Two forms of immediate reward reinforcement learning for exploratory data analysis

We review two forms of immediate reward reinforcement learning: in the first of these, the learner is a stochastic node while in the second the individual unit is deterministic but has stochastic synapses. We illustrate the first method on the problem of Independent Component Analysis. Four learning rules have been developed from the second perspective and we investigate the use of these learning rules to perform linear projection techniques such as principal component analysis, exploratory projection pursuit and canonical correlation analysis. The method is very general and simply requires a reward function which is specific to the function we require the unit to perform. We also discuss how the method can be used to learn kernel mappings and conclude by illustrating its use on a topology preserving mapping.     

Family study of schizophrenia: exploratory analysis for relevant factors

Two hundred twenty-nine schizophrenic patients, diagnosed according to DSM-III criteria, and their first degree relatives (except children) were studied. HLA A1 and CRAG A1 antigens were used as markers of probable dopaminergic pathology and, therefore, as possible indicators of genetic homogeneity that might identify subgroups of families with a specific and recognizable liability. Data were subjected to a logistic analysis in which the dependent variable was the presence of schizophrenia spectrum disorders in relatives, and the independent variables were the presence or absence of HLA A1 and CRAG A1 antigens, the sex of the proband, the sex of the relative, the severity of illness in the proband, and the type of relationship. The results for the entire sample demonstrate that the type of relationship, the sex of the proband, and the sex of the relative have significant effects on the risk of disorder in the relatives. In addition, the presence in the proband of one of the CRAG A1 antigens is a valid classification criterion for identifying a relatively homogeneous subgroup of families of schizophrenic patients.     

Improved learning algorithms for mixture of experts in multiclass classification

Mixture of experts (ME) is a modular neural network architecture for supervised learning. A double-loop Expectation-Maximization (EM) algorithm has been introduced to the ME architecture for adjusting the parameters and the iteratively reweighted least squares (IRLS) algorithm is used to perform maximization in the inner loop [Jordan, M.I., Jacobs, R.A. (1994). Hierarchical mixture of experts and the EM algorithm, Neural Computation, 6(2), 181–214]. However, it is reported in literature that the IRLS algorithm is of instability and the ME architecture trained by the EM algorithm, where IRLS algorithm is used in the inner loop, often produces the poor performance in multiclass classification. In this paper, the reason of this instability is explored. We find out that due to an implicitly imposed incorrect assumption on parameter independence in multiclass classification, an incomplete Hessian matrix is used in that IRLS algorithm. Based on this finding, we apply the Newton–Raphson method to the inner loop of the EM algorithm in the case of multiclass classification, where the exact Hessian matrix is adopted. To tackle the expensive computation of the Hessian matrix and its inverse, we propose an approximation to the Newton–Raphson algorithm based on a so-called generalized Bernoulli density. The Newton–Raphson algorithm and its approximation have been applied to synthetic data, benchmark, and real-world multiclass classification tasks. For comparison, the IRLS algorithm and a quasi-Newton algorithm called BFGS have also been applied to the same tasks. Simulation results have shown that the use of the proposed learning algorithms avoids the instability problem and makes the ME architecture produce good performance in multiclass classification. In particular, our approximation algorithm leads to fast learning. In addition, the limitation of our approximation algorithm is also empirically investigated in this paper.     

Improved imaging of rat hindfoot prints for walking track analysis.

Walking track analysis is an investigative technique that allows a researcher to assess objectively the functional capacity of the limb in the rat sciatic nerve model. This study has shown that paint and paper is a better method for imaging the rat foot plantar surface than other methods described in the literature. The block printing paint has the distinct advantage of keeping radial diffusion error to a minimum. Paint more correctly images the important anatomical landmarks of the rat plantar surface. There is better traction (less slippage) rendering the prints easier to interpret. The authors conclude that the paint and paper is a superior method to the other methods currently available.     

