In contrast to symbolic or associationist representations, I advocate a third form of representing information that employs geometrical structures. I argue that this form is appropriate for modelling concept learning. By using the geometrical structures of what I call conceptual spaces, I define properties and concepts. A learning model that shows how properties and concepts can be learned in a simple but naturalistic way is then presented. I also discuss the advantages of the geometric approach over the (...) symbolic and associationist traditions. (shrink)

Humans learn new concepts extremely fast. One or two examples of a new concept are often sufficient for us to grasp its meaning. Traditional theories of concept formation, such as symbolic or connectionist representations, have problems explaining the quick learning exhibited by humans. In contrast to these representations, I advocate a third form of representing categories, which employs geometric structures. I argue that this form is appropriate for modeling concept learning. By using the geometric structures (...) of what I call “conceptual spaces,” I define properties and concepts. A learning model that shows how properties and concepts can be learned in a simple but naturalistic way is then presented. This model accounts well for the role of similarity judgments in concept learning. Finally, as an application, the concept representations are used to give an analysis of nonmonotonic reasoning. (shrink)

Linguistic and non-linguistic pattern learning have been studied separately, but we argue for a comparative approach. Analogous inductive problems arise in phonological and visual pattern learning. Evidence from three experiments shows that human learners can solve them in analogous ways, and that human performance in both cases can be captured by the same models. We test GMECCS, an implementation of the Configural Cue Model in a Maximum Entropy phonotactic-learning framework with a single free parameter, against the alternative (...) hypothesis that learners seek featurally simple algebraic rules. We study the full typology of patterns introduced by Shepard, Hovland, and Jenkins (), instantiated as both phonotactic patterns and visual analogs, using unsupervised training. Unlike SHJ, Experiments 1 and 2 found that both phonotactic and visual patterns that depended on fewer features could be more difficult than those that depended on more features, as predicted by GMECCS but not by rule-seeking. GMECCS also correctly predicted performance differences between stimulus subclasses within each pattern. A third experiment tried supervised training to elicit simple rule-seeking phonotactic learning, but cue-based behavior persisted. We conclude that similar cue-based cognitive processes are available for phonological and visual concept learning, and hence that studying either kind of learning can lead to significant insights about the other. (shrink)

Susan Carey's account of Quinean bootstrapping has been heavily criticized. While it purports to explain how important new concepts are learned, many commentators complain that it is unclear just what bootstrapping is supposed to be or how it is supposed to work. Others allege that bootstrapping falls prey to the circularity challenge: it cannot explain how new concepts are learned without presupposing that learners already have those very concepts. Drawing on discussions of concept learning from the philosophical literature, (...) this article develops a detailed interpretation of bootstrapping that can answer the circularity challenge. The key to this interpretation is the recognition of computational constraints, both internal and external to the mind, which can endow empty symbols with new conceptual roles and thus new contents. (shrink)

While Tenenbaum and Griffiths impressively consolidate and extend Shepard's research in the areas of stimulus representation and generalization, there is a need for complexity measures to be developed to control the flexibility of their “hypothesis space” approach to representation. It may also be possible to extend their concept learning model to consider the fundamental issue of representational adaptation. [Tenenbaum & Griffiths].

In contrast to symbolic or associationist representations, I advocate a third form of representing information that employs geometrical structures. I argue that this form is appropriate for modelling concept learning. By using the geometrical structures of what I call conceptual spaces, I define properties and concepts. A learning model that shows how properties and concepts can be learned in a simple but naturalistic way is then presented. I also discuss the advantages of the geometric approach over the (...) symbolic and associationist traditions. (shrink)

Where does a probabilistic language-of-thought (PLoT) come from? How can we learn new concepts based on probabilistic inferences operating on a PLoT? Here, I explore these questions, sketching a traditional circularity objection to LoT and canvassing various approaches to addressing it. I conclude that PLoT-based cognitive architectures can support genuine concept learning; but, currently, it is unclear that they enjoy more explanatory breadth in relation to concept learning than alternative architectures that do not posit any LoT.

