Contributions by metaplasticity to solving the Catastrophic Forgetting Problem.

Catastrophic forgetting (CF) refers to the sudden and severe loss of prior information in learning systems when acquiring new information. CF has been an Achilles heel of standard artificial neural networks (ANNs) when learning multiple tasks sequentially. The brain, by contrast, has solved this problem during evolution. Modellers now use a variety of strategies to overcome CF, many of which have parallels to cellular and circuit functions in the brain. One common strategy, based on metaplasticity phenomena, controls the future rate of change at key connections to help retain previously learned information. However, the metaplasticity properties so far used are only a subset of those existing in neurobiology. We propose that as models become more sophisticated, there could be value in drawing on a richer set of metaplasticity rules, especially when promoting continual learning in agents moving about the environment.

Hopfield networks as a model of prototype-based category learning: A method to distinguish trained, spurious, and prototypical attractors.

We present an investigation of the potential use of Hopfield networks to learn neurally plausible, distributed representations of category prototypes. Hopfield networks are dynamical models of autoassociative memory which learn to recreate a set of input states from any given starting state. These networks, however, will almost always learn states which were not presented during training, so called spurious states. Historically, spurious states have been an undesirable side-effect of training a Hopfield network and there has been much research into detecting and discarding these unwanted states. However, we suggest that some of these states may represent useful information, namely states which represent prototypes of the categories instantiated in the network's training data. It would be desirable for a memory system trained on multiple instance tokens of a category to extract a representation of the category prototype. We present an investigation showing that Hopfield networks are in fact capable of learning category prototypes as strong, stable, attractors without being explicitly trained on them. We also expand on previous research into the detection of spurious states in order to show that it is possible to distinguish between trained, spurious, and prototypical attractors.

Preschool children's environmental exposures: a case-control epidemiological study of the presence of asthma-like symptoms.

The heterogeneity of asthma and asthma-like symptoms results in difficulty defining, diagnosing, and therefore estimating environmental exposures and associations with wheezing or asthma-like symptoms. Determining the disease burden for young children is particularly challenging. In the study described in this article, counter-matched sampling design was used to select participants from the Woman, Infants, and Children (WIC) program for this nested case-control study (N = 691, with 412 controls). Data were collected through structured interviews. Exposure to wood or oil smoke, soot, or exhaust was significantly associated with early-life asthma, as was exposure to cockroaches. Multivariate analyses showed that increasing age, male gender, presence of allergies (although not the type of allergies), and the presence of eczema at birth predicted wheezing behaviors in the authors' study. The authors estimated the prevalence of wheezing behavior in a population of low-income preschool children was 31% with prevalence rates higher among African-American children as compared to other races/ethnicities. Fifty-one percent of those children whose caregivers reported wheezing, however, had not received a diagnosis of asthma. Further study is recommended to compare the differences in the wheezing experiences between those diagnosed with asthma and those who are undiagnosed, with the intent of designing primary prevention interventions tailored to parents and caregivers of young children.

Neural network modelling and dynamical system theory: are they relevant to study the governing dynamics of association football players?

Recent studies have explored the organization of player movements in team sports using a range of statistical tools. However, the factors that best explain the performance of association football teams remain elusive. Arguably, this is due to the high-dimensional behavioural outputs that illustrate the complex, evolving configurations typical of team games. According to dynamical system analysts, movement patterns in team sports exhibit nonlinear self-organizing features. Nonlinear processing tools (i.e. Artificial Neural Networks; ANNs) are becoming increasingly popular to investigate the coordination of participants in sports competitions. ANNs are well suited to describing high-dimensional data sets with nonlinear attributes, however, limited information concerning the processes required to apply ANNs exists. This review investigates the relative value of various ANN learning approaches used in sports performance analysis of team sports focusing on potential applications for association football. Sixty-two research sources were summarized and reviewed from electronic literature search engines such as SPORTDiscus, Google Scholar, IEEE Xplore, Scirus, ScienceDirect and Elsevier. Typical ANN learning algorithms can be adapted to perform pattern recognition and pattern classification. Particularly, dimensionality reduction by a Kohonen feature map (KFM) can compress chaotic high-dimensional datasets into low-dimensional relevant information. Such information would be useful for developing effective training drills that should enhance self-organizing coordination among players. We conclude that ANN-based qualitative analysis is a promising approach to understand the dynamical attributes of association football players.

Visualizing changes in lower body coordination with different types of foot orthoses using self-organizing maps (SOM).

Human movement involves the coordination of individual segments controlled by the central nervous system and powered by the muscles. However, visualization of this high-dimensional coordination between kinematic and kinetic parameters is challenging. The purposes of this study were (a) to identify differences in lower extremity coordination between different types of foot orthoses using Kohonen self-organizing maps (SOM) and (b) to demonstrate the SOM visualization of high-dimensional coordination in gait. This study used gait data for twenty subjects while running in four different orthotic conditions (control, posted, molded, and posted-molded) from a previous study. Data for one exemplar participant was used to demonstrate the visualization technique. In this visualization, areas on an output map represent certain characteristics of the gait cycle. By comparing trials of gait in different orthotic conditions a visual analysis of high-dimensional coordination is possible. Posting orthoses were shown to reduce and molded orthoses were shown to increase ankle mobility, respectively. However, when posting and molding were combined, the effects of the molded orthoses over-rode those of the posted orthoses. In fact, trials using posted-molded orthoses enhanced the effects of molded orthoses. SOMs may contribute to a better understanding of changes in the coordination of kinematic and kinetic variables at certain phases of the gait cycle under different conditions.