An exploratory analysis of correlates of recovery

OBJECTIVE: The concept of recovery has received increasing emphasis in the delivery of services to persons with schizophrenia. This study was an initial effort to develop an empirically based model of factors associated with a recovery orientation. METHODS: The authors reanalyzed data from 825 persons with schizophrenia who were assessed in the Schizophrenia Patient Outcomes Research Team (PORT) client survey. Multiple regression models were used to identify client and service use variables associated with each of four domains identified as important to a recovery orientation: life satisfaction, hope and optimism, knowledge about mental illness and services, and empowerment. RESULTS: In each regression model, the strongest relationship was observed between recovery orientation and lower severity of depressive symptoms. Both receipt of family psychoeducation and fewer side effects of medications were significantly and positively related to three of the four recovery domains. Psychotic symptoms were associated with less life satisfaction. Receipt of various services, including day treatment and legal services, was positively associated with knowledge about illness and services. CONCLUSIONS: Severity of psychiatric symptoms, a core feature of the biomedical perspective of mental illness, was negatively associated with a recovery orientation, and use of a variety of standard services were positively associated with a recovery orientation. Thus a polarized view of biomedical and recovery perspectives on mental illness may be unfounded, given that these perspectives appear to be mutually reinforcing.     

An exploratory survey of opportunities and barriers to creative leisure activity for people with learning disabilities

The present study examines an aspect of lifestyle that has received little attention, namely creative arts leisure occupations, and explores the opportunities and barriers to participation that confront community‐dwelling adults with learning disabilities. A survey of the managers of 34 residential homes in two major cities in the UK was carried out (reflecting a 54% return rate). They reported that art activities (including painting and pottery) were the most commonly available form of creative leisure occupation, and drama the least available. Creative occupations tended to occur in specialist settings, such as day centres, and a minority of residents participated in mainstream community groups and environments. Managers perceived the major barrier to creative leisure participation to be expense (i.e. for the activity itself and for support staff). Further barriers included insufficient staff to accommodate individual's personal interests, transport problems and unwelcoming community resources/attitudes. The results of the study indicate that many residents can engage in art within either mainstream and specialist resources. Other active creative occupations are not as accessible. Given its dependency on the voluntary return of questionnaires, the present survey may have provided a benign view of creative leisure provision. Even so, the barriers that were identified confirm some continuing problems with accessing community/mainstream facilities.     

A computer-controlled Y-maze for the analysis of vibrissotactile discrimination learning in mice.

An automatized, computer-controlled Y-maze is described in which mice are trained to discriminate between two vibrissotactile, and/or visual stimuli, other modalities or combinations of modalities being testable as well. Movements of the mouse are recorded by photocells and monitored on a computer screen. Forward passage of the mouse is ensured by movable gates, and, if necessary, by brief air blows. Wrong choices are punished by air blows as well. The controller is a Hewlett-Packard Series 80 microcomputer with a 16 channel parallel input/output interface; programs are in BASIC. The program analyzes start latency, decision and homing time, side preference, inspections and ultimate choice, as well as choice strategies based on discrimination, left/right habits and short-term memory. Thus we can determine the nature of discrimination errors, and establish individual behavioral profiles of the animals. Results are both printed alphanumerically and plotted. The apparatus may be used for studying sensory physiology as well as cerebral lateralization, and drug or gene effects on memory and learning.     

SEE: a tool for the visualization and analysis of rodent exploratory behavior

The complexity of exploratory behavior creates a need for a visualization and analysis tool that will highlight regularities and help generating new hypotheses about the structure of this behavior. The hypotheses can then be formulated as algorithms that capture the patterns and quantify them. SEE is a Mathematica based software developed by us for the exploration of exploratory behavior. The raw data for SEE are a time series of the animal 's coordinates in space sampled at a rate that allows a meaningful computation of speeds. SEE permits: (i) a visualization of the path of the animal and a computation of the dynamics of activity; (ii) a decomposition of the path into several modes of motion (1st gear, 2nd gear, etc.) and a computation of the typical maximal speeds, the spatial spread, and the proportion of each of these modes; and(iii) a visualization of the location in the environment of stopping episodes, along with their dwell time. These visualizations highlight the presence of preferred places, including the animal's so-called home base, and permits a computation of the spatio-temporal diversity in the location of stopping episodes. The software also: (i) decomposes the animal's path into round trips from the home base, called 'excursions', and computes the number of stops per excursion; (ii) generates a visualization of the phase space (path+speed, traced in a three-dimensional graph) of any progression segment or list of such segments; and (iii) produces a visualization of the way places in the animal's operational world are connected to each other. SEE also permits the definition and computation of behavioral endpoints across any section of any database of raw data. The range of applicability of SEE to various experimental set ups, tracking procedures, species, and preparations is addressed in the discussion.     

