Search PubMed⌕ Search

SEARCH · Search PubMed

Results for “representation learning”

Search indexed PubMed citations on genomics, clinical trials, systematic reviews and public health. Explore titles, authors and supplied subject terms, then open the PubMed record.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 1,117 records · Page 62Linked to original sources
PubMed1988

Looking at faces: first-order and second-order features as determinants of facial appearance.

The encoding and relative importance of first-order (discrete) and second-order (configural) features in mental representations of unfamiliar faces have been investigated. Nonmetric multidimensional scaling (KYST) was carried out on similarity judgments of forty-one photographs of faces (homogeneous with respect to sex, race, facial expression, and, to a lesser extent, age). A large set of ratings, measurements, and ratios of measurements of the faces was regressed against the three-dimensional KYST solution in order to determine the first-order and second-order features used to judge similarity. Parameters characterizing both first-order and second-order features emerged as important determinants of facial similarity. First-order feature parameters characterizing the appearance of the eyes, eyebrows, and mouth, and second-order feature parameters characterizing the position of the eyes, spatial relations between the internal features, and chin shape correlated with the dimensions of the KYST solution. There was little difference in the extent to which first-order and second-order features were encoded. Two higher-level parameters, age and weight, were also used to judge similarity. The implications of these results for mental representations of faces are discussed.

Attention↗
PubMed2001

Principle explanation and strategic schema abstraction in problem solving.

This study was undertaken to examine the effects of strategic schema-acquisition tasks (problem comparison or problem construction) and the method of principle explanation (abstract or embedded principle method) on schema acquisition. Ninety-eight subjects studied a set of problems in probability, presented according to either method of principle explanation. Half the subjects in each principle-explanation group were then asked to compare analogous problems, and the rest constructed new analogous problems. To determine whether subjects generalized problem schemas, they were given new analogous problems to solve. The results showed that when the abstract principle method was used, schema acquisition was better in problem comparison; but with the embedded principle method, schema acquisition was better in problem construction. Results were discussed in relation to the importance of some fit between the presentation of problem information and the processes that will draw from or build on this information in tasks designed to allow novice problem solvers to acquire advanced problem representations.

Adolescent↗
PubMed1991

Perception of three-dimensional angular rotation.

In three experiments, difference thresholds (dLs) and points of subjective equality (PSEs) for three-dimensional (3-D) rotation simulations were examined. In the first experiment, observers compared pairs of simulated spheres that rotated in polar projection and that differed in their structure (points plotted in the volume vs. on the surface), axis of rotation (vertical, y, vs. horizontal, x), and magnitude of rotation (20 degrees-70 degrees). DLs were lowest (7%) when points were on the surface and when at least one sphere rotated around the y-axis and varied with changes in the independent variables. PSEs were closest to objective equality when points were on the surface of both spheres and when both spheres rotated about the x-axis. In the second experiment, subjects provided direct estimates of the rotations of the same spheres. Results suggested a reasonable agreement between PSEs for the indirect-scaling and direct-estimate procedures. The third experiment varied sphere diameter (and therefore mean linear velocity of stimulus elements) and showed that although rotation judgments are biased by mean linear velocity, they are not likely to be made solely on the basis of that information. These and past results suggest a model whereby recovery of structure is conducted by low-level motion-detecting mechanisms, whereas rotation (and other) judgments are based on a higher level representation.

Acceleration↗
PubMed2006

Brain plasticity and functional losses in the aged: scientific bases for a novel intervention.

