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 811 records · Page 45Linked to original sources
PubMed2004

Conceptual precursors to language.

Because human languages vary in sound and meaning, children must learn which distinctions their language uses. For speech perception, this learning is selective: initially infants are sensitive to most acoustic distinctions used in any language, and this sensitivity reflects basic properties of the auditory system rather than mechanisms specific to language; however, infants' sensitivity to non-native sound distinctions declines over the course of the first year. Here we ask whether a similar process governs learning of word meanings. We investigated the sensitivity of 5-month-old infants in an English-speaking environment to a conceptual distinction that is marked in Korean but not English; that is, the distinction between 'tight' and 'loose' fit of one object to another. Like adult Korean speakers but unlike adult English speakers, these infants detected this distinction and divided a continuum of motion-into-contact actions into tight- and loose-fit categories. Infants' sensitivity to this distinction is linked to representations of object mechanics that are shared by non-human animals. Language learning therefore seems to develop by linking linguistic forms to universal, pre-existing representations of sound and meaning.

Adolescent↗
PubMed1987

Possible learned detection of exogenous brain frequency electromagnetic fields: a case study.

Zener card representations were presented visually for 15 sec. every 30 sec. for a total of 40 trials per session over 17 sessions to a single subject who displayed temporal lobe lability. During the 15 sec. before the presentation of one of the symbols (target) either a specific 4-Hz magnetic field pattern or a 1-Hz or 7-Hz magnetic field was presented bilaterally at the level of the temporal lobes. Field strengths were in the order of milligauss (about 1000X background values). The subject was required to anticipate the next symbol for each trial. Only when the 4-Hz field preceded the target symbol did the subject's accuracy of guessing remain systematically above (50%) chance (20%) over the sessions. These results suggest that exogenous magnetic fields may become discriminative stimuli through temporal association. Evidence of habituation was also obtained.

Brain↗
PubMed2005

Towards an explicit account of implicit learning.

PURPOSE OF REVIEW: The human brain supports acquisition mechanisms that can extract structural regularities implicitly from experience without the induction of an explicit model. Reber defined the process by which an individual comes to respond appropriately to the statistical structure of the input ensemble as implicit learning. He argued that the capacity to generalize to new input is based on the acquisition of abstract representations that reflect underlying structural regularities in the acquisition input. We focus this review of the implicit learning literature on studies published during 2004 and 2005. We will not review studies of repetition priming ('implicit memory'). Instead we focus on two commonly used experimental paradigms: the serial reaction time task and artificial grammar learning. Previous comprehensive reviews can be found in Seger's 1994 article and the Handbook of Implicit Learning. RECENT FINDINGS: Emerging themes include the interaction between implicit and explicit processes, the role of the medial temporal lobe, developmental aspects of implicit learning, age-dependence, the role of sleep and consolidation. SUMMARY: The attempts to characterize the interaction between implicit and explicit learning are promising although not well understood. The same can be said about the role of sleep and consolidation. Despite the fact that lesion studies have relatively consistently suggested that the medial temporal lobe memory system is not necessary for implicit learning, a number of functional magnetic resonance studies have reported medial temporal lobe activation in implicit learning. This issue merits further research. Finally, the clinical relevance of implicit learning remains to be determined.

Brain↗
PubMed2007

Prediction-based fingerprints of protein-protein interactions.

