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 955 records · Page 53Linked to original sources

Sleep to remember.

Recently, compelling evidence has accumulated that links sleep to learning and memory. Sleep has been identified as a state that optimizes the consolidation of newly acquired information in memory. Consolidation is an active process that is presumed to rely on the covert reactivation and reorganization of newly encoded representations. Hippocampus-dependent memories benefit primarily from slow-wave sleep (SWS), whereas memories not depending on the hippocampus show greater gains over periods containing high amounts of rapid eye movement sleep. One way sleep does this is by establishing different patterns of neurotransmitters and neurohormone secretion between sleep stages. Another central role for consolidating memories is played by the slow oscillation, that is, the oscillating field potential change dominating SWS. The emergence of slow oscillations in neocortical networks depends on the prior use of these networks for encoding of information. Via efferent pathways, they synchronize the occurrence of sharp wave ripples accompanying memory reactivations in the hippocampus with thalamocortical spindle activity. Thus, hippocampal memories are fed back into neocortical networks at a time when these networks are depolarized and, because of concurrent spindle activity, can most sensitively react to these inputs with plastic changes underlying the formation of long-term memory representations.

Animals↗

Representation of serial order in humans: a comparison to the findings with monkeys (Cebus apella).

In a number of studies, serially organized behavior in humans has been examined using a procedure developed for use with pigeons and monkeys. There have been few direct comparisons, however, between the data collected with humans and that collected with nonhumans, and none with respect to the interesting latency effects noted with nonhumans. The purpose of this experiment was to make this comparison. Human subjects were trained to respond to five simultaneously presented stimuli (A, B, C, D, and E) in a specific order (A-->B-->C-->D-->E) and were then tested with all 10 pairwise combinations of the five stimuli, followed by all 10 triplet combinations of the five stimuli. Mirroring the findings with monkeys (Cebus apella), humans showed a first-item effect, a missing-item effect, and a symbolic-distance effect. These results suggest that during the course of learning the five-item serial-order task humans form an internal representation of the series and access that representation to guide their behavior.

Adolescent↗

Learning directions of objects specified by vision, spatial audition, or auditory spatial language.

The modality by which object azimuths (directions) are presented affects learning of multiple locations. In Experiment 1, participants learned sets of three and five object azimuths specified by a visual virtual environment, spatial audition (3D sound), or auditory spatial language. Five azimuths were learned faster when specified by spatial modalities (vision, audition) than by language. Experiment 2 equated the modalities for proprioceptive cues and eliminated spatial cues unique to vision (optic flow) and audition (differential binaural signals). There remained a learning disadvantage for spatial language. We attribute this result to the cost of indirect processing from words to spatial representations.

Adult↗

Navigation around London by a taxi driver with bilateral hippocampal lesions.

The time-scale of hippocampal involvement in supporting episodic memory remains a keenly debated topic, with disagreement over whether its role is temporary or permanent. Recently, there has been interest in how navigation by hippocampally-compromised patients in environments learned long ago speaks to this issue. However, identifying patients with damage that is primarily hippocampal, control subjects matched for navigation experience, and testing their in situ navigation, present substantial problems. We met these challenges by using a highly accurate and interactive virtual reality simulation of central London (UK) to assess the navigation ability of a licensed London taxi driver who had sustained bilateral hippocampal damage. In this test, patient TT and matched control taxi drivers drove a virtual London taxi along the streets they had first learned 40 years before. We found that the hippocampus is not required for general orientation in the city either in first person or survey perspectives, detailed topographical knowledge of landmarks and their spatial relationships, or even for active navigation along some routes. However, in his navigation TT was very reliant on main artery or 'A' roads, and became lost when navigation depended instead on non-A roads. We conclude that the hippocampus in humans is necessary for facilitating navigation in places learned long ago, particularly where complex large-scale spaces are concerned, and successful navigation requires access to detailed spatial representations.

Aged↗

Environment-spatial conditional learning in rats with selective lesions of medial septal cholinergic neurons.