Aging is associated with progressive losses in function across multiple systems, including sensation, cognition, memory, motor control, and affect. The traditional view has been that functional decline in aging is unavoidable because it is a direct consequence of brain machinery wearing down over time. In recent years, an alternative perspective has emerged, which elaborates on this traditional view of age-related functional decline. This new viewpoint--based upon decades of research in neuroscience, experimental psychology, and other related fields--argues that as people age, brain plasticity processes with negative consequences begin to dominate brain functioning. Four core factors--reduced schedules of brain activity, noisy processing, weakened neuromodulatory control, and negative learning--interact to create a self-reinforcing downward spiral of degraded brain function in older adults. This downward spiral might begin from reduced brain activity due to behavioral change, from a loss in brain function driven by aging brain machinery, or more likely from both. In aggregate, these interrelated factors promote plastic changes in the brain that result in age-related functional decline. This new viewpoint on the root causes of functional decline immediately suggests a remedial approach. Studies of adult brain plasticity have shown that substantial improvement in function and/or recovery from losses in sensation, cognition, memory, motor control, and affect should be possible, using appropriately designed behavioral training paradigms. Driving brain plasticity with positive outcomes requires engaging older adults in demanding sensory, cognitive, and motor activities on an intensive basis, in a behavioral context designed to re-engage and strengthen the neuromodulatory systems that control learning in adults, with the goal of increasing the fidelity, reliability, and power of cortical representations. Such a training program would serve a substantial unmet need in aging adults. Current treatments directed at age-related functional losses are limited in important ways. Pharmacological therapies can target only a limited number of the many changes believed to underlie functional decline. Behavioral approaches focus on teaching specific strategies to aid higher order cognitive functions, and do not usually aspire to fundamentally change brain function. A brain-plasticity-based training program would potentially be applicable to all aging adults with the promise of improving their operational capabilities. We have constructed such a brain-plasticity-based training program and conducted an initial randomized controlled pilot study to evaluate the feasibility of its use by older adults. A main objective of this initial study was to estimate the effect size on standardized neuropsychological measures of memory. We found that older adults could learn the training program quickly, and could use it entirely unsupervised for the majority of the time required. Pre- and posttesting documented a significant improvement in memory within the training group (effect size 0.41, p<0.0005), with no significant within-group changes in a time-matched computer using active control group, or in a no-contact control group. Thus, a brain-plasticity-based intervention targeting normal age-related cognitive decline may potentially offer benefit to a broad population of older adults.

Aged↗
PubMed1996

Merging Back-propagation and Hebbian Learning Rules for Robust Classifications.

By imposing saturation requirements on hidden-layer neural activations, a new learning algorithm is developed to improve robustness on classification performance of a multi-layer Perceptron. Derivatives of the sigmoid functions at hidden-layers are added to the standard output error with relative significance factors, and the total error is minimized by the steepest-descent method. The additional gradient-descent terms become Hebbian, and this new algorithm merges two popular learning algorithms, i.e., error back-propagation and Hebbian learning rules. Only slight modifications are needed for the standard back-propagation algorithm, and additional computational requirements are negligible. This saturation requirement effectively reduces output sensitivity to the input, which results in improved robustness and better generalization for classifier networks. Also distributed representations at hidden-layers are successfully suppressed to accomplish efficient utilization of hidden neurons. Computer simulations demonstrates much faster learning convergence as well as improved robustness for classifications and hetero-associations of binary patterns. Copyright 1996 Elsevier Science Ltd

Journal Article↗
PubMed1994

Significance of distributed representation in the output layer of a neural network in a pattern recognition task.

In the cerebral cortex, it is assumed that information is represented by the activity pattern of an assembly of neurons and the synaptic efficacies among them. A distributed representation of pattern is incorporated in the output layer of a neural network with an error back-propagation algorithm, in order to study its technological merits. The network has three layers, which consist of a 32 x 32 array of units (1024) for the input layer, 6-25 units for the hidden layer and 12 units for the output layer. 12 triangular patterns with a variety of parameters are presented to the input layer. Three output-layer units are assigned to each input figure. After initial learning, the network responds to the learned pattern with high accuracy. In addition, it responds with high accuracy to similar but unpresented patterns, showing a generalisation for patterns. The network shows resistance to unit de-activation procedures. When the input layer is exposed to the learned pattern, the hidden-layer units show associative activation pattern. These results indicate that the organisation of information representation in the output layer in a neural network strongly influences both the performance of the whole network and information representation in the hidden layer.

Association↗
PubMed2006

Symbols as self-emergent entities in an optimization process of feature extraction and predictions.

