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 685 records · Page 38Linked to original sources
PubMed1993

Answering the connectionist challenge: a symbolic model of learning the past tenses of English verbs.

Supporters of eliminative connectionism have argued for a pattern association-based explanation of language learning and language processing. They deny that explicit rules and symbolic representations play any role in language processing and cognition in general. Their argument is based to a large extent on two artificial neural network (ANN) models that are claimed to be able to learn the past tenses of English verbs (Rumelhart & McClelland, 1986, Parallel distributed processing, Vol. 2, Cambridge, MA: MIT Press; MacWhinney & Leinbach, 1991, Cognition, 40, 121-157). In this article we critically review Rumelhart and McClelland's as well as MacWhinney and Leinbach's ANN models and conclude that they do not succeed in the assigned task of learning the past tenses of English verbs. In order to answer their challenge to the symbolic processing approach, we present our symbolic pattern associator (SPA)-a general-purpose pattern associator that can learn to associate arbitrary discrete patterns. We carried out several experiments with the SPA using the same set of verbs that was used in MacWhinney and Leinbach's simulation with more realistic training and testing procedures. The SPA outperformed the connectionist models by a wide margin in the accuracy of learning, and successful inductive generalizations to unseen verbs. Our SPA has very natural and psychologically realistic explanations to many psychological effects such as U-shaped learning curve, and is much closer to human subjects in predicting past tense of the pseudo-verbs. In contrast to ANNs, whose internal representations are entirely opaque, the SPA can represent the acquired knowledge in the form of production rules that allow for further higher-level processing and integration, resulting in linguistically realistic associative templates for irregular verbs and production rules for regular verbs. In the light of these findings, we conclude that eliminative connectionists' vision of cognition as simple pattern association and pattern recognition without symbolic representation is inadequate. Pattern association as such does not imply rule-less or cue-based models of language acquisition or of human learning in general.

Cognition↗
PubMedDate not supplied

Place cell rigidity correlates with impaired spatial learning in aged rats.

In humans and in animals, some aged individuals are severely impaired in learning and memory capacity whereas others perform as well as young adults. In the present study, the spatial memory capacity of young and aged rats was characterized by the Morris water maze task, and then firing patterns of hippocampal "place cells" were assessed as the animals explored a familiar environment and a geometrically-altered version of the environment. Spatial representations of hippocampal cells in young and memory-intact aged rats changed upon exposure to the altered environment. In contrast, spatial representations of many cells in aged, memory-impaired rats were unaffected by the environmental alteration. Furthermore, combining all groups, the extent to which spatial representations distinguished the familiar and altered environments predicted learning capacity in the water maze. These findings suggest that a major component of memory impairment in aging may be the failure of the hippocampus to encode subtle differences in contextual information that differ across multiple experiences, such as the sequence of training trials in the water maze.

Aging↗
PubMed2006

Discrimination training alters object representations in human extrastriate cortex.

Visual object recognition relies critically on learning. However, little is known about the effect of object learning in human visual cortex, and in particular how the spatial distribution of training effects relates to the distribution of object and face selectivity across the cortex before training. We scanned human subjects with high-resolution functional magnetic resonance imaging (fMRI) while they viewed novel object classes, both before and after extensive training to discriminate between exemplars within one of these object classes. Training increased the strength of the response in visual cortex to trained objects compared with untrained objects. However, training did not simply induce a uniform increase in the response to trained objects: the magnitude of this training effect varied substantially across subregions of extrastriate cortex, with some showing a twofold increase in response to trained objects and others (including the right fusiform face area) showing no significant effect of training. Furthermore, the spatial distribution of training effects could not be predicted from the spatial distribution of either pretrained responses or face selectivity. Instead, training changed the spatial distribution of activity across the cortex. These findings support a dynamic view of the ventral visual pathway in which the cortical representation of an object category is continuously modulated by experience.