Cholinergic medial septal neurons may regulate several aspects of hippocampal function, including place field stability and spatial working memory. Monkeys with damage to septal cholinergic neurons are impaired in visual-spatial conditional learning tasks; however, this candidate function of septal cholinergic neurons has not been studied extensively in the rat. In the present study, rats with selective lesions of cholinergic neurons in the medial septum and vertical limb of the diagonal band of Broca (MS/VDB), made with 192 IgG-saporin, were tested on a conditional associative learning task. In this task, which we term "environment-spatial" conditional learning, the correct location of a spatial response depended on the array of local environmental cues. MS/VDB-lesioned rats were impaired when the two parts of the conditional problem were presented concurrently, but not when one environment had been learned before the full conditional problem was presented. Our findings suggest that cholinergic MS/VDB neurons participate in some aspects of conditional associative learning in rats. They may also shed light on the involvement of cholinergic projections to the hippocampus in modulating and remodeling hippocampal spatial representations.

Animals↗

Late-training amnesic deficits in probabilistic category learning: a neurocomputational analysis.

Building upon earlier behavioral models of animal and human learning, we explore how a psychobiological model of animal conditioning can be applied to amnesic category learning. In particular, we show that the late-training deficit found in Knowlton, Squire, and Gluck's 1994 study of amnesic category learning can be understood as a natural consequence of Gluck and Myers's (1993) theory of hippocampal-region function, a theory that has heretofore been applied only to studies of animal learning. When applied to Knowlton et al.'s category learning task, Gluck and Myers's model assumes that the hippocampal region induces new stimulus representations over multiple training trials that reflect stimulus-stimulus regularities in the training set. As such, the model expects an advantage for control subjects over hippocampal-damaged amnesic patients only later in training when control subjects have developed new hippocampal-dependent stimulus representations; in contrast, both groups are expected to show equivalent performance early in training. A potentially analogous early/late distinction is described for animal studies of stimulus generalization. Our analyses suggest that careful comparisons between early and late-training differences in learning may be an important factor in understanding amnesia and the neural bases of both animal and human learning.

Amnesia↗

The child's route into reading and what can go wrong.

Two strands of linguistic development critically important for successful reading acquisition are outlined. One of these ontogenetic roots concerns phonological development and projects onto word decoding. The other root concerns the development of vocabulary and syntax projecting onto reading comprehension. Language development starts very early in infancy when the child learns to categorize the speech sounds according to the pattern typical of the mother tongue. Equipped with these sound categories the child is ready to learn to understand and to use new words. The precise phonological representation of words will facilitate the important development of phonological awareness. The paper reviews some longitudinal research and training studies indicating the causal direction of the relation between phonological awareness and reading. Preventive implications are pointed out. The preventive power of the second strand of development is discussed with reference to vocabulary exposure in different social environments and to informal early literacy socialization. In particular, the benefits of the prototypical situation of reading aloud to children are discussed. Finally, the complexity of the causal relationships between different aspects of early language development and later reading is pointed out.

Awareness↗

Numerical bifurcation analysis of distance-dependent on-center off-surround shunting neural networks.

On-center off-surround shunting neural networks are often applied as models for content-address-able memory (CAM), the equilibria being the stored memories. One important demand of biological plausible CAMs is that they function under a broad range of parameters, since several parameters vary due to postnatal maturation or learning. Ellias, Cohen and Grossberg have put much effort into showing the stability properties of several configurations of on-center off-surround shunting neural networks. In this article we present numerical bifurcation analysis of distance-dependent on-center off-surround shunting neural networks with fixed external input. We varied four parameters that may be subject to postnatal maturation: the range of both excitatory and inhibitory connections and the strength of both inhibitory and excitatory connections. These analyses show that fold bifurcations occur in the equilibrium behavior of the network by variation of all four parameters. The most important result is that the number of activation peaks in the equilibrium behavior varies from one to many if the range of inhibitory connections is decreased. Moreover, under a broad range of the parameters the stability of the network is maintained. The examined network is implemented in an ART network, Exact ART, where it functions as the classification layer F2. The stability of the ART network with the F2-field in different dynamic regimes is maintained and the behavior is functional in Exact ART. Through a bifurcation the learning behavior of Exact ART may even change from forming local representations to forming distributed representations.

Cybernetics↗

Infants' emerging ability to represent occluded object motion.