Artificial neural networks for analyzing inter-limb coordination: the golf chip shot.

Motor control research relies on theories, such as coordination dynamics, adapted from physical sciences to explain the emergence of coordinated movement in biological systems. Historically, many studies of coordination have involved inter-limb coordination of relatively few degrees of freedom. This study looked at the high-dimensional inter-limb coordination used to perform the golf chip shot toward six different target distances. This study also introduces a visualization of high-dimensional coordination relevant within the coordination dynamics theoretical framework. A specific type of Artificial Neural Network (ANN), the Self-Organizing Map (SOM), was used for the analysis. In this study, the trajectory of consecutive best-matching nodes on the output map was used as a collective variable and subsequently fed into a second SOM which was used to create visualization of coordination stability. The SOM trajectories showed changes in coordination between movement patterns used for short chip shots and movement patterns used for long chip shots. The attractor diagrams showed non-linear phase transitions for three out of four players. The methods used in this study may offer a solution for researchers from a coordination dynamics perspective who intend to use data obtained from discrete high-dimensional movements.

Beyond the cathedral: building trust to engage the African American community in health promotion and disease prevention.

Effective efforts to eliminate health disparities must be grounded in strong community partnerships and trusting relationships between academic institutions and minority communities. However, there are often barriers to such efforts, including the frequent need to rely on time-limited funding mechanisms that take categorical approaches. This article provides an overview of health promotion and disease prevention projects implemented through the Community Outreach and Information Dissemination Core (COID) of the Center for Minority Health, within the Graduate School of Public Health at the University of Pittsburgh. The COID is one of five Cores that comprised the University of Pittsburgh's NIH Excellence in Partnerships for Community Outreach, and Research on Disparities in Health and Training (EXPORT Health) funded from 2002 to 2007 by the National Center on Minority Health and Health Disparities. Based in large part on the success of the community engagement activities, in 2007, the National Center on Minority Health and Health Disparities, National Institutes of Health, designated the CMH as a Research Center of Excellence on Minority Health Disparities. COID major initiatives included the Community Research Advisory Board, Health Disparity Working Groups, Health Advocates in Reach, Healthy Class of 2010, and the Healthy Black Family Project. Lessons learned may provide guidance to other academic institutions, community-based organizations, and health departments who seek to engage minority communities in changing social norms to support health promotion and disease prevention.

Iron overload in the Asian community.

Hereditary hemochromatosis is an iron overload disorder that can lead to the impairment of multiple organs and is caused by mutations in one or more different genes. Type 1 hemochromatosis is the most common form of the disease and results from mutations in the HFE gene. Juvenile hemochromatosis (JH) is the most severe form, usually caused by mutations in hemojuvelin (HJV) or hepcidin (HAMP). The autosomal dominant form of the disease, type 4, is due to mutations in the SLC40A1 gene, which encodes for ferroportin (FPN). Hereditary hemochromatosis is commonly found in populations of European origin. By contrast, hemochromatosis in Asia is rare and less well understood and can be masked by the presence of iron deficiency and secondary iron overload from thalassemia. Here, we provide a comprehensive report of hemochromatosis in a group of patients of Asian origin. We have identified novel mutations in HJV, HAMP, and SLC40A1 in countries not normally associated with hereditary hemochromatosis (Pakistan, Bangladesh, Sri Lanka, and Thailand). Our family studies show a high degree of consanguinity, highlighting the increased risk of iron overload in many countries of the developing world and in countries in which there are large immigrant populations from these regions.

A model of cardinality blindness in inferotemporal cortex.

A classifier is cardinality invariant if it can classify more than one token of a single type at a time. We present a convolutional neural network (CNN) model of inferotemporal cortex (IT) and show that it is cardinality invariant. While the CNN is designed with translation invariance in mind, cardinality invariance is an emergent property. We speculate that translation invariance may lead to cardinality invariance in general, and particularly in IT. Recent investigations have shown that cells in IT are indeed cardinality blind. We also explore the implications of a cardinality blind classifier for vision overall, concentrating on visual attention and search.

Memory retention--the synaptic stability versus plasticity dilemma.

Memory maintenance is widely believed to involve long-term retention of the synaptic weights that are set within relevant neural circuits during learning. However, despite recent exciting technical advances, it has not yet proved possible to confirm experimentally this intuitively appealing hypothesis. Artificial neural networks offer an alternative methodology as they permit continuous monitoring of individual connection weights during learning and retention. In such models, ongoing alterations in connection weights are required if a network is to retain previously stored material while learning new information. Thus, the duration of synaptic change does not necessarily define the persistence of a memory; rather, it is likely that a regulated balance of synaptic stability and synaptic plasticity is required for optimal memory retention in real neuronal circuits.

A robust method for distinguishing between learned and spurious attractors.

Hopfield/constraint satisfaction type networks can be used to learn (autoassociate) patterns. Random inputs to the network will sometimes converge on states which are learned patterns, and sometimes converge on states which are unlearned/spurious. It would be useful for many reasons to be able to tell whether or not a given state was learned or spurious. In this paper we present a robust and general method, based on 'energy profiles', which allows us to make this distinction. We briefly describe related research, and note links with the study of recall, recognition and familiarity in the psychological literature.