Discrimination, Psychological↗
PubMed2003

Motor adaptation to different dynamic environments is facilitated by indicative context stimuli.

When humans are exposed to external forces while performing arm movements, they adapt by compensating for these novel forces. The basis of this learning process is thought to be a neural representation that models the relation between all forces acting upon the system and the kinematic effects they produce, called inverse dynamic model (IDM). The present study investigated whether and how the predictability of a given external force affects the selection of an appropriate motor response to compensate for such force. Adult human subjects ( N=32) held a handle that could rotate around the elbow joint and learned to perform goal-directed forearm flexion movements, while an external velocity-dependent negative damping force was applied that assisted forearm movement. Subjects were randomly assigned to two groups. In the associative group, the applied damping force was always associated with a specific initial position. Thus, after initial learning, the force application became predictable. In the non-associative group, where the same movements were performed, the applied force was independent of the initial position, so that no association between force and location could be formed. We found that only the associative group significantly reduced target error when damping was present. That is, the location cue aided these subjects in generating dynamic responses in the appropriate limb. Our results indicate that motor adaptation to different dynamic environments can be facilitated by indicative stimuli.

Adult↗
PubMed2003

Multiple shifts in the representation of a motor sequence during the acquisition of skilled performance.

When do learning-related changes in performance occur? Here we show that the knowledge of a sequence of movements evolves through several distinctive phases that depend on two critical factors: the amount of practice as well as the passage of time. Our results show the following. (i) Within a given session, large performance gains constituted a signature for motor novelty. Such gains occurred only for newly introduced conditions irrespective of the absolute level of performance. (ii) A single training session resulted in both immediate but also time-dependent, latent learning hours after the termination of practice. Time in sleep determined the time of expression of these delayed gains. Moreover, the delayed gains were sequence-specific, indicating a qualitative change in the representation of the task within 24 h posttraining. (iii) Prolonged training resulted in additional between-session gains that, unlike the effects of a single training session, were confined to the trained hand. Thus, the effects of multisession training were qualitatively different than the immediate and time-dependent effects of a single session. Altogether, our results indicate multiple time-dependent shifts in the representation of motor experience during the acquisition of skilled performance.

Adolescent↗
PubMed1996

Associative and perceptual learning and the concept of memory systems.

An introductory review is followed by some new experimental data and a final discussion. The primate temporal lobe contains multiple qualitatively distinct memory systems. The functional properties of these memory systems can be explained by reference to the nature of the afferent information which they process, rather than by reference to any putative specialization in memory processing. In this way, the plasticity of 'memory systems' in associative memory is probably similar in principle to the plasticity of 'perceptual systems' in perceptual learning. Therefore, it is important to consider the relationship between perceptual and associative learning. Two experiments investigated perceptual learning in the Rhesus monkey (Macaca mulatta). Substantial perceptual learning was observed both with complex scenes and with simple colours. Two hypotheses as to the basis of perceptual learning are discussed. A physiological hypothesis is that training with a particular set of stimuli expands the cortical representation of those stimuli. This can explain the effects in both experiments. A psychological hypothesis is that perceptual learning is produced by learned associations among multiple features of complex stimuli. This can explain the effects in Expt. 1 but not in Expt. 2. The psychological associative hypothesis is therefore redundant. Furthermore, associative learning can itself be viewed as an expansion of the cortical representation of a complex event. Thus, the distinction between perceptual and memory systems will need to be abandoned as deeper understanding of cortical plasticity is achieved.

