Neural Datalog Through Time: Informed Temporal Modeling via Logical Specification
This addresses the challenge of overfitting in temporal event prediction for domains with large event sets, though it appears incremental as it combines existing logical and neural approaches.
The paper tackles the problem of predicting future events from past patterns when event types are numerous, by proposing neural probabilistic models derived from Datalog programs that encode domain knowledge. The result shows improved prediction in synthetic and real-world domains.
Learning how to predict future events from patterns of past events is difficult when the set of possible event types is large. Training an unrestricted neural model might overfit to spurious patterns. To exploit domain-specific knowledge of how past events might affect an event's present probability, we propose using a temporal deductive database to track structured facts over time. Rules serve to prove facts from other facts and from past events. Each fact has a time-varying state---a vector computed by a neural net whose topology is determined by the fact's provenance, including its experience of past events. The possible event types at any time are given by special facts, whose probabilities are neurally modeled alongside their states. In both synthetic and real-world domains, we show that neural probabilistic models derived from concise Datalog programs improve prediction by encoding appropriate domain knowledge in their architecture.