"""The main abstract classes."""
import logging
from abc import ABCMeta, abstractmethod
from typing import Set, List, Tuple, Iterable, TypeVar, Generic, ClassVar, Optional
from owlapy.class_expression import OWLClassExpression
from owlapy.owl_ontology import OWLOntology
from owlapy.util import iter_count
from .data_struct import Experience
from .utils import read_csv
from collections import OrderedDict
_N = TypeVar('_N') #:
_KB = TypeVar('_KB', bound='AbstractKnowledgeBase') #:
logger = logging.getLogger(__name__)
# @TODO:CD: Each Class definiton in abstract.py should share a prefix, e.g., BaseX or AbstractX.
# @TODO:CD: All imports must be located on top of the script
from owlapy import owl_expression_to_dl
[docs]
class EncodedLearningProblem(metaclass=ABCMeta):
"""Encoded Abstract learning problem for use in Scorers."""
__slots__ = ()
[docs]
class EncodedPosNegLPStandardKind(EncodedLearningProblem, metaclass=ABCMeta):
"""Encoded Abstract learning problem following pos-neg lp standard."""
__slots__ = ()
# @TODO: Why we need Generic[_N] and if we need it why we di not use it in all other abstract classes?
[docs]
class AbstractScorer(Generic[_N], metaclass=ABCMeta):
"""
An abstract class for quality functions.
"""
__slots__ = ()
name: ClassVar[str]
def __init__(self, *args, **kwargs):
"""Create a new quality function."""
pass
[docs]
def score_elp(self, instances: set, learning_problem: EncodedLearningProblem) -> Tuple[bool, Optional[float]]:
"""Quality score for a set of instances with regard to the learning problem.
Args:
instances (set): Instances to calculate a quality score for.
learning_problem: Underlying learning problem to compare the quality to.
Returns:
Tuple, first position indicating if the function could be applied, second position the quality value
in the range 0.0--1.0.
"""
if len(instances) == 0:
return False, 0
# @TODO: It must be moved to the top of the abstracts.py
from ontolearn.learning_problem import EncodedPosNegLPStandard
if isinstance(learning_problem, EncodedPosNegLPStandard):
tp = len(learning_problem.kb_pos.intersection(instances))
tn = len(learning_problem.kb_neg.difference(instances))
fp = len(learning_problem.kb_neg.intersection(instances))
fn = len(learning_problem.kb_pos.difference(instances))
return self.score2(tp=tp, tn=tn, fp=fp, fn=fn)
else:
raise NotImplementedError(learning_problem)
[docs]
@abstractmethod
def score2(self, tp: int, fn: int, fp: int, tn: int) -> Tuple[bool, Optional[float]]:
"""Quality score for a coverage count.
Args:
tp: True positive count.
fn: False negative count.
fp: False positive count.
tn: True negative count.
Returns:
Tuple, first position indicating if the function could be applied, second position the quality value
in the range 0.0--1.0.
"""
pass
# @TODO:CD: Why there is '..' in AbstractNode
[docs]
def apply(self, node: 'AbstractNode', instances, learning_problem: EncodedLearningProblem) -> bool:
"""Apply the quality function to a search tree node after calculating the quality score on the given instances.
Args:
node: search tree node to set the quality on.
instances (set): Instances to calculate the quality for.
learning_problem: Underlying learning problem to compare the quality to.
Returns:
True if the quality function was applied successfully
"""
assert isinstance(learning_problem, EncodedLearningProblem), \
f'Expected EncodedLearningProblem but got {type(learning_problem)}'
assert isinstance(node, AbstractNode), \
f'Expected AbstractNode but got {type(node)}'
# @TODO: It must be moved to the top of the abstracts.py
from ontolearn.search import _NodeQuality
assert isinstance(node, _NodeQuality), \
f'Expected _NodeQuality but got {type(_NodeQuality)}'
ret, q = self.score_elp(instances, learning_problem)
if q is not None:
node.quality = q
return ret
[docs]
class AbstractHeuristic(Generic[_N], metaclass=ABCMeta):
"""Abstract base class for heuristic functions.
Heuristic functions can guide the search process."""
__slots__ = ()
@abstractmethod
def __init__(self):
"""Create a new heuristic function."""
pass
[docs]
@abstractmethod
def apply(self, node: _N, instances, learning_problem: EncodedLearningProblem):
"""Apply the heuristic on a search tree node and set its heuristic property to the calculated value.