A dynamical model for the analysis and acceleration of learning in feedforward networks.

A dynamical system model is derived for feedforward neural networks with one layer of hidden nodes. The model is valid in the vicinity of flat minima of the cost function that rise due to the formation of clusters of redundant hidden nodes with nearly identical outputs. The derivation is carried out for networks with an arbitrary number of hidden and output nodes and is, therefore, a generalization of previous work valid for networks with only two hidden nodes and one output node. The Jacobian matrix of the system is obtained, whose eigenvalues characterize the evolution of learning. Flat minima correspond to critical points of the phase plane trajectories and the bifurcation of the eigenvalues signifies their abandonment. Following the derivation of the dynamical model, we show that identification of the hidden nodes clusters using unsupervised learning techniques enables the application of a constrained application (Dynamically Constrained Back Propagation-DCBP) whose purpose is to facilitate prompt bifurcation of the eigenvalues of the Jacobian matrix and, thus, accelerate learning. DCBP is applied to standard benchmark tasks either autonomously or as an aid to other standard learning algorithms in the vicinity of flat minima. Its application leads to significant reduction in the number of required epochs for convergence.     

Uncovering shape signatures of resting‐state functional connectivity by geometric deep learning on Riemannian manifold

Functional neural activities manifest geometric patterns, as evidenced by the evolving network topology of functional connectivities (FC) even in the resting state. In this work, we propose a novel manifold‐based geometric neural network for functional brain networks (called “Geo‐Net4Net” for short) to learn the intrinsic low‐dimensional feature representations of resting‐state brain networks on the Riemannian manifold. This tool allows us to answer the scientific question of how the spontaneous fluctuation of FC supports behavior and cognition. We deploy a set of positive maps and rectified linear unit (ReLU) layers to uncover the intrinsic low‐dimensional feature representations of functional brain networks on the Riemannian manifold taking advantage of the symmetric positive‐definite (SPD) form of the correlation matrices. Due to the lack of well‐defined ground truth in the resting state, existing learning‐based methods are limited to unsupervised methodologies. To go beyond this boundary, we propose to self‐supervise the feature representation learning of resting‐state functional networks by leveraging the task‐based counterparts occurring before and after the underlying resting state. With this extra heuristic, our Geo‐Net4Net allows us to establish a more reasonable understanding of resting‐state FCs by capturing the geometric patterns (aka. spectral/shape signature) associated with resting states on the Riemannian manifold. We have conducted extensive experiments on both simulated data and task‐based functional resonance magnetic imaging (fMRI) data from the Human Connectome Project (HCP) database, where our Geo‐Net4Net not only achieves more accurate change detection results than other state‐of‐the‐art counterpart methods but also yields ubiquitous geometric patterns that manifest putative insights into brain function.     

Neural independent component analysis by 'maximum-mismatch' learning principle.

The aim of the present paper is to apply Sudjanto-Hassoun theory of Hebbian learning to neural independent component analysis. The basic learning theory is first recalled and expanded in order to make it suitable for a network of non-linear complex-weighted neurons; then its interpretation and application is shown in the context of blind separation of complex-valued sources. Numerical results are given in order to assess the effectiveness of the proposed learning theory and the related separation algorithm on telecommunication signals; a comparison with other existing techniques finally helps assessing the performances and computational requirements of the proposed algorithm.     

Genetic dissection of mouse exploratory behaviour.

A large variety of apparatus and procedures are being employed to measure mouse exploratory behaviour. Definitions of what constitutes exploration also vary widely. The present article reviews two studies whose results permit a genetic dissection of behaviour displayed in an open-field situation. The results agree that factors representing exploration and stress/fear underlie this type of behaviour. Both factors appear to be linked to neuroanatomical variation in the sizes of the hippocampal intra- and infrapyramidal mossy fibre terminal fields. Multivariate analysis of genetic correlations may render important insights into the structure of behaviour and its relations with neuroanatomical and neurophysiological systems.     

