Analysis of Dynamical Recognizers

By Alan D. Blair, Jordan B. Pollack. 1997

How researchers see this paper

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

Citations in the corpus

Deep learning in neural networks: An overview (2014, methodology): “LSTM clearly outperformed previous RNNs on tasks that require learning the rules of regular languages describable by deterministic Finite State Automata (FSAs) (Watrous and Kuhn, 1992; Casey, 1996; Siegelmann, 1992; Blair and Pollack, 1997; Kalinke and Lehmann, 1998; Zeng et al., 1994; Manolios and ”
LSTM recurrent networks learn simple context-free and context-sensitive languages (2001, background): “For the gates, , is a logistic sigmoid with range [0, 1 ].”
Rule Extraction from Recurrent Neural Networks: ATaxonomy and Review (2005, methodology): “The consistency over diﬀerent parameter settings (of the quantisation parameter q in the equipartitioned RNN-RE algo-rithm) has also been proposed as an indicator of regularity in the underlying network (Blair & Pollack 1997).”
Context-free and context-sensitive dynamics in recurrent neural networks (2000, background): “It is also well known that a large class of RNNs are based on analog components and are not limited to ﬁnite state computation (Siegelmann, 1999; Moore, 1998; Blair and Pollack, 1997; Kolen, 1994).”
Architectural Bias in Recurrent Neural Networks: Fractal Analysis (2002, background): “The underlying dynamics which emerges through training has been investigated in the work of Blair, Bod¶en, Pollack and others (Blair & Pollack, 1997; Bod¶en & Blair, 2002; Rodriguez, Wiles & Elman, 1999).”
On learning context-free and context-sensitive languages (2002, background): “to do so even when trained on regular languages [ 2 ].”
Simple Recurrent Networks Learn Context-Free and Context-Sensitive Languages by Counting (2001, unknown): “…machine, and one should treat it as a continuous-space dynamical system (Pollack, 1991; Kolen, 1994; Tani & Fukumura, 1995; Casey, 1996; Hoelldobler, Kalinke, & Lehmann, 1997; Blair & Pollack, 1997; Tino, Horne, Giles, & Collingwood, 1998; Tabor & Tanenhaus, 1999; Rodriguez, Wiles, & Elman, 1999).”
A Recurrent Neural Network that Learns to Count (1999, background): “Blair and Pollack (1997) and Pollack (1991) showed how an RNN that learns an RL can represent an in® nite state machine as one looks at higher resolution of phase space partitions, even though the RNN looks like an FSM at lower resolution.”
Fractal encoding of context‐free grammars in connectionist networks (2000, background): “Intriguingly, this is the character of a number of recent results connecting symbolic computation with multi-stable dynamical systems (Pollack, 1991; Barnsley, 1993 [1988]; Blair & Pollack, 1997; Moore, 1998; Tinˇo & Dorffner, 1998; Tinˇo, 1999).”
Incremental training of first order recurrent neural networks to predict a context-sensitive language (2003, methodology): “For more details on this interpretation see for example (Pollack, 1991; Blair and Pollack, 1997; Kremer, 1999; Carrasco et al., 2000).”