Args:
node: Node to set the heuristic on.
instances (set, optional): Set of instances covered by this node.
learning_problem: Underlying learning problem to compare the heuristic to.
"""
pass
[docs]
class AbstractFitness(metaclass=ABCMeta):
"""Abstract base class for fitness functions.
Fitness functions guide the evolutionary process."""
__slots__ = ()
name: ClassVar[str]
@abstractmethod
def __init__(self):
"""Create a new fitness function."""
pass
[docs]
@abstractmethod
def apply(self, individual):
"""Apply the fitness function on an individual and set its fitness attribute to the calculated value.
Args:
individual: Individual to set the fitness on.
"""
pass
[docs]
class BaseRefinement(Generic[_N], metaclass=ABCMeta):
"""
Base class for Refinement Operators.
Let C, D \\in N_c where N_c os a finite set of concepts.
* Proposition 3.3 (Complete and Finite Refinement Operators) [1]
* ρ(C) = {C ⊓ T} ∪ {D \\| D is not empty AND D \\sqset C}
* The operator is finite,
* The operator is complete as given a concept C, we can reach an arbitrary concept D such that D subset of C.
*) Theoretical Foundations of Refinement Operators [1].
*) Defining a top-down refimenent operator that is a proper is crutial.
4.1.3 Achieving Properness [1]
*) Figure 4.1 [1] defines of the refinement operator.
[1] Learning OWL Class Expressions.
Attributes:
kb (AbstractKnowledgeBase): The knowledge base used by this refinement operator.
"""
__slots__ = 'kb'
kb: _KB
@abstractmethod
def __init__(self, knowledge_base: _KB):
"""Construct a new base refinement operator.
Args:
knowledge_base: Knowledge base to operate on.
"""
self.kb = knowledge_base
[docs]
@abstractmethod
def refine(self, *args, **kwargs) -> Iterable[OWLClassExpression]:
"""Refine a given concept.
Args:
ce (OWLClassExpression): Concept to refine.
Returns:
New refined concepts.
"""
pass
[docs]
def len(self, concept: OWLClassExpression) -> int:
"""The length of a concept.
Args:
concept: The concept to measure the length for.
Returns:
Length of concept according to some metric configured in the knowledge base.
"""
return self.kb.concept_len(concept)
[docs]
class AbstractNode(metaclass=ABCMeta):
"""Abstract search tree node."""
__slots__ = ()
@abstractmethod
def __init__(self):
"""Create an abstract search tree node."""
pass
[docs]
def __str__(self):
"""String representation of node, by default its internal memory address."""
addr = hex(id(self))
addr = addr[0:2] + addr[6:-1]
return f'{type(self)} at {addr}'
[docs]
def __repr__(self):
return self.__str__()
[docs]
class AbstractOEHeuristicNode(metaclass=ABCMeta):
"""Abstract Node for the CELOEHeuristic heuristic function.
This node must support quality, horizontal expansion (h_exp), is_root, parent_node and refinement_count.
"""
__slots__ = ()
@property
@abstractmethod
def quality(self) -> Optional[float]:
"""Get the quality of the node.
Returns:
Quality of the node.
"""
pass
@property
@abstractmethod
def h_exp(self) -> int:
"""Get horizontal expansion.
Returns:
Horizontal expansion.
"""
pass
@property
@abstractmethod
def is_root(self) -> bool:
"""Is this the root node?
Returns:
True if this is the root node, otherwise False.
"""
pass
@property
@abstractmethod
def parent_node(self: _N) -> Optional[_N]:
"""Get the parent node.
Returns:
Parent node.
"""
pass
@property
@abstractmethod
def refinement_count(self) -> int:
"""Get the refinement count for this node.
Returns:
Refinement count.
"""
pass
@property
@abstractmethod
def heuristic(self) -> Optional[float]:
"""Get the heuristic value.
Returns:
Heuristic value.
"""
pass
@heuristic.setter
@abstractmethod
def heuristic(self, v: float):
"""Set the heuristic value."""
pass
[docs]
class AbstractConceptNode(metaclass=ABCMeta):
"""Abstract search tree node which has a concept."""
__slots__ = ()
@property
@abstractmethod
def concept(self) -> OWLClassExpression:
"""Get the concept representing this node.
Returns:
The concept representing this node.
"""
pass
[docs]
class AbstractKnowledgeBase(metaclass=ABCMeta):
"""Abstract knowledge base."""
