arXiv AI recent: DeepSWIP: Quotient-WMC Counterfactuals for Neural Probabilistic Logic Programs
Researchers introduced DeepSWIP, a single-world counterfactual semantics for DeepProbLog programs.,DeepSWIP uses neural materialization to reduce fixed-context neural predicates to ordina...
DeepSWIP is a neurosymbolic system that combines neural perception with probabilistic logic, and it requires a causal semantics for interventions and evidence.,The system is exact relative to the learned materialized FCM under finite grounding and unique-supported-model assumptions, and it identi...