Learning and development in neural networks--the importance of prior experience

Infants can discriminate between familiar and unfamiliar grammatical patterns expressed in a vocabulary that is distinct from that used earlier during familiarization (Cognition 70(2) (1999) 109; Science 283 (1999) 77). Various models have captured the data, although each required that discrimination be distinct, in terms of the computational process, from familiarization. This article describes a simple recurrent network (SRN), equipped only with the assumption that it should predict what comes next, which models the data without distinguishing between familiarization and discrimination. To accomplish this, the SRN requires pre-training on a range of sequences instantiating different structures and different vocabulary items to those used subsequently during familiarization and test. Pre-training enables the network to avoid replacing structure acquired during familiarization with structure experienced at test. An equivalent enabling condition may underpin infants' resistance to catastrophic interference between the different structures and vocabulary items to which they are exposed.     

Toward a Comprehensive Strategy for Effective Practitioner–Scientist Partnerships and Larger-Scale Community Health and Well-Being

This article articulates joint priorities for the fields of prevention science and community psychology. These priorities are intended to address issues raised by the frequent observation of natural tensions between community practitioners and scientists. The first priority is to expand the knowledge base on practitioner–scientist partnerships, particularly on factors associated with positive outcomes within communities. To further articulate this priority, the paper first discusses the rapid growth in community-based partnerships and the emergent research on them. Next described is an illustrative research project on a partnership model that links state university extension and public school delivery systems. The article then turns to the second, related priority of future capacity-building for diffusion of effective partnership-based interventions to achieve larger-scale health and well-being across communities. It outlines two salient tasks: clarification of a conceptual framework and the formulation of a comprehensive capacity-building strategy for diffusion. The comprehensive strategy would require careful attention to the expansion of networks of effective partnerships, partnership-based research agendas, and requisite policy-making.     

A formal model for structuring local perceptions of environmental space

The problem of constructing computational models of space that mimic those found in human spatial reasoning is addressed. This paper extends a formal model that addressed point objects in a 2D world to include the presence of line objects and barriers. The problem consists of determining an appropriate global model for the spatial configuration given a sequence of local views or observations. A formal model is proposed based on constraint network theory. The model leads to the identification of stable regions within which perceptual information about the environment changes slowly, and zones of transition within which the perceptual information changes more quickly. A special case of transition zones is a gateway, which may be viewed as a kind of discontinuity or singularity in the model. It is expected that the model obtained will be used to mimic human mental representations of navigable outdoor environments.     

Transitive choices by a simple, fully connected, backpropagation neural network: implications for the comparative study of transitive inference.

In search of the minimal requirements for transitive reasoning, a simple neural network was trained and tested on the non-verbal version of the conventional "five-term-series task" – a paradigm used with human adults, children and a variety of non-human species. The transitive performance of the network was analogous in several aspects to that reported for children and animals. The three effects usually associated with transitive choices i.e. "symbolic distance", "lexical marking" and "end-anchor", were also clearly shown by the neural network. In a second experiment, where the training conditions were manipulated, the network failed to match the behavioural pattern reported for human adults in the test following an ordered presentation of the premises. However, it mimicked young children's performance when tested with a novel comparison term. Although we do not intend to suggest a new model of transitive inference, we conclude, in line with other authors, that a simple error-correcting rule can generate transitive behaviour similar to the choice pattern of children and animals in the binary form of the five-term-series task without requiring high-order logical or paralogical abilities. The analysis of the training history and of the final internal structure of the network reveals the associative strategy employed. However, our results indicate that the scope of the associative strategy used by the network might be limited. The extent to which the conventional five-term-series task, in absence of appropriate manipulations of training and testing conditions, is suitable to detect cognitive differences across species is also discussed on the basis of our results. Accepted after revision: 29 May 2001 Electronic Publication     

Finding MAPs using strongly equivalent high order recurrent symmetric connectionist networks

Belief revision is the problem of finding the most plausible explanation for an observed set of evidences. It has many applications in various scientific domains like natural language understanding, medical diagnosis and computational biology. Bayesian Networks (BN) is an important probabilistic graphical formalism widely used for belief revision tasks. In BN, belief revision can be achieved by finding the maximum a posteriori (MAP) assignment. Finding MAP is an NP-Hard problem. In previous work, we showed how to find the MAP assignment in BN using High Order Recurrent Neural Networks (HORN) through an intermediate representation of Cost-Based Abduction. This method eliminates the need to explicitly construct the energy function in two steps, objective and constraints. This paper builds on that previous work by providing the theoretical foundation and proving that the resultant HORN used to find MAP is strongly equivalent to the original BN it tries to solve.     

We're Closer than I Thought: Social Network Heterogeneity,Morality, and Political Persuasion

Literature in the area of social networks indicates that increases in perceived social network attitudinal heterogeneity generate increased openness to attitude change. Recent evidence in the area of morality, however, shows that morally based attitudes are particularly resistant to persuasion and can result in the rejection of disagreeing others. Positing that considering morality would reduce network influence, an experiment varied moral cues presented along with a non‐network persuasive message while holding the actual content constant. Results demonstrate that morality and network composition interact to predict persuasion, such that when people are not cued to consider morality increased network heterogeneity predicts increased persuasion, but when identical messages are presented in a way that invokes morality the impact of network heterogeneity disappears or even reverses marginally. This interactive effect was replicated in two very different political issues: gay adoption and nationalized healthcare. Implications for persuasion by morally motivated sources independent of the effects of specific moral arguments are discussed.     

Social Support and the Religious Dimensions of Close Ties

Despite the influence of social support on physical and mental health, few studies have examined why some close ties are more supportive than others. Though religion provides a rich context for social interaction and a meaningful social identity, it has received little attention in the social support literature. A growing literature on religion and health offers insight into how religion affects social support processes. Using dyadic network data derived from the nationally representative 2006 Portraits of American Life Study, I examine how the religious dimensions of close, nonhousehold ties relate to provision of social support. Results from logistic regression analyses indicate that (controlling for a range of other social tie characteristics) same‐faith ties are significantly more likely to be sources of help “in times of need,” while religious discussion is a strong predictor of receiving both help and advice. The effect of religious homophily is strongest for evangelical Protestants and African‐American Protestants. My findings underscore the need for social support researchers to consider the role of religion in shaping support processes in close relationships.