__slots__ = ()
# CD: This function is used as "a get method". Insteadf either access the atttribute directly
# or use it as a property @abstractmethod
[docs]
def ontology(self) -> OWLOntology:
"""The base ontology of this knowledge base."""
pass
[docs]
def describe(self) -> None:
"""Print a short description of the Knowledge Base to the info logger output."""
properties_count = iter_count(self.ontology.object_properties_in_signature()) + iter_count(
self.ontology.data_properties_in_signature())
logger.info(f'Number of named classes: {iter_count(self.ontology.classes_in_signature())}\n'
f'Number of individuals: {self.individuals_count()}\n'
f'Number of properties: {properties_count}')
[docs]
@abstractmethod
def clean(self) -> None:
"""This method should reset any caches and statistics in the knowledge base."""
raise NotImplementedError
[docs]
@abstractmethod
def individuals_count(self) -> int:
"""Total number of individuals in this knowledge base."""
pass
[docs]
@abstractmethod
def individuals_set(self, *args, **kwargs) -> Set:
"""Encode an individual, an iterable of individuals or the individuals that are instances of a given concept
into a set.
Args:
arg (OWLNamedIndividual): Individual to encode.
arg (Iterable[OWLNamedIndividual]): Individuals to encode.
arg (OWLClassExpression): Encode individuals that are instances of this concept.
Returns:
Encoded set representation of individual(s).
"""
pass
[docs]
@abstractmethod
def concept_len(self, ce: OWLClassExpression) -> int:
"""Calculate the length of a concept.
Args:
ce: The concept to measure the length for.
Returns:
Length of concept.
"""
pass
[docs]
class AbstractLearningProblem(metaclass=ABCMeta):
"""Abstract learning problem."""
__slots__ = ()
@abstractmethod
def __init__(self, *args, **kwargs):
"""Create a new abstract learning problem."""
pass
[docs]
@abstractmethod
def encode_kb(self, knowledge_base: AbstractKnowledgeBase) -> 'EncodedLearningProblem':
"""Encode the learning problem into the knowledge base."""
pass
[docs]
class LBLSearchTree(Generic[_N], metaclass=ABCMeta):
"""Abstract search tree for the Length based learner."""
[docs]
@abstractmethod
def get_most_promising(self) -> _N:
"""Find most "promising" node in the search tree that should be refined next.
Returns:
Most promising search tree node.
"""
pass
[docs]
@abstractmethod
def add_node(self, node: _N, parent_node: _N, kb_learning_problem: EncodedLearningProblem):
"""Add a node to the search tree.
Args:
node: Node to add.
parent_node: Parent of that node.
kb_learning_problem: Underlying learning problem to compare the quality to.
"""
pass
[docs]
@abstractmethod
def clean(self):
"""Reset the search tree state."""
pass
[docs]
@abstractmethod
def get_top_n(self, n: int) -> List[_N]:
"""Retrieve the best n search tree nodes.
Args:
n: Maximum number of nodes.
Returns:
List of top n search tree nodes.
"""
pass
[docs]
@abstractmethod
def show_search_tree(self, root_concept: OWLClassExpression, heading_step: str):
"""Debugging function to print the search tree to standard output.
Args:
root_concept: The tree is printed starting from this search tree node.
heading_step: Message to print at top of the output.
"""
pass
[docs]
@abstractmethod
def add_root(self, node: _N, kb_learning_problem: EncodedLearningProblem):
"""Add the root node to the search tree.
Args:
node: Root node to add.
kb_learning_problem: Underlying learning problem to compare the quality to.
"""
pass
[docs]
class DepthAbstractDrill:
"""
Abstract class for Convolutional DQL concept learning.