An exploratory analysis of behavioral health care use within families

OBJECTIVES: The authors studied enrollees in employer-sponsored managed health plans to determine the extent of multiple behavioral health consumers within families, use of behavioral health services by employees who have another family member using such services, congruence of diagnoses between two family members using behavioral health services, and the effect on covered charge per behavioral health consumer when other family members use behavioral health services. METHODS: Claims data from 911 plans sold or managed by a single managed behavioral health care company were examined. The plans provided coverage for 724,789 employees and covered about 1.7 million lives. Family members of employees were identified by the relationship codes on the claims. Service utilization rates were calculated for employees overall and for employees who had spouses or children who used behavioral health services. Mean and median covered charges were determined and were examined by number and type of consumers in the family. RESULTS: The use of behavioral health services was greater among employees whose children or spouses used behavioral health services. Utilization rates varied by the child's or spouse's diagnoses. More than 50 percent of male employees whose children received treatment for a depressive disorder also received such treatment. Congruence of diagnoses within families was noted. Covered charges per person generally increased with the number of family members who used behavioral health services. CONCLUSION: Greater knowledge about patterns of use of behavioral health services within families may help in improving access to care and developing more effective family interventions.     

Sexual socialization in lesbian-parent families: an exploratory analysis

Few studies have examined the sexual socialization of children within lesbian-parent families, despite evidence that these children may experience benefits in this regard. Qualitative interviews were conducted with 10 partnered, lesbian mothers to explore what and how participants taught their children about sexuality-related issues. Themes related to what participants taught their children about these topics included diverse notions of sexual orientation and reproduction. Themes related to how participants taught these concepts included tag-teaming with their partners (i.e., participants shared with their partners the task of teaching their children about sexuality-related issues), although differences in how the partners carried out this task were identified. This study demonstrated a diversity of experiences of lesbian-parent families, often seemingly in relation to family contextual factors (e.g., how children were conceived), and has potential implications for family researchers, practitioners, and, perhaps, all parents.     

An exploratory factor analysis of the children's depression rating scale-revised

An exploratory factor analysis was performed in a clinical sample of 314 children and adolescents to investigate the factor structure of the Children's Depression Rating Scale-Revised (CDRS-R; Poznanski et al. 1984). A maximum likelihood method followed by a Promax rotation yielded five factors: observed depressive mood, anhedonia, morbid thoughts, somatic symptoms and reported depressive mood. The age group and gender differences on the factors scores are evaluated. After controlling for gender, the adolescents had more severe depression in terms of observed depressive mood, anhedonia, and somatic symptoms. After controlling for age groups, girls had higher scores for reported depressive mood.     

New metrics for analysis of dendritic branching patterns demonstrating similarities and differences in ON and ON-OFF directionally selective retinal ganglion cells.

The morphology and dendritic branching patterns of retinal ganglion cells have been studied in Golgi-impregnated, whole-mount preparations of rabbit retina. Among a large number of morphological types identified, two have been found that correspond to the morphology of ON and ON-OFF directionally selective (DS) ganglion cells identified in other studies. These two kinds of DS ganglion cell are compared with each other, as well as with examples of class I, class II, and class III cells, defined here with reference to our previous studies. Cell body, dendritic field size and branching pattern are analyzed in this paper and levels of dendritic stratification are examined in the following paper. ON DS ganglion cells are about 10% larger in soma size and about 5 times the dendritic field area of ON-OFF DS ganglion cells, when compared at the same retinal location. These two morphological types of ganglion cell can be said to define the upper and lower bounds of an intermediate range of cell body and dendritic field sizes within the whole population of ganglion cells. Nevertheless, in previous physiological studies receptive field sizes of the two types were shown to be similar. This discrepancy between morphological and physiological evidence is considered in the Discussion in terms of a model of the excitatory receptive field of ON-OFF DS ganglion cells incorporating starburst amacrine cells. A new set of metrics is introduced here for the quantitative analysis and characterization of the branching pattern of neuronal arborizations. This method compares the lengths of terminal and preterminal dendritic branches (treated separately), as a function of the distances of their origins from the soma, viewed graphically in a two-dimensional scatter plot. These values are derived from computer-aided 3D logging of the dendritic trees, and distance from the soma is measured as the shortest distance tracked along the dendritic branches. From these metrics of the "branch length distributions," scale-independent branching statistics are derived. These make use of mean branch lengths and distances, slopes of lines fitted to the distributions, and elliptical indices of scatter in the distributions. By these measures, ON and ON-OFF DS ganglion cells have similar branching patterns, which they share to varying degrees with functionally unrelated class III.1 ganglion cells. The scale of the branching patterns of ON and ON-OFF DS cells and their degree of uniformity are different, however.(ABSTRACT TRUNCATED AT 400 WORDS)