In the mammalian cortex the early sensory processing can be characterized as feature extraction resulting in local and analogue low-level representations. As a direct consequence, these map directly to the environment, but interpretation under natural conditions is ambiguous. In contrast, high-level representations for cognitive processing, e.g. language, require symbolic representations characterized by expression and syntax. The representations are binary, structured and disambiguated. However, do these fundamental functional distinctions translate into a fundamental distinction of the respective brain areas and their anatomical and physiological properties? Here we argue that the distinction between early sensory processing and higher cognitive functions may not be based on structural differences of cortical areas; instead similar learning principles acting on input signals with different statistics give rise to the observed variations of function. Firstly, we give an account of present research describing neuronal properties at early stages of sensory systems as a consequence of an optimization process over the set of natural stimuli. Secondly, addressing a stage following early visual processing we suggest to extend the unsupervised learning scheme by including predictive processes. These contain the widely used objective of temporal coherence as a special case and are a powerful approach to resolve ambiguities. Furthermore, in combination with a prior on the bandwidth of information exchange between units it leads to a condensation of information. Thirdly, as a crucial step, not only are predictive units optimized, but the selectivity of the feature extractors are adapted to allow optimal predictability. Thus, over and beyond making useful predictions, we propose that the predictability of a stimulus be in itself a selection criterion for further processing. In a hierarchical system the combined optimization process leads to entities that represent condensed pieces of knowledge and that are not analogue anymore. Instead, these entities work as arguments in a framework of transformations that realize predictions. Thus, the criteria of predictability and condensation in an optimization of sensory representations relate directly to the two defining properties of symbols of expression and syntax. In this paper, we sketch an unsupervised learning process that gradually transforms analogue local representations into discrete binary representations by means of four hypotheses. We propose that in this optimization process acting in a hierarchical system, entities emerge at, higher levels that fulfil the criteria defining symbols, instantiating qualitatively different representations at similarly structured low and high levels.

Animals↗
PubMed2000

Computational approaches to sensorimotor transformations.

Behaviors such as sensing an object and then moving your eyes or your hand toward it require that sensory information be used to help generate a motor command, a process known as a sensorimotor transformation. Here we review models of sensorimotor transformations that use a flexible intermediate representation that relies on basis functions. The use of basis functions as an intermediate is borrowed from the theory of nonlinear function approximation. We show that this approach provides a unifying insight into the neural basis of three crucial aspects of sensorimotor transformations, namely, computation, learning and short-term memory. This mathematical formalism is consistent with the responses of cortical neurons and provides a fresh perspective on the issue of frames of reference in spatial representations.

Animals↗
PubMed2000

The persistence of structural priming: transient activation or implicit learning?

Structural priming in language production is a tendency to recreate a recently uttered syntactic structure in different words. This tendency can be seen independent of specific lexical items, thematic roles, or word sequences. Two alternative proposals about the mechanism behind structural priming include (a) short-term activation from a memory representation of a priming structure and (b) longer term adaptation within the cognitive mechanisms for creating sentences, as a form of procedural learning. Two experiments evaluated these hypotheses, focusing on the persistence of structural priming. Both experiments yielded priming that endured beyond adjacent sentences, persisting over 2 intervening sentences in Experiment 1 and over 10 in Experiment 2. Although memory may have short-term consequences for some components of this kind of priming, the persisting effects are more compatible with a learning account than a transient memory account.

Adult↗
PubMed2005

Does swarming cause honey bees to update their solar ephemerides?

Spatial orientation in the social insects offers several examples of specialized learning mechanisms that underlie complex learning tasks. Here we study one of these systems: the processes by which honey bees update, or fail to update, their memories of the sun's daily pattern of movement (the solar ephemeris function) in relation to the landscape. Specifically, we ask whether bees that have initially learned the solar ephemeris function relative to a conspicuous treeline at their natal site can later realign the ephemeris to a differently oriented treeline. We first confirm and clarify an earlier finding that bees transplanted passively (by being carried) do not re-learn the solar ephemeris in relation to the new treeline. When they cannot detect the sun directly, as on overcast days, these transplanted bees use a solar ephemeris function appropriate for their natal site, despite days or weeks of experience at the new site. We then ask whether bees put through a swarming process as they are transplanted are induced to re-learn the solar ephemeris function at the new site, as swarming is a natural process wherein bees transplant themselves. Most of the swarmed bees failed to re-learn, even though they did extensive learning flights (in comparison with those of non-swarmed controls) as they first emerged from the hive at the new site. We hypothesize that the bees' representation of the solar ephemeris function is stored in an encapsulated cognitive module in which the ephemeris is inextricably linked to the reference landscape in which it was learned.

Animal Communication↗
PubMed1989

Monitors: key mechanisms and roles in the development and aging of the consciousness and self.

