The Induction of Dynamical Recognizers

By Jordan B. Pollack. 1991

How researchers see this paper

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

In Model Architectures

Citations in the corpus

Inferring Algorithmic Patterns with Stack-Augmented Recurrent Nets (2015, methodology): “For example, Tabor (2000) uses a model based on pushdown dynamical automata and Bod´en & Wiles (2000) use a sequential cascaded network (Pollack, 1991 & Schmidhuber (1997)) which shows interesting results on these tasks.”
Weakly Supervised Memory Networks (2015, background): “Subsequent work in the 1990’s explored other types of memory (Pollack, 1991; Das et al., 1992; Mozer & Das, 1993).”
CHILD: A First Step Towards Continual Learning (1997, background): “Transition hierarchies resemble recurrent neural networks such as those introduced by Elman (1993), Jordan (1986), Robinson and Fallside (1987), Pollack (1991), etc., in that they learn temporal tasks using connectionist methods.”
Incremental training of first order recurrent neural networks to predict a context-sensitive language (2003, methodology): “…as machine states, discrete time recurrent neural networks have been interpreted as neural state machines in analogy to a type of generalised finite state machine called the Mealy –Moore Machine (Blair & Pollack, 1997; Carrasco et al., 2000; Hopcroft & Ullman, 1979; Kremer, 1999; Pollack, 1991).”
An Empirical Evaluation of Rule Extraction from Recurrent Neural Networks (2017, methodology): “We select a set of seven relatively simple grammars, which are originally suggested by Tomita (Tomita, 1982a) and widely studied (Pollack, 1991; Omlin & Giles, 1996c; Wa-trous & Kuhn, 1992) and use them for an empirical study for extracting rules from RNN.”
End-To-End Memory Networks (2015, background): “Subsequent work in the 1990’s explored other types of memory [18, 5, 16].”
Self-Organization of Spatio-Temporal Hierarchy via Learning of Dynamic Visual Image Patterns on Action Sequences (2015, background): “This can be inferred from the nonlinear dynamic system theory of iterated function systems (IFS) [24, 25].”
Real-Time Computing Without Stable States: A New Framework for Neural Computation Based on Perturbations (2002, methodology): “…computation in biological neural systems are based on constructions of arti(cid:142)cial neural networks that simulate Turing machines or other models for digital computation (Pollack, 1991 ; Giles, Miller, Chen, Sun, & Lee, 1992 ; Siegelmann & Sontag, 1994 ; Hyoetyniemi, 1996 ; Moore, 1998).”
On learning context-free and context-sensitive languages (2002, background): “Specifically, second-order Sequential Cascaded Networks (SCNs; [12]) are able to learn to process strings well outside the training set in a manner which naturally scales with the length of strings [4].”
Divide and Conquer with Neural Networks (2017, background): “In particular, dynamic learning appears in works such as [11] and [17].”
A Machine Learning Method for Extracting Symbolic Knowledge from Recurrent Neural Networks (2004, background): “The relationship between finite state machines and recurrent neural networks has been explored in various ways (Elman, 1990; Watrous & Kuhn, 1992a; Pollack, 1991; Giles, Miller, Chen, Chen, et al., 1992; Zeng, Goodman, & Smyth, 1993).”