"""
def __init__(self, path_of_embeddings, reward_func, learning_rate=None,
num_episode=None, num_episodes_per_replay=None, epsilon=None,
num_of_sequential_actions=None, max_len_replay_memory=None,
representation_mode=None, batch_size=None, epsilon_decay=None, epsilon_min=None,
num_epochs_per_replay=None, num_workers=None, verbose=0):
self.name = 'DRILL'
self.instance_embeddings = read_csv(path_of_embeddings)
if not self.instance_embeddings:
print("No embeddings found")
self.embedding_dim = None
else:
self.embedding_dim = self.instance_embeddings.shape[1]
self.reward_func = reward_func
self.representation_mode = representation_mode
assert representation_mode in ['averaging', 'sampling']
# Will be filled by child class
self.heuristic_func = None
self.num_workers = num_workers
# constants
self.epsilon = epsilon
self.learning_rate = learning_rate
self.num_episode = num_episode
self.num_of_sequential_actions = num_of_sequential_actions
self.num_epochs_per_replay = num_epochs_per_replay
self.max_len_replay_memory = max_len_replay_memory
self.epsilon_decay = epsilon_decay
self.epsilon_min = epsilon_min
self.batch_size = batch_size
self.verbose = verbose
self.num_episodes_per_replay = num_episodes_per_replay
# will be filled
self.optimizer = None # torch.optim.Adam(self.model_net.parameters(), lr=self.learning_rate)
self.seen_examples = dict()
self.emb_pos, self.emb_neg = None, None
self.start_time = None
self.goal_found = False
self.experiences = Experience(maxlen=self.max_len_replay_memory)
[docs]
def attributes_sanity_checking_rl(self):
assert len(self.instance_embeddings) > 0
assert self.embedding_dim > 0
if self.num_workers is None:
self.num_workers = 4
if self.epsilon is None:
self.epsilon = 1
if self.learning_rate is None:
self.learning_rate = .001
if self.num_episode is None:
self.num_episode = 1
if self.num_of_sequential_actions is None:
self.num_of_sequential_actions = 3
if self.num_epochs_per_replay is None:
self.num_epochs_per_replay = 1
if self.max_len_replay_memory is None:
self.max_len_replay_memory = 256
if self.epsilon_decay is None:
self.epsilon_decay = 0.01
if self.epsilon_min is None:
self.epsilon_min = 0
if self.batch_size is None:
self.batch_size = 1024
if self.verbose is None:
self.verbose = 0
if self.num_episodes_per_replay is None:
self.num_episodes_per_replay = 2
[docs]
@abstractmethod
def init_training(self, *args, **kwargs):
"""
Initialize training for a given E+,E- and K.
"""
[docs]
@abstractmethod
def terminate_training(self):
"""
Save weights and training data after training phase.
"""
[docs]
class DRILLAbstractTree:
"""Abstract Tree for DRILL."""
@abstractmethod
def __init__(self):
self._nodes = dict()
[docs]
def __len__(self):
return len(self._nodes)
[docs]
def __getitem__(self, item):
return self._nodes[item]
[docs]
def __setitem__(self, k, v):
self._nodes[k] = v
[docs]
def __iter__(self):
for k, node in self._nodes.items():
yield node
[docs]
def get_top_n_nodes(self, n: int, key='quality'):
self.sort_search_tree_by_decreasing_order(key=key)
for ith, dict_ in enumerate(self._nodes.items()):
if ith >= n:
break
k, node = dict_
yield node
[docs]
def redundancy_check(self, n):
if n in self._nodes:
return False
return True
@property
def nodes(self):
return self._nodes
[docs]
@abstractmethod
def add(self, *args, **kwargs):
pass
[docs]
def sort_search_tree_by_decreasing_order(self, *, key: str):
if key == 'heuristic':
sorted_x = sorted(self._nodes.items(), key=lambda kv: kv[1].heuristic, reverse=True)
elif key == 'quality':
sorted_x = sorted(self._nodes.items(), key=lambda kv: kv[1].quality, reverse=True)
elif key == 'length':
sorted_x = sorted(self._nodes.items(), key=lambda kv: len(kv[1]), reverse=True)
else:
raise ValueError('Wrong Key. Key must be heuristic, quality or concept_length')
self._nodes = OrderedDict(sorted_x)
[docs]
def best_hypotheses(self, n=10) -> List:
assert self.search_tree is not None
assert len(self.search_tree) > 1
return [i for i in self.search_tree.get_top_n_nodes(n)]
[docs]
def show_search_tree(self, top_n=100):
"""
Show search tree.
"""
predictions = list(self.get_top_n_nodes(top_n))
print('######## Search Tree ###########\n')
for ith, node in enumerate(predictions):
print(f"{ith + 1}-\t{owl_expression_to_dl(node.concept)} | Quality:{node.quality}| Heuristic:{node.heuristic}")
print('\n######## Search Tree ###########\n')
return predictions
[docs]
def show_best_nodes(self, top_n, key=None):
assert key
self.sort_search_tree_by_decreasing_order(key=key)
return self.show_search_tree('Final', top_n=top_n + 1)
[docs]
@staticmethod
def save_current_top_n_nodes(key=None, n=10, path=None):
"""
Save current top_n nodes.
"""
assert path
assert key
assert isinstance(n, int)
pass
[docs]
def clean(self):
self._nodes.clear()