42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174 | @dataclass
class ProbLogSolver(Solver):
"""
A solver that uses problog.
[problog](https://dtai.cs.kuleuven.be/problog/index.html) is a
probabilistic logic programming language.
Example:
>>> from typedlogic.integrations.frameworks.pydantic import FactBaseModel
>>> from typedlogic import SentenceGroup, PredicateDefinition, Forall, Variable
>>> solver = ProbLogSolver()
>>> solver.add_predicate_definition(PredicateDefinition(predicate="AncestorOf", arguments={'ancestor': 'str', 'descendant': 'str'}))
>>> x = Variable('x')
>>> y = Variable('y')
>>> z = Variable('z')
>>> tr_axiom = Forall([x, y, z], (Term('AncestorOf', x, y) & Term('AncestorOf', y, z)) >> Term('AncestorOf', x, z))
>>> solver.add_sentence(tr_axiom)
>>> solver.add_probabilistic_fact(Term('AncestorOf', 'p1', 'p1a'), 0.5)
>>> solver.add_probabilistic_fact(Term('AncestorOf', 'p1a', 'p1aa'), 0.5)
>>> model = solver.model()
>>> print(model.term_probabilities[Term('AncestorOf', 'p1', 'p1aa')])
0.25
"""
exec_name: str = field(default="problog")
profile: ClassVar[Profile] = MixedProfile(Probabilistic(), AllowsComparisonTerms(), MultipleModelSemantics())
problog_args: Optional[Dict[str, Any]] = None
def models(self, **kwargs) -> Iterator[ProbabilisticModel]:
compiler = ProbLogCompiler()
program = compiler.compile(self.base_theory)
result = self._evaluate_program(program, **kwargs)
m = ProbabilisticModel()
for term, prob in result.items():
plt_term = compiler.decompile_term(term)
reified = Probability(prob, That(plt_term))
reified_as_term = reified.to_model_object()
if not isinstance(reified_as_term, Term):
raise ValueError(f"Expected a term, got {reified_as_term}")
m.ground_terms.append(reified_as_term)
m.term_probabilities[plt_term] = prob
yield m
def _evaluate_program(self, program: str, **kwargs) -> Dict[Any, float]:
"""Evaluate a ProbLog program and return query probabilities."""
for k, v in DEFAULT_PROBLOG_ARGS.items():
if k not in kwargs:
kwargs[k] = v
if self.problog_args:
for k, v in self.problog_args.items():
if k not in kwargs:
kwargs[k] = v
try:
return get_evaluatable().create_from(PrologString(program), **kwargs).evaluate()
except InconsistentEvidenceError as e:
raise UnsatisfiableEvidenceError(
"ProbLog evidence is inconsistent; no probabilistic model can satisfy the asserted evidence. "
f"ProbLog reported: {e}"
) from e
def model(self) -> ProbabilisticModel:
models = list(self.models())
if len(models) == 0:
raise UnsatisfiableEvidenceError(
"ProbLog produced no models; the evidence may be unsatisfiable for this probabilistic theory."
)
if len(models) > 1:
raise AmbiguousModelError(
"ProbLog produced multiple models; ProbLogSolver.model() expects exactly one model. "
"Use ProbLogSolver.models() to inspect all models."
)
return models[0]
def check(self) -> Solution:
try:
models = list(self.models())
except UnsatisfiableEvidenceError:
return Solution(satisfiable=False)
sat = len(models) > 0
return Solution(satisfiable=sat)
def dump(self) -> str:
compiler = ProbLogCompiler()
return compiler.compile(self.base_theory)
def prove(self, sentence: Sentence) -> Optional[bool]:
"""Prove Datalog-safe implications by checking counterexample probability."""
compiler = ProbLogCompiler()
try:
counterexample_sentences = counterexample_proof_sentences(
sentence,
predicate=COUNTEREXAMPLE_PREDICATE,
)
except NotInProfileError:
return None
program = compiler.compile(self.base_theory, include_queries=False)
proof_clauses: list[str] = []
prolog_config = compiler.prolog_config()
try:
for proof_sentence in counterexample_sentences:
for rule in to_horn_rules(proof_sentence, allow_disjunctions_in_head=False, allow_goal_clauses=True):
proof_clauses.append(as_prolog(rule, config=prolog_config))
query = Term("query", Term(COUNTEREXAMPLE_PREDICATE))
for rule in to_horn_rules(query, allow_disjunctions_in_head=False, allow_goal_clauses=True):
proof_clauses.append(as_prolog(rule, config=prolog_config))
except NotInProfileError as e:
logger.info(f"Cannot prove sentence {sentence} with counterexample transform due to {e}")
return None
if proof_clauses:
program = f"{program}\n" + "\n".join(proof_clauses)
try:
result = self._evaluate_program(program)
except UnsatisfiableEvidenceError:
return None
counterexample = Term(COUNTEREXAMPLE_PREDICATE)
for term, probability in result.items():
if compiler.decompile_term(term) == counterexample:
return probability == 0
return None
def add_probabilistic_fact(self, fact: Sentence, probability: float) -> None:
pr_sent = Probability(probability, That(fact))
self.add(pr_sent)
def add_evidence(self, fact: Sentence, truth_value: bool) -> None:
ev = Evidence(fact, truth_value)
self.add(ev)
|