The emerging ability to represent an oscillating moving object over occlusions was studied in 7-21-week-old infants. The object moved at 0.25 Hz and was either occluded at the center of the trajectory (for 0.3 s) or at one turning point (for 0.7 s). Each trial lasted for 20 s. Both eye and head movements were measured. By using two kinds of motion, sinusoidal (varying velocity) and triangular (constant velocity), infants' ability to take velocity change into account when predicting the reappearance of the moving object was tested. Over the age period studied, performance at the central occluder progressed from almost total ignorance of what happened to consistent predictive behavior. From around 12 weeks of age, infants began to form representations of the moving object that persisted over temporary occlusions. At around 5 months of age these representations began to incorporate the dynamics of the represented motion. Strong learning effects were obtained over single trials, but there was no evidence of retention between trials. The individual differences were profound.

Electrooculography↗

Visual detection in monkeys with blindsight.

Monkeys with unilateral striate cortical removal show residual visual abilities in their affected hemifield. To learn whether the monkeys, like patients with blindsight, lose the phenomenal representation of the visual stimuli they nevertheless respond to, we first studied their ability to localize a briefly presented target in either hemifield. By varying the luminance of the stimuli we determined their visual sensitivity, which was reduced by 0.3-1.5 log units in the impaired hemifield; suprathreshold stimuli yielded almost perfect performance. We then presented two tests designed to show whether the monkeys categorized visual stimuli in the impaired field in the same manner as they categorize them in the normal field. In the first test, they had to respond differently according to whether one or two lights were presented, with the relative position of the two stimuli in a pair being varied. Whenever one of the paired stimuli lay in the impaired hemifield, two of the three monkeys consistently ignored it, and responded as if it had been a single stimulus in the good field. In the second test, trials consisting of a single stimulus light were interleaved with blank trials. The monkey touched the position of the light or made a different response, indicating that no light had appeared. All three monkeys responded to a light of supra-threshold luminance presented in the impaired field as if it were a blank trial. These results suggest that monkeys with striate cortical destruction, like neurological patients with similar lesions, have blindsight rather than phenomenal vision when they have to detect brief static visual targets.

Animals↗

eLoom and Flatland: specification, simulation and visualization engines for the study of arbitrary hierarchical neural architectures.

eLoom is an open source graph simulation software tool, developed at the University of New Mexico (UNM), that enables users to specify and simulate neural network models. Its specification language and libraries enables users to construct and simulate arbitrary, potentially hierarchical network structures on serial and parallel processing systems. In addition, eLoom is integrated with UNM's Flatland, an open source virtual environments development tool to provide real-time visualizations of the network structure and activity. Visualization is a useful method for understanding both learning and computation in artificial neural networks. Through 3D animated pictorially representations of the state and flow of information in the network, a better understanding of network functionality is achieved. ART-1, LAPART-II, MLP, and SOM neural networks are presented to illustrate eLoom and Flatland's capabilities.

Computer Simulation↗

The dissection of selection in person perception: inhibitory processes in social stereotyping.

Although people simultaneously belong to multiple social categories, any one of these competing representations can dominate the categorization process. It is surprising therefore to learn that only a few studies have considered the question of how people are categorized when multiple categorizations are available. In addition, relatively little is known about the cognitive mechanisms through which these categorization effects are realized. In the reported research, we attempted to extend recent ideas from work on selective attention to shed some light on these fundamental issues in social perception. Our basic contention was that following the initial identification of a person's applicable categories, the categorization process is driven by the interplay of both excitatory and inhibitory mechanisms. The results of 3 studies supported this contention. We discuss our findings in the wider context of contemporary issues in social stereotyping.

Adult↗

Global and local control of choice behavior by cyclically varying outcome probabilities.

Subjects asked to simulate decision makers in a competitive bidding situation chose repeatedly between two alternatives; reward probability varied according to a sine wave function of time for one alternative but was held constant over time for the other. Learning functions for choice probability exhibited the wavelike pattern predicted by a statistical learning model. However, on later transfer trials, when success probability was independent of subjects' choices, their choice behavior continued to follow a wavelike function rather than approaching the constant .5 level predicted by the learning model. A possible basis in memory for the transfer performance was revealed in subjects' sketches of the remembered pattern of variation in reward probabilities. It is concluded that choice performance is controlled by a mixture of local feedback, in a manner described by the learning model, and more global information encoded in an abstract memory representation, with the balance of the influence shifting toward the latter when local feedback becomes uninformative.

Adult↗

Haiku: language, communication, and hypnosis.

