model: ObservationModelBase, transition_model: TransitionModelBase,
-                 proposal_model: ProposalModelBase, resampling_criterion: ResamplingCriterionBase,
-                 resampling_method: ResamplerBase, name='AuxiliaryParticleFilter'):
-        super(AuxiliaryParticleFilter, self).__init__(observation_model, transition_model,
-                                                      proposal_model, resampling_criterion, resampling_method,
-                                                      name=name)
-        raise NotImplementedError("This is not yet implemented")
-
-    @tf.function
-    def propose_and_weight(self, state: State, observation: tf.Tensor,
-                           inputs: tf.Tensor, seed=None):
-        """
-        :param state: State
-            current state of the filter
-        :param observation: tf.Tensor
-            observation to compare the state against
-        :param inputs: tf.Tensor
-            inputs for the observation_model
-        :return: Updated weights
-        """
-        float_n_particles = tf.cast(state.n_particles, float)
-        proposed_state = self._proposal_model.propose(state, inputs, observation)
-
-        log_weights = self._transition_model.loglikelihood(state, proposed_state, inputs)
-        log_weights = log_weights + self._observation_model.loglikelihood(proposed_state, observation)
-        log_weights = log_weights - self._proposal_model.loglikelihood(proposed_state, state, inputs, observation)
-        log_weights = log_weights + state.log_weights
-
-        log_likelihood_increment = tf.math.reduce_logsumexp(log_weights, 1)
-        log_likelihoods = state.log_likelihoods + log_likelihood_increment
-        normalized_log_weights = normalize(log_weights, 1, True)
-        normalized_log_weights = tf.clip_by_value(normalized_log_weights,
-                                                  MIN_RELATIVE_LOG_WEIGHT * float_n_particles,
-                                                  tf.constant(float('inf')))
-        normalized_log_weights = normalize(normalized_log_weights, 1, True)
-        return attr.evolve(proposed_state, weights=tf.math.exp(normalized_log_weights),
-                           log_weights=normalized_log_weights, log_likelihoods=log_likelihoods)