A network of interacting neural structures, called monitors, exists in the mammalian brain in which data derived from sensory inputs and from memory stores is precisely displayed within the brain. The key function of monitors is to provide an 'ultimate monitor', proposed to be the locus that generates the phenomenon of conscious self awareness, with information that defines or maps the positions of the parts of an individual with respect to each other and with respect to external objects or events at specific times. The resolution of at least some of these monitors (e.g. some concerned with vision) is extremely great and approaches, in the case of vision, the precision with which images of external objects are projected onto the retina. This conclusion is based on the fact that an individual is able to perceive visual images with an acuity that closely approximates the fineness of resolution of the retinal image. The sensory signals that provide information about body part positions and those that provide information about the exterior are evidently integrated with each other in a suitable hierarchy of monitors so as to provide a coherent representation of self-vs.-environment. The logical 'framework' monitor for this integrated display-mapping is proposed to be that that maps the body in space and it is proposed that the locations of objects perceived through the touch sense and senses that deal with more remote items in the environment become superimposed on a map that extends or extrapolates the body space map beyond the body's physical boundaries, a learning process that occurs during development. The ultimate monitor not only receives a display of the synthetic representations derived currently through the integrative functions defined above, but also is provided with at least four other inputs from other different classes of monitors. One of these is a monitoring system that generates timing signals needed to separate inputs into a time order and to assign an order to them. It is proposed that it is awareness of these timing signals by the ultimate monitor that is the essential and indispensible input that generates the phenomenon of awareness. A second input to the monitor that is the self is a selected part of its own activities. This awareness of what the ultimate monitor is receiving, doing or planning to do in the future is the characteristic necessary for awareness of self.(ABSTRACT TRUNCATED AT 400 WORDS)

Aging↗
PubMed2004

Identification of humans using gait.

We propose a view-based approach to recognize humans from their gait. Two different image features have been considered: the width of the outer contour of the binarized silhouette of the walking person and the entire binary silhouette itself. To obtain the observation vector from the image features, we employ two different methods. In the first method, referred to as the indirect approach, the high-dimensional image feature is transformed to a lower dimensional space by generating what we call the frame to exemplar (FED) distance. The FED vector captures both structural and dynamic traits of each individual. For compact and effective gait representation and recognition, the gait information in the FED vector sequences is captured in a hidden Markov model (HMM). In the second method, referred to as the direct approach, we work with the feature vector directly (as opposed to computing the FED) and train an HMM. We estimate the HMM parameters (specifically the observation probability B) based on the distance between the exemplars and the image features. In this way, we avoid learning high-dimensional probability density functions. The statistical nature of the HMM lends overall robustness to representation and recognition. The performance of the methods is illustrated using several databases.

Algorithms↗
PubMed2006

Information and communication technologies in higher education: evidence-based practices in medical education.

In contrast to traditional meta-analyses of research, an alternative overview and analysis of the research literature on the impact of information and communication technologies (ICT) in medical education is presented in this article. A distinction is made between studies that have been set up at the micro-level of the teaching and learning situation and studies on meso-level issues. At the micro-level, ICT is hypothesized to foster three basic information processing activities: presentation, organization, and integration of information. Next to this, ICT is expected to foster collaborative learning in the medical knowledge domain. Empirical evidence supports the potential of ICT to introduce students to advanced graphical representations but the studies also stress the importance of prior knowledge and the need for real-life tactile and practical experiences. The number of empirical studies focusing on the impact of ICT on information organization is restricted but the results suggest a positive impact on student attitudes and relevant learning gains. However, again, students need a relevant level of prior knowledge. Empirical studies focusing on the impact of ICT on information integration highlight the positive impact of ICT-based assessment and computer simulations; for the latter this is especially the case when novices are involved, and when they master the prerequisite ICT skills. Little empirical evidence is available regarding the impact of computer games. Research results support the positive impact of ICT-based collaboration but care has to be taken when skills development is pursued. At the meso-level, the available empirical evidence highlights the positive impact of ICT to promote the efficiency of learning arrangements. Research grounds the key position of ICT in a state-of-the-art medical curriculum. Recent developments focusing on repositories of learning materials for medical education have yet not been evaluated. The article concludes by stressing the need for evaluative studies, especially in the promising field of ICT-based collaborative learning. Furthermore, the importance to be attached to the position and qualifications of the teaching staff is emphasized.

Belgium↗
PubMedDate not supplied

Perception and awareness in phonological processing: the case of the phoneme.

The necessity of a "levels-of-processing" approach in the study of mental representations is illustrated by the work on the psychological reality of the phoneme. On the basis of both experimental studies of human behavior and functional imaging data, it is argued that there are unconscious representations of phonemes in addition to conscious ones. These two sets of mental representations are functionally distinct: the former intervene in speech perception and (presumably) production; the latter are developed in the context of learning alphabetic literacy for both reading and writing purposes. Moreover, among phonological units and properties, phonemes may be the only ones to present a neural dissociation at the macro-anatomic level. Finally, it is argued that even if the representations used in speech perception and those used in assembling and in conscious operations are distinct, they may entertain dependency relations.