The recognition of protein interaction sites is an important intermediate step toward identification of functionally relevant residues and understanding protein function, facilitating experimental efforts in that regard. Toward that goal, the authors propose a novel representation for the recognition of protein-protein interaction sites that integrates enhanced relative solvent accessibility (RSA) predictions with high resolution structural data. An observation that RSA predictions are biased toward the level of surface exposure consistent with protein complexes led the authors to investigate the difference between the predicted and actual (i.e., observed in an unbound structure) RSA of an amino acid residue as a fingerprint of interaction sites. The authors demonstrate that RSA prediction-based fingerprints of protein interactions significantly improve the discrimination between interacting and noninteracting sites, compared with evolutionary conservation, physicochemical characteristics, structure-derived and other features considered before. On the basis of these observations, the authors developed a new method for the prediction of protein-protein interaction sites, using machine learning approaches to combine the most informative features into the final predictor. For training and validation, the authors used several large sets of protein complexes and derived from them nonredundant representative chains, with interaction sites mapped from multiple complexes. Alternative machine learning techniques are used, including Support Vector Machines and Neural Networks, so as to evaluate the relative effects of the choice of a representation and a specific learning algorithm. The effects of induced fit and uncertainty of the negative (noninteracting) class assignment are also evaluated. Several representative methods from the literature are reimplemented to enable direct comparison of the results. Using rigorous validation protocols, the authors estimated that the new method yields the overall classification accuracy of about 74% and Matthews correlation coefficients of 0.42, as opposed to up to 70% classification accuracy and up to 0.3 Matthews correlation coefficient for methods that do not utilize RSA prediction-based fingerprints. The new method is available at http://sppider.cchmc.org.

Artificial Intelligence↗
PubMed1998

Functional reorganization of the rat motor cortex following motor skill learning.

Functional reorganization of the rat motor cortex following motor skill learning. J. Neurophysiol. 80: 3321-3325, 1998. Adult rats were allocated to either a skilled or unskilled reaching condition (SRC and URC, respectively). SRC animals were trained for 10 days on a skilled reaching task while URC animals were trained on a simple bar pressing task. After training, microelectrode stimulation was used to derive high resolution maps of the forelimb and hindlimb representations within the motor cortex. In comparison with URC animals, SRC animals exhibited a significant increase in mean area of the wrist and digit representations but a decrease in elbow/shoulder representation within the caudal forelimb area. No between-group differences in areal representation were found in either the hindlimb or rostral forelimb areas. These results demonstrate that motor skill learning is associated with a reorganization of movement representations within the rodent motor cortex.

Animals↗
PubMed2001

Spatial pattern learning in rats: control by an iterative pattern.

Rats searched in a matrix of vertical poles for food hidden on top of the poles. The only information available about the location of the food was the consistent spatial pattern of the baited poles, which was a checkerboard. This spatial pattern of hidden-food locations came to control the choices of poles made by the rats. The experiments ruled out the possibility that this control can be explained by the acquisition of simple response tendencies to move from pole to pole. Instead, this behavioral control of choices was attributed to the development of a representation of the checkerboard pattern of baited locations. Spatial pattern learning may have mechanisms in common with other forms of pattern learning.

Animals↗
PubMed2003

Consolidation during sleep of perceptual learning of spoken language.

Memory consolidation resulting from sleep has been seen broadly: in verbal list learning, spatial learning, and skill acquisition in visual and motor tasks. These tasks do not generalize across spatial locations or motor sequences, or to different stimuli in the same location. Although episodic rote learning constitutes a large part of any organism's learning, generalization is a hallmark of adaptive behaviour. In speech, the same phoneme often has different acoustic patterns depending on context. Training on a small set of words improves performance on novel words using the same phonemes but with different acoustic patterns, demonstrating perceptual generalization. Here we show a role of sleep in the consolidation of a naturalistic spoken-language learning task that produces generalization of phonological categories across different acoustic patterns. Recognition performance immediately after training showed a significant improvement that subsequently degraded over the span of a day's retention interval, but completely recovered following sleep. Thus, sleep facilitates the recovery and subsequent retention of material learned opportunistically at any time throughout the day. Performance recovery indicates that representations and mappings associated with generalization are refined and stabilized during sleep.

Humans↗
PubMed2006

The primate amygdala represents the positive and negative value of visual stimuli during learning.

Visual stimuli can acquire positive or negative value through their association with rewards and punishments, a process called reinforcement learning. Although we now know a great deal about how the brain analyses visual information, we know little about how visual representations become linked with values. To study this process, we turned to the amygdala, a brain structure implicated in reinforcement learning. We recorded the activity of individual amygdala neurons in monkeys while abstract images acquired either positive or negative value through conditioning. After monkeys had learned the initial associations, we reversed image value assignments. We examined neural responses in relation to these reversals in order to estimate the relative contribution to neural activity of the sensory properties of images and their conditioned values. Here we show that changes in the values of images modulate neural activity, and that this modulation occurs rapidly enough to account for, and correlates with, monkeys' learning. Furthermore, distinct populations of neurons encode the positive and negative values of visual stimuli. Behavioural and physiological responses to visual stimuli may therefore be based in part on the plastic representation of value provided by the amygdala.