Animals↗
PubMed1997

RESEARCH: Theory in Practice: Applying Participatory Democracy Theory to Public Land Planning

/ Application of participatory democracy theory to public participation in public land planning, while widely advocated, has not been closely examined. A case study is used here to explicate the application of participatory democracy concepts to public participation in public land planning and decision making. In this case, a Bureau of Land Management resource area manager decided to make a significant shift from the traditional public involvement process to a more participatory method-coordinated resource management (CRM). This case was assessed using document analysis, direct observation of CRM meetings, questionnaires, and interviews of key participants. These sources were used to examine the CRM case using participatory democracy concepts of efficacy, access and representation, continuous participation throughout planning, information exchange and learning, and decision-making authority. The case study suggests that social deliberation in itself does not ensure successful collaboration and that establishing rules of operation and decision making within the group is critical. Furthermore, conflicts between the concept of shared decision-making authority and the public land management agencies' accountability to Congress, the President, and the courts need further consideration.KEY WORDS: Case study; Coordinated resource management; Public participation; Administrative discretion; Representation; Consensus; Collaboration

Journal Article↗
PubMed2000

Learning to find a shape.

We studied the transition of stimuli from novel to familiar in visual search and in the guidance of attention to a particular object. Ability to identify an object improved dramatically over several days of training. The learning was specific for the object's position in the visual field, orientation and configuration. Improvement was initially localized to one or two positions near the fixation spot and then expanded radially to include the full area of the stimulus array. Characteristics of this learning process may reflect a shift in the cortical representation of complex features toward earlier stages in the visual pathway.

Adult↗
PubMed2005

A docking analysis of the statistical physics of protein-protein recognition.

We describe protein-protein recognition within the frame of the random energy model of statistical physics. We simulate, by docking the component proteins, the process of association of two proteins that form a complex. We obtain the energy spectrum of a set of protein-protein complexes of known three-dimensional structure by performing docking in random orientations and scoring the models thus generated. We use a coarse protein representation where each amino acid residue is replaced by its Voronoï cell, and derive a scoring function by applying the evolutionary learning program ROGER to a set of parameters measured on that representation. Taking the scores of the docking models to be interaction energies, we obtain energy spectra for the complexes and fit them to a Gaussian distribution, from which we derive physical parameters such as a glass transition temperature and a specificity transition temperature.

Biophysics↗
PubMed2006

Essentializing differences between women and men.

People represent many social categories, including gender categories, in essentialist terms: They see category members as sharing deep, nonobvious properties that make them the kinds of things they are. The present research explored the consequences of this mode of representation for social inferences. In two sets of studies, participants learned (a) that they were similar to a member of the other gender on a novel attribute, (b) that they were different from a member of the other gender on a novel attribute, or (c) just their own standing on a novel attribute. Results showed that participants made stronger inductive inferences about the attribute in question when they learned that it distinguished them from a member of the other gender than in the other conditions. We consider the implications of these results for the representation of social categories and for everyday social inference processes.

Adult↗
PubMed1999

Distribution of tactile learning and its neural basis.

The brain's sensory processing systems are modified during perceptual learning. To learn more about the spatial organization of learning-related modifications, we trained rats to utilize the sensory signal from a single intact whisker to carry out a behavioral task. Once a rat had mastered the task, we clipped its "trained" whisker and attached a "prosthetic" one to a different whisker stub. We then tested the rat to determine how quickly it could relearn the task by using the new whisker. We observed that rats were immediately able to use the prosthetic whisker if it were attached to the stub of the trained whisker but not if it were attached to a different stub. Indeed, the greater the distance between the trained and prosthetic whisker, the more trials were needed to relearn the task. We hypothesized that this "transfer" of learning between whiskers might depend on how much the representations of individual whiskers overlap in primary somatosensory cortex. Testing this hypothesis by using 100-electrode cortical recordings, we found that the overlap between the cortical response patterns of two whiskers accounted well for the transfer of learning between them: The correlation between the electrophysiological and behavioral data was very high (r = 0.98). These findings suggest that a topographically distributed memory trace for sensory-perceptual learning may reside in primary sensory cortex.

Animals↗
PubMed1994

Comparison-based learning: effects of comparing instances during category learning.

