The Algebraic Mind: Integrating Connectionism and Cognitive Science

By Gary F. Marcus. 2001

639 citations recorded by the source corpus.

Subfields: Cognitive Science & Psychology

How researchers see this paper

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

In Cognitive Science & Psychology

Citations in the corpus

The Next Decade in AI: Four Steps Towards Robust Artificial Intelligence (2020, background): “Collectively, these four assumptions about variables, bindings, instances, and operations over variables comprise the core of symbol-manipulation (Newell, 1980; Marcus, 2001).”
Neural Turing Machines (2014, background): “Addressing, fundamental to our work, is usually left out from computational models in neuroscience, though it deserves to be mentioned that Gallistel and King (Gallistel and King, 2009) and Marcus (Marcus, 2003) have argued that addressing must be implicated in the operation of the brain.”
What can the brain teach us about building artificial intelligence? (2016, background): “& Marcus, G. (2014) The scope and limits of simulation in cognition.”
Analysing Mathematical Reasoning Abilities of Neural Models (2019, background): “One area where human intelligence still differs and excels compared to neural models is discrete compositional reasoning about objects and entities, that “algebraically generalize” (Marcus, 2003).”
Reconciling deep learning with symbolic artificial intelligence: representing objects and relations (2019, background): “Signiﬁcantly, the shortcomings of deep learning align with the strengths of symbolic AI, which suggests the time is right for a reconciliation [15,9 ].”
On the Binding Problem in Artificial Neural Networks (2020, background): “Note that the identity of an object cannot be tied exclusively to its visible properties, as illustrated by the extreme example of a fairytale prince that is transformed into a frog (Marcus, 2003; Bambini et al., 2012).”
Toward an Integration of Deep Learning and Neuroscience (2016, methodology): “…for specialized functions that have been made throughout the computational neuroscience literature, and is inﬂuenced strongly by the models of O’Reilly, Eliasmith, Grossberg, Marcus, Hayworth and others (Marcus, 2001; O’Reilly, 2006; Eliasmith et al., 2012; Hayworth, 2012; Grossberg, 2013).”
Learning at a distance I. Statistical learning of non-adjacent dependencies. (2004, background): “In the first experiment, we present streams of speech in which precisely the same kinds of syllable regularities occur as in our previous studies, except that the patterned relations among syllables occur between non-adjacent syllables (with an intervening syllable that is unrelated).”
The atoms of neural computation (2014, background): “…recognition and working memory, humans might further re-purpose elements such as the “cortical column”, potentially providing for the capacity to concatenate “arbitrary” symbols (Marcus, 2001) or represent recursive structures (Hauser, Chomsky, & Fitch, 2002; Traxler, Boudewyn, & Loudermilk, 2012).”
An architecture for encoding sentence meaning in left mid-superior temporal cortex (2015, methodology): “In experiment 2, we test the hypothesis that lmSTC flexibly encodes these meanings (at least in part) by explicitly representing the values of the agent (“Who did it?”) and the patient (“To whom was it done?”) (5).”
Assembly pointers for variable binding in networks of spiking neurons (2016, background): “It has often been emphasized (see e.g. [22, 23]) that there is a need to understand brain mechanisms for variable binding.”
Generalization without Systematicity: On the Compositional Skills of Sequence-to-Sequence Recurrent Networks (2017, background): “However, the models are catastrophically affected by systematic differences between training and test sentences, of the sort that would be trivial for an agent equipped with an “algebraic mind” (Marcus, 2003).”
Iterated learning for emergent systematicity in VQA (2021, background): “Whether or not neural networks can exhibit systematic compositionality has been a subject of much debate in the research community (Fodor & Pylyshyn, 1988; Christiansen & Chater, 1994; Marcus, 1998; Phillips, 1998; Chang, 2002; Marcus, 2018; van der Velde et al., 2004; Botvinick & Plaut, 2009; Bower”
Commonsense reasoning about containers using radically incomplete information (2017, background): “…cognitive processes can be usefully described at the knowledge level at least partly in terms of symbolic representations and symbolic reasoning (Marcus, 2001). bending the object O through a tight passage, reasoning whether it is feasible to move O from A to B or not may require a very…”