Awareness↗
PubMed2000

Motor learning through the combination of primitives.

In this paper we discuss a new perspective on how the central nervous system (CNS) represents and solves some of the most fundamental computational problems of motor control. In particular, we consider the task of transforming a planned limb movement into an adequate set of motor commands. To carry out this task the CNS must solve a complex inverse dynamic problem. This problem involves the transformation from a desired motion to the forces that are needed to drive the limb. The inverse dynamic problem is a hard computational challenge because of the need to coordinate multiple limb segments and because of the continuous changes in the mechanical properties of the limbs and of the environment with which they come in contact. A number of studies of motor learning have provided support for the idea that the CNS creates, updates and exploits internal representations of limb dynamics in order to deal with the complexity of inverse dynamics. Here we discuss how such internal representations are likely to be built by combining the modular primitives in the spinal cord as well as other building blocks found in higher brain structures. Experimental studies on spinalized frogs and rats have led to the conclusion that the premotor circuits within the spinal cord are organized into a set of discrete modules. Each module, when activated, induces a specific force field and the simultaneous activation of multiple modules leads to the vectorial combination of the corresponding fields. We regard these force fields as computational primitives that are used by the CNS for generating a rich grammar of motor behaviours.

Animals↗
PubMed1996

Learning multiple causes by competition enhanced least mean square error reconstruction.

In this paper we studied a self-organization principle that input should be best reconstructed from a factorial distributed hidden representation, which has been addressed in the literature recently. An auto-encoder network is trained by the Least Mean Square Error Reconstruction (LMSER) while the redundance in the representation is reduced by a proposed anti-Hebbian scheme, in which a penalty term called Receptive Field Overlapping Index (RFOI) is combined into the objective function for enhancing competition among nodes in the network. Our learning scheme provides a way for balancing the cooperation and competition necessary for the self-organization process thus realizes the multiple causes model, which accounts for an observed data by combining assertions from the discovered causes or features in the data. Our experiment results demonstrate again the powerful information processing capability inherent to the popular weighted sum followed by sigmoid squashing. Comparing with previous probability theory based multiple causes models, our scheme is much easier to implement and quite reliable.

Algorithms↗
PubMed2005

Emphatic, interactive volume rendering to support variance in user expertise.

Various levels of representation, from abstract to schematic to realistic, have been exploited for millennia to facilitate the transfer of information from one individual to another. Learning complex information, such as that found in biomedicine, proves specifically problematic to many, and requires incremental, step-wise depictions of the information to clarify structural, functional, and procedural relationships.Emerging volume-rendering technique, such as non-photorealistic representation, coupled with advances in computational speeds, especially new graphical processing units, provide unique capabilities to explore the use of various levels of representation in interactive sessions. We have developed a system that produces images that simulate pictorial representations for both scientific and biomedical visualization. The system combines traditional and novel volume illustration techniques. We present examples from our efforts to distill representational techniques for both creative exploration and emphatic presentation for clarity. More specifically, we present our efforts to adapt these techniques for interactive simulation sessions being developed in a concurrent project for resident training in temporal bone dissection simulation. The goal of this effort is to evaluate the use of emphatic rendering to guide the user in an interactive session and to facilitate the learning of complex biomedical information, including structural, functional, and procedural information.

Computational Biology↗
PubMed2004

Doing without schema hierarchies: a recurrent connectionist approach to normal and impaired routine sequential action.

In everyday tasks, selecting actions in the proper sequence requires a continuously updated representation of temporal context. Previous models have addressed this problem by positing a hierarchy of processing units, mirroring the roughly hierarchical structure of naturalistic tasks themselves. The present study considers an alternative framework, in which the representation of context depends on recurrent connections within a network mapping from environmental inputs to actions. The ability of this approach to account for human performance was evaluated by applying it, through simulation, to a specific everyday task. The resulting model learned to deal flexibly with a complex set of sequencing constraints, encoding contextual information at multiple time scales within a single, distributed internal representation. Degrading this representation led to errors resembling those observed both in everyday behavior and in apraxia. Analysis of the model's function yielded numerous predictions relevant to both normal and apraxic performance.

Cognition↗