Amygdala↗
PubMed2003

Learning a single limb multijoint coordination pattern: the impact of a mechanical constraint on the coordination dynamics of learning and transfer.

The coordination dynamics of learning and transfer were studied in a single limb multijoint task requiring rhythmic elbow and wrist motions. Participants were required to learn a continuous 90 degrees relative phase pattern between the elbow and wrist such that an angle-angle plot of elbow and wrist motion produced a circle with a diameter of 80 degrees. Joint motion was restricted to elbow and wrist flexion-extension on the sagittal plane and the to-be-learned 90 degrees relative phase pattern was always practiced with the learning arm supine. Cycling frequency was controlled by a pacing metronome set at 0.75 Hz. Issues regarding effector-independent and effector-specific transfer were addressed with three transfer conditions: (1). learning arm prone (LP), (2). non-learning arm supine (NS), and (3). non-learning arm prone (NP). Four subjects learned the required relative phase (90 degrees ) and amplitude (80 degrees ) pattern with their dominant arm and four with their non-dominant arm. The experiment produced three main findings with regard to elbow-wrist control processes: First, seven of eight participants spontaneously produced a wrist-lagging coordination pattern (wrist motion lagged elbow motion) in learning to produce a continuous relative phase pattern of 90 degrees between the elbow and wrist. The wrist-lagging pattern may emerge as a result of the central nervous system exploiting the transfer of angular momentum from the elbow to the wrist as the elbow rotates up and down. The influence of interactive torque on elbow-wrist coordination represents an important mechanical constraint on the selection of intralimb coordination strategies during learning. The transfer conditions revealed that this mechanical constraint was effector-independent with regard to ipsilateral limb transfer (LP) and contralateral limb transfer (NS and NP). Second, consistent transfer of the learned relative phase pattern across ipsilateral and contralateral conditions demonstrates an effector-independent representation for this control variable. The effector-independent and effector-specific nature of joint amplitude transfer was dependent to some degree on learning arm, dominant or non-dominant, and the amount of practice, 1 day versus 5 days. Third, learning of the required 90 degrees relative phase pattern may be characterized as a phase transition leading to the formation of a stable attractor in the elbow-wrist coordination landscape. The above findings are discussed with respect to motor programming and coordination dynamic viewpoints on effector-independent and effector-specific aspects of motor equivalence.

Adult↗
PubMed2004

The neurobiology of category learning.

Many recent studies have examined the neural basis of category learning. Behavioral neuroscience results suggest that both the prefrontal cortex and the basal ganglia play important category-learning roles; neurons that develop category-specific firing properties are found in both regions, and lesions to both areas cause category-learning deficits. Similar studies indicate that the inferotemporal cortex does not mediate the learning of new categories. The cognitive neuroscience literature on category learning appears contradictory until the results are partitioned according to the type of category-learning task that was used. Three major tasks can be identified: rule based, information-integration, and prototype-distortion. Recent results are consistent with the hypotheses that (a) learning in rule-based tasks requires working memory and executive attention and is mediated by frontal-striatal circuits, (b) learning in information-integration tasks requires procedural memory and is mediated primarily within the basal ganglia, and (c) learning in prototype-distortion tasks depends on multiple memory systems, including the perceptual representation system.

Animals↗
PubMed2006

Experience-dependent representation of visual categories in parietal cortex.

