Hybrid Neural/Symbolic Models (11/15/2018)
Content:
- Incorporating Neural Features in Symbolic Models
- Incorporating Symbolic Features in Neural Models
- Extracting Symbolic Models from Neural Models
- Reading Material: Neural-symbolic Models
In class I'll be mainly talking about connections between RNNs and automata. The following will not be on the quiz, but are here for reference:
- Reference: Extraction of Rules from Discrete-time Recurrent Neural Networks
- Reference: Extracting Automata from Recurrent Neural Networks Using Queries and Counterexamples
- Reference: Recurrent Neural Networks as Weighted Language Recognizers
- Reference: On the Practical Computational Power of Finite Precision RNNs for Language Recognition
- Reference: SoPa: Bridging CNNs, RNNs, and Weighted Finite-State Machines
- Reference: Rational Recurrences