This article will illustrate how effective hypnotic communication closely resembles the Haiku form. Working with the Haiku form is an effective and dynamic approach that encourages the therapists to keep their awareness sharpened and observation astute. Haiku is not just a type of poetry; it is a way of looking at the world with a heightened level of attentiveness. Crafting effective and evocative hypnotic suggestions requires that the therapist become immersed in the world of passion, images, sounds, sights, opposites, humor, creativity, and perceptive consciousness. Enhancing our skills of observation is an important aspect of the continuing experience of the hypnotherapist. The Haiku method can help us enhance our observation and utilize what we observe in developing evocative hypnotic suggestions that help the client access their internal representational systems to stimulate their healing response. A systematized method for learning to write Haiku is presented.

Communication↗

Optimizing the use of vision in manual aiming: the role of practice.

The purpose of this study was to determine how subjects learn to adjust the characteristics of their manual aiming movements in order to make optimal use of the visual information and reduce movement error. Subjects practised aiming (120 trials) with visual information available for either 400 msec or 600 msec. Following acquisition, they were transferred to conditions in which visual information was available for either more or less time. Over acquisition, subjects appeared to reduce target-aiming error by moving to the target area more quickly in order to make greater use of vision when in the vicinity of the target. With practice, there was also a reduction in the number of modifications in the movement. After transfer, both performance and kinematic data indicated that the time for which visual information was available was a more important predictor of aiming error than the similarity between training and transfer conditions. These findings are not consistent with a strong "specificity of learning" position. They also suggest that, if some sort of general representation or motor programme develops with practice, that representation includes rules or procedures for the utilization of visual feedback to allow for the on-line adjustment of the goal-directed movement.

Adult↗

Preschool children's mapping of number words to nonsymbolic numerosities.

Five-year-old children categorized as skilled versus unskilled counters were given verbal estimation and number word comprehension tasks with numerosities 20-120. Skilled counters showed a linear relation between number words and nonsymbolic numerosities. Unskilled counters showed the same linear relation for smaller numbers to which they could count, but not for larger number words. Further tasks indicated that unskilled counters failed even to correctly order large number words differing by a 2 : 1 ratio, whereas they performed well on this task with smaller numbers, and performed well on a nonsymbolic ordering task with the same numerosities. These findings provide evidence that large, approximate numerosity representations become linked to number words around the time that children learn to count to those words reliably.

Adolescent↗

Hereditary cerebellar ataxia progressively impairs force adaptation during goal-directed arm movements.

We investigated how humans with hereditary cerebellar degeneration [spinocerebellar ataxia (SCA) type 6 and 8, n = 9] and age- and sex-matched healthy controls (n = 9) adapted goal-directed arm movements to an unknown external force field. We tested whether learning could be generalized to untrained regions in the workspace, an aspect central to the idea of an internal model, and if any learning could be retained. After removal of the force field, SCA patients showed little or no learning-related aftereffects indicating that repeated force-field exposure never led to successful force compensation. In contrast, healthy control subjects quickly adapted their movements to the new force field. The difference in force adaptation was significant for movements to targets that required both the shoulder and elbow joint (P < 0.001). Moreover, the generalization of learned movements to targets outside the learned workspace was prevented by the cerebellar degeneration (P < 0.01). Retention of force adaptation was significantly lower in SCA patients (P = 0.003). The severity of ataxia in SCA patients correlated negatively with the extent of learning (r = -0.84, P = 0.004). Our findings imply that progressive loss of cerebellar function gradually impairs force adaptation. The failure to generalize learning suggests that cerebellar degeneration prevents the formation of an internal representation of the limb dynamics.

Adaptation, Physiological↗

Blind source separation by sparse decomposition in a signal dictionary.

The blind source separation problem is to extract the underlying source signals from a set of linear mixtures, where the mixing matrix is unknown. This situation is common in acoustics, radio, medical signal and image processing, hyperspectral imaging, and other areas. We suggest a two-stage separation process: a priori selection of a possibly overcomplete signal dictionary (for instance, a wavelet frame or a learned dictionary) in which the sources are assumed to be sparsely representable, followed by unmixing the sources by exploiting the their sparse representability. We consider the general case of more sources than mixtures, but also derive a more efficient algorithm in the case of a nonovercomplete dictionary and an equal numbers of sources and mixtures. Experiments with artificial signals and musical sounds demonstrate significantly better separation than other known techniques.

Journal Article↗