The central focus of this article is the epistemological assumption that knowledge could be generated based on human beings’ experiences and over their conceptions of the world. Logical characterisation of human inductive learning over their produced conceptions within terminological systems and providing a logical background for theorising over the Human Concept Learning Problem (HCLP) in terminological systems are the main contributions of this research. In order to make a linkage between ‘Logic’ and ‘Cognition’, Description Logics (DLs) will (...) be employed to provide a logical description and analysis of actual human inductive reasoning (and learning). This research connects with the topics ‘logic & learning’, ‘cognitive modelling’, and ‘terminological knowledge representation’. (shrink)

The current study examined whether pigeons could learn to use abstract concepts as the basis for conditionally switching behavior as a function of time. Using a mid-session reversal task, experienced pigeons were trained to switch from matching-to-sample (MTS) to non-matching-to-sample (NMTS) conditional discriminations within a session. One group had prior training with MTS, while the other had prior training with NMTS. Over training, stimulus set size was progressively doubled from 3 to 6 to 12 stimuli to promote abstract concept (...) development. Prior experience had an effect on the initial learning at each of the set sizes but by the end of training there were no group differences, as both groups showed similar within-session linear matching functions. After acquiring the 12-item set, abstract-concept learning was tested by placing novel stimuli at the beginning and end of a test session. Prior matching and non-matching experience affected transfer behavior. The matching experienced group transferred to novel stimuli in both the matching and non-matching portion of the sessions using a matching rule. The non-matching experienced group transferred to novel stimuli in both portions of the session using a non-matching rule. The representations used as the basis for mid-session reversal of the conditional discrimination behaviors and subsequent transfer behavior appears to have different temporal sources. The implications for the flexibility and organization of complex behaviors are considered. (shrink)

In the context of the conceptual spaces framework, it has been argued that a natural concept is represented by a convex region in a similarity space. The convexity requirement has been defended on grounds of cognitive economy: among other benefits, concepts represented by convex regions have been said to be easily learnable, or more easily than concepts represented by nonconvex regions. There is some evidence that concepts in use are represented by regions that are convex, or at least almost (...) so. There is so far no evidence that concepts represented by convex regions are more easily learnable than ones represented by regions that satisfy topological criteria that are somewhat less stringent, most notably that of connectedness. This note presents the outcomes from computational studies carried out on perceptual color space as well as on a shape space for representing various container objects, indicating that convexity indeed facilitates learning more than does connectedness. The studies use the training of neural nets as a model of human learning. (shrink)

Complex cognitive processes like concept learning involve a mixture of redundant explicit and implicit processes that are active simultaneously. This aspect of cognitive architecture creates difficulties in determining the influence of consciousness on processing. We propose that the interaction between an individual's regulatory focus and the reward structure of the current task influences the degree to which explicit processing is active. Thus, by manipulating people's motivational state and the nature of the task they perform, we can vary the (...) influence of conscious processing in cognitive performance. We demonstrate the utility of this view by focusing on studies in which people acquire new perceptual concepts by learning to classify them. This technique will allow us to better tease apart the roles of explicit and implicit processing in a variety of cognitive tasks. (shrink)

The aberrations of a gene can influence it and the functions of its neighbour genes in gene interaction network, leading to the development of carcinogenesis of normal cells. In consideration of gene interaction network as a complex network, previous studies have made efforts on the driver attribute filling of genes via network properties of nodes and network propagation of mutations. However, there are still obstacles from problems of small size of cancer samples and the existence of drivers without property of (...) network neighbours, limiting the discovery of cancer driver genes. To address these obstacles, we propose an efficient modularity subspace based concept learning model. Our model can overcome the curse of dimensionality due to small samples via dimension reduction in the task of attribute concept learning and explore the features of genes through modularity subspace beyond the network neighbours. The evaluation analysis also demonstrates the superiority of our model in the task of driver attribute filling on two gene interaction networks. Generally, our model shows a promising prospect in the application of interaction network analysis of tumorigenesis. (shrink)

This article proposes a new model of human concept learning that provides a rational analysis of learning feature‐based concepts. This model is built upon Bayesian inference for a grammatically structured hypothesis space—a concept language of logical rules. This article compares the model predictions to human generalization judgments in several well‐known category learning experiments, and finds good agreement for both average and individual participant generalizations. This article further investigates judgments for a broad set of 7‐feature concepts—a (...) more natural setting in several ways—and again finds that the model explains human performance. (shrink)

A new model of the development of temporal concepts is described that assumes that there are substantial changes in how children think about time in the early years. It is argued that there is a shift from understanding time in an event-dependent way to an event-independent understanding of time. Early in development, very young children are unable to think about locations in time independently of the events that occur at those locations. It is only with development that children begin to (...) have a proper grasp of the distinction between past, present, and future, and represent time as linear and unidirectional. The model assumes that although children aged 2 to 3 years may categorize events differently depending on whether they lie in the past or the future, they may not be able to understand that whether an event is in the past or future is something that changes as time passes and varies with temporal perspective. Around 4 to 5 years, children understand how causality operates in time, and can grasp the systematic relations that obtain between different locations in time, which provides the basis for acquiring the conventional clock and calendar system. (shrink)