Factorizing Declarative and Procedural Knowledge in Structured, Dynamical Environments
international conference on learning representations, 2020.
We explore separate factorization of procedural and declarative knowledge in modular recurrent neural networks.
Modeling a structured, dynamic environment like a video game requires keeping track of the objects and their states (declarative knowledge) as well as predicting how objects behave (procedural knowledge). Black-box models with a monolithic hidden state often fail to apply procedural knowledge consistently and uniformly, i.e., they lack sy...More
PPT (Upload PPT)