Siamese recurrent networks learn first-order logic reasoning and exhibit zero-shot compositional generalization.
Mathijs MulWillem H. ZuidemaPublished in: CoRR (2019)
Keyphrases
- first order logic
- recurrent networks
- probabilistic reasoning
- knowledge representation
- expressive power
- logical rules
- proof procedure
- temporal knowledge
- theorem proving
- inference rules
- theorem prover
- anti unification
- automated reasoning
- propositional logic
- recurrent neural networks
- markov networks
- probabilistic graphical models
- transitive closure
- feed forward
- biologically inspired
- horn clauses
- predicate calculus
- knowledge base
- relational algebra
- constraint databases
- markov logic networks
- probabilistic logic
- inductive logic programming
- highly expressive
- quantifier elimination
- quantifier free
- reasoning tasks
- data complexity
- probabilistic model
- max margin
- knowledge representation and reasoning