Categorization is a process by which the brain assigns meaning to sensory stimuli. Through experience, we learn to group stimuli into categories, such as 'chair', 'table' and 'vehicle', which are critical for rapidly and appropriately selecting behavioural responses. Although much is known about the neural representation of simple visual stimulus features (for example, orientation, direction and colour), relatively little is known about how the brain learns and encodes the meaning of stimuli. We trained monkeys to classify 360 degrees of visual motion directions into two discrete categories, and compared neuronal activity in the lateral intraparietal (LIP) and middle temporal (MT) areas, two interconnected brain regions known to be involved in visual motion processing. Here we show that neurons in LIP--an area known to be centrally involved in visuo-spatial attention, motor planning and decision-making-robustly reflect the category of motion direction as a result of learning. The activity of LIP neurons encoded directions of motion according to their category membership, and that encoding shifted after the monkeys were retrained to group the same stimuli into two new categories. In contrast, neurons in area MT were strongly direction selective but carried little, if any, explicit category information. This indicates that LIP might be an important nexus for the transformation of visual direction selectivity to more abstract representations that encode the behavioural relevance, or meaning, of stimuli.

Animals↗
PubMed1998

Disconjugate memory-guided saccades to disparate targets: evidence for 3D sensitivity.

The saccadic system has been traditionally regarded as two-dimensional (horizontal, vertical) and basically conjugate in the two eyes. However, saccades to disparate targets (e.g., targets in real three-dimensional space that are located in different directions and at different distances) are naturally disconjugate. We report here that memory-guided saccades to a disparate target flashed 1 s earlier become disconjugate following repeated trials. After 15 min of repetition, the disconjugacy persists even when the target to be remembered is no longer disparate. This suggests fast memory-based learning. Learning, however, fails to occur if, during the repetition trials, the memory delay is 2 s. These findings suggest that the saccadic system has access to a 3D representation of targets and is gifted with 3D short-term memory and learning capacity.

Adaptation, Ocular↗
PubMed2006

Effects of associative learning on age differences in task-set switching.

Costs of switching between tasks may disappear when subjects are able to learn associations between tasks, stimuli, and responses (cf. Rogers, R. D., & Monsell, S. (1995). Costs of a predictable switch between simple cognitive tasks. Journal of Experimental Psychology: General, 124, 207-231). The first aim of this study was to examine this possibility by manipulating stimulus-set size. We expected that costs of switching between tasks would be strongly reduced under conditions of small stimulus-set sizes (n=4) as compared to large stimulus-set sizes (n=96) with increasing time on task. The second aim was to determine whether younger as well as older adults were able to create associations between task components. As age differences in task switching are often found to be larger when response mappings are incompatible we also investigated interactions with response compatibility. Results of our study indicated that practice effects on switch costs were much more pronounced for small than large stimulus-set sizes, consistent with the view that the strength of associations between task components facilitates task switching. Furthermore, we found that practice benefits on task switching for small stimulus-set sizes were sensitive to age and response compatibility. In contrast to younger adults, who showed a reduction of switch costs for both response mapping conditions, older adults showed a reduction of switch costs only when response mappings were compatible. That is, older adults showed less associative learning when the currently irrelevant task feature had to be suppressed, supporting the view that older adults have primarily problems in separating overlapping task-set representations.

Adult↗
PubMed1992

An introduction to model-based imaging.

The purpose of this paper is to clarify the distinction between the recognition of form, i.e. pattern recognition, and the interpretation of visual scenes, i.e. image understanding. Pattern recognition is part of image understanding, but the latter also includes cognitive tasks such as learning and inference. The key to developing image-understanding systems is to concentrate on the representation and use of models. This paper is a brief outline of the components of a model-based image-understanding system. First, the notions of iconic, categorical and symbolic knowledge are described. Although they appear to be disparate, the common notion is that the image understanding is based on recognizing concepts and not recognizing form. Next, the notion of a concept is defined, followed by representation techniques and control strategies for using concepts. Last, an example is given of an image-understanding system that learns to recognize concepts such as radiographic projections of teeth in panoramic radiographs.

Expert Systems↗
PubMed1999

Pretraining prevents spatial learning impairment after saturation of hippocampal long-term potentiation.

