Letting structure emerge: connectionist and dynamical systems approaches to cognition

McClelland, James L.; Botvinick, Matthew; Noelle, David C.; Plaut, David C.; Rogers, Timothy T.; Seidenberg, Mark S.; Smith, Linda B.

doi:10.1016/j.tics.2010.06.002

Cited by 380 publications

(261 citation statements)

References 61 publications

Supporting

Mentioning

250

Contrasting

Unclassified

Order By: Relevance

“…In one way, the answer is yes: because neural networks are universal approximators, it is always possible to construct one that approximates the input-output functionality of a specific Bayesian model. In practice, however, the answer is usually no: the two methods have very different strengths and weaknesses, and therefore their value as modelling tools varies depending on the questions being asked (see Griffiths, Chater, Kemp, Perfors, and Tenenbaum (2010) and McClelland et al (2010) for a more thorough discussion of these issues). One difference is that connectionist models make certain commitments about representation that make it difficult to capture explicit symbolic knowledge, of the sort that is commonly incorporated into cognitive theories.…”

Section: Where Does It All Come From?mentioning

confidence: 99%

A tutorial introduction to Bayesian models of cognitive development

et al. 2011

View full text Add to dashboard Cite

We present an introduction to Bayesian inference as it is used in probabilistic models of cognitive development. Our goal is to provide an intuitive and accessible guide to the what, the how, and the why of the Bayesian approach: what sorts of problems and data the framework is most relevant for, and how and why it may be useful for developmentalists. We emphasize a qualitative understanding of Bayesian inference, but also include information about additional resources for those interested in the cognitive science applications, mathematical foundations, or machine learning details in more depth. In addition, we discuss some important interpretation issues that often arise when evaluating Bayesian models in cognitive science.

show abstract

Section: Where Does It All Come From?mentioning

confidence: 99%

A tutorial introduction to Bayesian models of cognitive development

et al. 2011

View full text Add to dashboard Cite

show abstract

“…Already in their first months of life, infants rapidly learn to recognize complex objects and events in their visual input (1)(2)(3). Probabilistic learning models, as well as connectionist and dynamical models, have been developed in recent years as powerful tools for extracting the unobserved causes of sensory signals (4)(5)(6). Some of these models can efficiently discover significant statistical regularities in the observed signals, which may be subtle and of high order, and use them to construct world models and guide behavior (7)(8)(9)(10).…”

mentioning

confidence: 99%

From simple innate biases to complex visual concepts

Ullman

Harari

Dorfman

2012

Proc. Natl. Acad. Sci. U.S.A.

View full text Add to dashboard Cite

Early in development, infants learn to solve visual problems that are highly challenging for current computational methods. We present a model that deals with two fundamental problems in which the gap between computational difficulty and infant learning is particularly striking: learning to recognize hands and learning to recognize gaze direction. The model is shown a stream of natural videos and learns without any supervision to detect human hands by appearance and by context, as well as direction of gaze, in complex natural scenes. The algorithm is guided by an empirically motivated innate mechanism-the detection of "mover" events in dynamic images, which are the events of a moving image region causing a stationary region to move or change after contact. Mover events provide an internal teaching signal, which is shown to be more effective than alternative cues and sufficient for the efficient acquisition of hand and gaze representations. The implications go beyond the specific tasks, by showing how domain-specific "proto concepts" can guide the system to acquire meaningful concepts, which are significant to the observer but statistically inconspicuous in the sensory input.A basic question in cognitive development is how we learn to understand the world on the basis of sensory perception and active exploration. Already in their first months of life, infants rapidly learn to recognize complex objects and events in their visual input (1-3). Probabilistic learning models, as well as connectionist and dynamical models, have been developed in recent years as powerful tools for extracting the unobserved causes of sensory signals (4-6). Some of these models can efficiently discover significant statistical regularities in the observed signals, which may be subtle and of high order, and use them to construct world models and guide behavior (7-10). However, even powerful statistical models have inherent difficulties with natural cognitive concepts, which depend not only on statistical regularities in the sensory input but also on their significance and meaning to the observer. For example, in learning to understand actions and goals, an important part is identifying the agents' hands, their configuration, and their interactions with objects (1-3). This is an example in which significant and meaningful features can be nonsalient and highly variable and therefore difficult to learn. Our testing shows that current computational methods for general object detection (11-13) applied to large training data do not result by themselves in automatically learning about hands. In contrast, detecting hands (14), paying attention to what they are doing (15, 16), and using them to make inferences and predictions (1-3, 17) are natural for humans and appear early in development. How is it possible for infants to acquire such concepts in early development?A large body of developmental studies has suggested that the human cognitive system is equipped through evolution with basic innate structures that facilitate the acquisition of meaningf...

show abstract

“…[2]), the working hypothesis is that -at a less fine-grained, more abstract level of description -the functions computed by intuitive processes are often well approximated by symbolic descriptions. The intuitive mental processes that actually compute these functions do not however admit a symbolic description: these processes require subsymbolic, connectionist descriptions.…”

Section: Effective Procedures Vs Intuitive Cognitive Processesmentioning

confidence: 99%

“…Then V F and V R are combined by the tensor product to form the actual representational (vector) space for trees V S ≡ V F ⊗ V R . 2 Finally, the filler-role decomposition and vectorial realizations are combined: the vector realization of s, ψ S (s) ∈ V S , is the sum of the vector realizations of the filler-role bindings of s, each of which uses the tensor product to bind together the encodings of the filler and the role:…”

mentioning

confidence: 99%

Subsymbolic Computation Theory for the Human Intuitive Processor

Smolensky

2012

Lecture Notes in Computer Science

View full text Add to dashboard Cite

Abstract. The classic theory of computation initiated by Turing and his contemporaries provides a theory of effective procedures -algorithms that can be executed by the human mind, deploying cognitive processes constituting the conscious rule interpreter. The cognitive processes constituting the human intuitive processor potentially call for a different theory of computation. Assuming that important functions computed by the intuitive processor can be described abstractly as symbolic recursive functions and symbolic grammars, we ask which symbolic functions can be computed by the human intuitive processor, and how those functions are best specified -given that these functions must be computed using neural computation. Characterizing the automata of neural computation, we begin the construction of a class of recursive symbolic functions computable by these automata, and the construction of a class of neural networks that embody the grammars defining formal languages.

show abstract

Letting structure emerge: connectionist and dynamical systems approaches to cognition

Cited by 380 publications

References 61 publications

A tutorial introduction to Bayesian models of cognitive development

A tutorial introduction to Bayesian models of cognitive development

From simple innate biases to complex visual concepts

Subsymbolic Computation Theory for the Human Intuitive Processor

Contact Info

Product

Resources

About