When learning about a category, people often compare new instances with similar old instances and notice features common to the compared instances. Five experiments demonstrate that such comparisons cause features common to compared instances to be considered more important for the category than equally frequent features that are not common to compared instances. Experiment 1 shows that what is learned depends on which instances are compared. Experiment 2 investigates the conditions under which comparison-based learning occurs. The next experiments find that these comparisons affect subjective feature frequency (Experiment 3) and sensitivity to feature correlations (Experiment 4). Experiment 5 shows that comparisons during early learning affect what is learned from later instances. The discussion focuses on the implications for models of category representation.

Color Perception↗
PubMedDate not supplied

The parahippocampus subserves topographical learning in man.

The hippocampus has been proposed as the site of neural representation of large-scale environmental space, based upon the identification of place cells (neurons with receptive fields for current position in the environment) within the rat hippocampus and the demonstration that hippocampal lesions impair place learning in the rat. The inability to identify place cells within the monkey hippocampus and the observation that unilateral hippocampal lesions do not selectively impair topographic behavior in humans suggest that alternate regions may subserve this function in man. To examine the contribution of the hippocampus and adjacent medial-temporal lobe structures to topographic learning in the human, a 'virtual' maze was used as a task environment during functional magnetic resonance imaging studies. During the learning and recall of topographic information, medial-temporal activity was confined to the para- hippocampal gyri. This activity accords well with the lesion site known to produce topographical disorientation in humans. Activity was also observed in cortical areas known to project to the parahippocampus and previously proposed to contribute to a network subserving spatially guided behavior.

Adult↗
PubMed2005

Arthroscopic rotator cuff repair: the learning curve.

PURPOSE: The purpose of this study was to answer the question: How many cases are required for a surgeon to become proficient in performing arthroscopic rotator cuff repair? We hypothesize that as surgical experienced is gained, learning can be quantitatively shown by a significant decrease in operative time. TYPE OF STUDY: Prospective case series. METHODS: Rotator cuff repair time (RCRT) in minutes (as well as other time components comprising total surgical time) was recorded for 100 consecutive patients having arthroscopic rotator cuff repair performed by a single surgeon beginning with his first case in private practice. Mean RCRTs for consecutive blocks of 10 cases were compared. Learning is graphically represented by plotting the RCRT by case number and generating a logarithmic trend curve. A best-fit linear equation (y = mx + b) allows comparison of the initial 10 cases with the subsequent 90 cases, where m , the slope, represents the rate of decrease in RCRT (learning). RESULTS: Mean RCRT decreased significantly (P < .05) from the first block of 10 cases to the second block of 10 cases. There were no significant changes in mean RCRT when comparing other consecutive blocks of 10 cases. The slope of the line fitting the first block of 10 cases is -8.75; the slope (m) of the line fitting the subsequent 90 cases is -0.23. There is no significant difference in mean RCRT when cases are stratified by tear size. CONCLUSIONS: Graphic representation of RCRT by case number generates a learning curve whereby learning is quantitatively shown as a significant decrease in operative time as surgical experience is gained. CLINICAL RELEVANCE: Qualification of the learning curve for arthroscopic rotator cuff repair provides a guide for orthopaedic surgeons contemplating the expected time line for acquiring proficiency in this technique.

Arthroscopy↗
PubMed2003

The time course of spoken word learning and recognition: studies with artificial lexicons.

The time course of spoken word recognition depends largely on the frequencies of a word and its competitors, or neighbors (similar-sounding words). However, variability in natural lexicons makes systematic analysis of frequency and neighbor similarity difficult. Artificial lexicons were used to achieve precise control over word frequency and phonological similarity. Eye tracking provided time course measures of lexical activation and competition (during spoken instructions to perform visually guided tasks) both during and after word learning, as a function of word frequency, neighbor type, and neighbor frequency. Apparent shifts from holistic to incremental competitor effects were observed in adults and neural network simulations, suggesting such shifts reflect general properties of learning rather than changes in the nature of lexical representations.

