A Connectionist Symbol Manipulator that Discovers the Structure of Context-Free Languages

By Michael C. Mozer, Sreerupa Das. 1992

Subfields: Natural Language Processing, Representation & Self-Supervised Learning

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Natural Language Processing

In Representation & Self-Supervised Learning

Citations in the corpus

Inferring Algorithmic Patterns with Stack-Augmented Recurrent Nets (2015, methodology): “Also, as in (Das et al., 1993; Mozer & Das, 1993), their model is train on supervised data, while we aim at learning on the more challenging task of sequence prediction.”
On the Binding Problem in Artificial Neural Networks (2020, background): “They are commonly found in RNNs and, when paired with an attention mechanism that attends to a different object at each step, can serve as object representations.”
Object Files and Schemata: Factorizing Declarative and Procedural Knowledge in Dynamical Systems (2020, methodology): “entations. Modularity is the guiding principle of the model we propose, which we call SCOFF, an acronym for schema / object-ﬁle factorization. Like other neural net models with external memory [e.g., Mozer and Das, 1993, Graves et al., 2016, Sukhbaatar et al., 2015], SCOFF includes a set of slots wh”
Weakly Supervised Memory Networks (2015, background): “For example, Das et al. (1992) and Mozer & Das (1993) introduced an explicit stack with push and pop operations which has been revisited recently by Joulin & Mikolov (2015) in the context of an RNN model.”
On Learning to Think: Algorithmic Information Theory for Novel Combinations of Reinforcement Learning Controllers and Recurrent Neural World Models (2015, background): “Let both C and M be RNNs or similar general parallel-sequential computers [229, 47, 175, 232, 231, 103, 80, 303].”
One Big Net For Everything (2018, methodology): “Many RNN-like models can be used to build general computers, e.g., RNNs controlling push-down automata [9, 41] or other types of differentiable memory [20] including differentiable fast weights [63, 65], as well as closely related RNN-based meta-learners [66, 25, 51].”
TerpreT: A Probabilistic Programming Language for Program Induction (2016, background): “, 1989) or a scratch pad simplified to a stack (Mozer and Das, 1992), and the controlling network learned when to push data to and pop (load) data from that stack.”
End-To-End Memory Networks (2015, background): “Subsequent work in the 1990’s explored other types of memory [18, 5, 16].”
Continual learning in reinforcement environments (1995, background): “Neural-network algorithms that learn to use a stack [67, 101] do exist.”
A Perspective on Objects and Systematic Generalization in Model-Based RL (2019, background): “Das et al. (1992); Mozer & Das (1993); Reed & de Freitas (2015); Graves et al. (2016)) or fast weights (eg.”
Fractal encoding of context‐free grammars in connectionist networks (2000, methodology): “Many have used an external stack (Giles et al ., 1990; Sun et al ., 1990; Das et al ., 1993; Mozer & Das, 1993; Zheng et al ., 1994).”