Spatial learning is impaired by NMDA receptor antagonists at doses that block hippocampal long-term potentiation (LTP). The deficit is not observed in animals that have received spatial or nonspatial pretraining in a different water maze. To determine whether this conditional impairment reflects debilitating sensorimotor effects of NMDA receptor antagonists in na¿ve animals, we compared spatial learning in na¿ve and pretrained animals in which induction of LTP was blocked by a saturation procedure with no obvious effects on sensorimotor functions. Rats with unilateral hippocampal lesions were implanted with multiple bipolar stimulation electrodes in the angular bundle and a recording electrode in the dentate gyrus of the intact hemisphere. Half of the rats were pretrained to find a hidden platform in a water maze. A week later, pretrained and na¿ve rats received either high-frequency (HF) or low-frequency (LF) stimulation at 2 hr intervals, until no further LTP could be induced. The stimulation did not interefere with performance on a balance task or a visual platform task. After stimulation, all rats were trained in a second water maze. Whereas na¿ve HF animals were impaired, pretrained HF animals acquired the new task rapidly and searched as extensively around the platform as LF control animals. These results suggest that pretraining prevents disruption of spatial learning after saturation of LTP in the absence of sensorimotor impairment, that hippocampal LTP might not be crucial for spatial representation per se, and that LTP may be involved only when spatial and contextual or procedural learning take place simultaneously.

Animals↗
PubMed2002

[Introduction to neuronal plasticity].

Neuronal plasticity is observed throughout animal kingdom. It underlies processes of functional restoration after brain injury or peripheral deafferentation and it also makes learning and memory possible. The paper reviews briefly the conditions which, according to Hebb's rule, are required for neuronal plasticity to occur. It stresses the importance of convergence of signals, coincidence and repetition. The roles of NMDA receptors and neuromodulators are described. The paper reviews current concepts and data concerning anatomical changes that accompany plasticity of cortical representations.

Animals↗
PubMed2025

CrossAttOmics: multiomics data integration with cross-attention.

MOTIVATION: Advances in high throughput technologies enabled large access to various types of omics. Each omics provides a partial view of the underlying biological process. Integrating multiple omics layers would help have a more accurate diagnosis. However, the complexity of omics data requires approaches that can capture complex relationships. One way to accomplish this is by exploiting the known regulatory links between the different omics, which could help in constructing a better multimodal representation. RESULTS: In this article, we propose CrossAttOmics, a new deep-learning architecture based on the cross-attention mechanism for multiomics integration. Each modality is projected in a lower dimensional space with its specific encoder. Interactions between modalities with known regulatory links are computed in the feature representation space with cross-attention. The results of different experiments carried out in this article show that our model can accurately predict the types of cancer by exploiting the interactions between multiple modalities. CrossAttOmics outperforms other methods when there are few paired training examples. Our approach can be combined with attribution methods like LRP to identify which interactions are the most important. AVAILABILITY AND IMPLEMENTATION: The code is available at https://github.com/Sanofi-Public/CrossAttOmics and https://doi.org/10.5281/zenodo.15065928. TCGA data can be downloaded from the Genomic Data Commons Data Portal. CCLE data can be downloaded from the depmap portal.

Humans↗
PubMed2003

Contiguity and contingency in action-effect learning.

According to the two-stage model of voluntary action, the ability to perform voluntary action is acquired in two sequential steps. Firstly, associations are acquired between representations of movements and of the effects that frequently follow them. Secondly, the anticipation or perception of an acquired action effect primes the movement that has been learnt to produce this effect; the acquired action-effect associations thus mediate the selection of actions that are most appropriate to achieve an intended action goal. If action-effect learning has an associative basis, it should be influenced by factors that are known to affect instrumental learning, such as the temporal contiguity and the probabilistic contingency of movement and effect. In two experiments, the contiguity or the contingency between key presses and subsequent tones was manipulated in various ways. As expected, both factors affected the acquisition of action-effect relations as assessed by the potency of action effects to prime the corresponding action in a later behavioral test. In particular, evidence of action-effect associations was obtained only if the effect of the action was delayed for no more than 1 s, if the effect appeared more often in the presence than in the absence of the action, or if action and effect were entirely uncorrelated but the effect appeared very often. These findings support the assumption that the control of voluntary actions is based on action-effect representations that are acquired by associative learning mechanisms.

Adult↗