Artificial Intelligence↗
PubMed2005

Neuroplasticity, learning and recovery after stroke: a critical evaluation of constraint-induced therapy.

Constraint-induced movement therapy (CIMT) has been hailed as a radical new approach to stroke rehabilitation. The guiding theory is that impairment of hand function is exacerbated by learned non-use and that this in turn leads to a loss of cortical representation of the upper limb. It is claimed that these processes can be reversed by two weeks of constraint of the unaffected limb combined with intensive practice in use of the paretic hand, and numerous small-scale studies have suggested that CIMT can lead to large improvements in function more than a year after stroke. However, the theory of learned non-use is open to question and there is uncertainty about the nature of the improvements induced by CIMT. The greatest effect seems to be increased spontaneous use of the hand, either through reduction of learned non-use or by overcoming the sense of effort during movement. There is also evidence of some improvement on dexterity tests but no studies have analysed in detail whether this reflects reduction of basic motor impairment or learning of compensatory movement strategies. The current weight of evidence is in favour of compensatory learning. Cortical changes detected by transcranial magnetic stimulation (TMS) or functional imaging may reflect this compensatory motor skill learning rather than restoration of representations lost due to the infarct or non-use of the limb. If future studies confirm this then the clinical implication is that direct teaching of unimanual or bimanual compensatory strategies might be a more productive approach than constraint.

Brain↗
PubMed2004

Associative learning shapes the neural code for stimulus magnitude in primary auditory cortex.

Since the dawn of experimental psychology, researchers have sought an understanding of the fundamental relationship between the amplitude of sensory stimuli and the magnitudes of their perceptual representations. Contemporary theories support the view that magnitude is encoded by a linear increase in firing rate established in the primary afferent pathways. In the present study, we have investigated sound intensity coding in the rat primary auditory cortex (AI) and describe its plasticity by following paired stimulus reinforcement and instrumental conditioning paradigms. In trained animals, population-response strengths in AI became more strongly nonlinear with increasing stimulus intensity. Individual AI responses became selective to more restricted ranges of sound intensities and, as a population, represented a broader range of preferred sound levels. These experiments demonstrate that the representation of stimulus magnitude can be powerfully reshaped by associative learning processes and suggest that the code for sound intensity within AI can be derived from intensity-tuned neurons that change, rather than simply increase, their firing rates in proportion to increases in sound intensity.

Acoustic Stimulation↗
PubMed2006

Neural representations during sleep: from sensory processing to memory traces.

In the course of a day, the brain undergoes large-scale changes in functional modes, from attentive wakefulness to the deepest stage of sleep. The present paper evaluates how these state changes affect the neural bases of sensory and cognitive representations. Are organized neural representations still maintained during sleep? In other words, despite the absence of conscious awareness, do neuronal signals emitted during sleep contain information and have a functional relevance? Through a critical evaluation of the animal and human literature, neural representations at different levels of integration (from the most elementary sensory level to the most cognitive one) are reviewed. Recordings of neuronal activity in animals at presentation of neutral or significant stimuli show that some analysis of the external word remains possible during sleep, allowing recognition of behaviorally relevant stimuli. Event-related brain potentials in humans confirm the preservation of some sensory integration and discriminative capacity. Behavioral and neuroimaging studies in humans substantiate the notion that memory representations are reactivated and are reorganized during post-learning sleep; these reorganisations may account for the beneficial effects of sleep on behavioral performance. Electrophysiological results showing replay of neuronal sequences in animals are presented, and their relevance as neuronal correlates of memory reactivation is discussed. The reviewed literature provides converging evidence that structured neural representations can be activated during sleep. Which reorganizations unique to sleep benefit memory representations, and to what extent the operations still efficient in processing environmental information during sleep are similar to those underlying the non-conscious, automatic processing continually at work in wakefulness, are challenging questions open to investigation.

Animals↗