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 1,387 records · Page 77Linked to original sources
PubMed1989

[The physician and the computer. 29. Learning systems and neuron networks].

The authors discuss basic properties of the process of learning as a form of artificial intelligence. They deal with the application of adaptive and learning systems in the area of identification, regulation, manipulation and recognition. A very important perspective trend is learning in a medium of neuron networks which are actually software or hardware simulation of the functional architecture of the CNS. The authors discuss peculiar features of representation of knowledge, structure and mode of function of this neuron network. Finally the authors give examples of the solution of optimalization and classification tasks within the framework of neurocomputers.

Artificial Intelligence↗
PubMed2001

Irrelevant response effects improve serial learning in serial reaction time tasks.

Tones were introduced into a serial reaction time (SRT) task to serve as redundant response effects. Experiment 1 showed that the tones improved serial learning with a 10-element stimulus sequence, but only if the tone effects were mapped onto the responses contingently. Experiment 2 demonstrated that switching to noncontingent response-effect mapping increased SRT only when participants had previously adapted to contingent response-effect mapping. In Experiment 3, the beneficial influence of contingent tone effects on serial learning occurred only when there was sufficient time between the response effects and the next imperative stimuli. The results are discussed in terms of the ideomotor principle. It is claimed that an internal representation of the to-be-produced tone effects develops and gains control over the execution of the response sequence.

Adult↗
PubMedDate not supplied

Category-use effects in children.

Three experiments examined the changes in category representation that take place when children use exemplars for tasks other than classification. In Experiments 1 and 2, 6- and 10-year-old children learned to classify exemplars of a novel category and then used the same exemplars in an inferential prediction task. In a subsequent classification task, features that were predictive for both classification and inference were classified more accurately than features that were predictive only of category membership. Experiment 3 showed that features with multiple uses were also more likely to be retrieved in feature listing. The findings show that children's category representations are affected by the way exemplars are used after they have been categorized.

Child↗
PubMed2006

Learning in trance: functional brain imaging studies and neuropsychology.

This study examined the fundamental question, whether verbal memory processing in hypnosis and in the waking state is mediated by a common neural system or by distinct cortical areas. Seven right-handed volunteers (25.4 years, sd 3.1) with high-hypnotic susceptibility scores were PET-scanned while encoding/retrieving word associations either in hypnosis or in the waking state. Word-pairs were visually presented and highly imaginable, but not semantically related (e.g. monkey-street). The presentation of pseudo-words served as a reference condition. An emission scan was recorded after each intravenous administration of O-15 water. Encoding under hypnosis was associated with more pronounced bilateral activations in the occipital cortex and the prefrontal areas as compared to learning in the waking state. During memory retrieval of word-pairs which had been previously learned under hypnosis, activations were found in the occipital lobe and the cerebellum. Under both experimental conditions precuneus and prefrontal cortex showed a consistent bilateral activation which was most distinct when the learning had taken place under hypnosis. In order to further analyze the effect of hypnosis on imagery-mediated learning, we administered sets of high-imagery word-pairs and sets of abstract words. In the first experimental condition word-pair associations were presented visually. In the second condition it was found that highly hypnotisable persons recalled significantly more high-imagery words under hypnosis as compared to low-hypnotisables both in the visual and auditory modality. Furthermore, high-imagery words were also better recalled by the highly hypnotisable subjects during the non-hypnotic condition. The memory effect was consistently present under both, immediate and delayed recall conditions. Taken together, the findings advance our understanding of the neural representation that underlies hypnosis and the neuropsychological correlates of hypnotic susceptibility.

Adult↗
PubMed1995

Configural association theory and the hippocampal formation: an appraisal and reconfiguration.

Sutherland and Rudy ([1989] Psychobiology 17:129-144) proposed that the hippocampal system is critical to normal learning and memory because of its function as the central part of a configural association system. This system constructs a unique representation of the joint occurrence of the independent elements of a compound. There is evidence consistent with the theory's predictions, however, there also are data that unambiguously demonstrate that, under some conditions, animals lacking an intact hippocampal system acquire configural associations. Thus, Sutherland and Rudy's fundamental assumption cannot be correct. To integrate the supporting and contradictory data, we propose two simple modifications of our position: 1) The critical neural system for configural associations is in cortical circuitry outside the hippocampus, and 2) the output from the hippocampal formation contributes to configural processing by selectively enhancing, thereby making more salient, cortical units representing stimulus conjunctions. This enhancement has two important effects: 1) It decreases the similarity between the configural units representing the co-occurrence of cues and the units representing the cues, and 2) It increases the rate at which the configural units can acquire associative strength. The modified theory explains why damage to the hippocampal formation only impairs learning on a subset of nonlinear discrimination problems. It also integrates recent data on the effects of hippocampal formation damage on conditioning involving context cues and makes novel predictions about performance on nonlinear discrimination problems and place learning.

Association Learning↗
PubMed2003

Skill learning in mirror reading: how repetition determines acquisition.

Practice makes perfect, but the role of repetitions in skill learning is not yet fully understood. For example, given a similar number of trials on a given task, it is debated whether repeating and non-repeating items are learned by the same neural process. When one is given training with both types of items--does one learn two separate skills, or only one? Here we show, using a mirror reading task, that practice trials with trial-unique words, and practice trials with repeated words, count towards learning to a different degree. There was no interaction between the time-course of learning repeated and unique words even within the same individuals given mixed training. While repeated words were learned faster than unique words, the repetitions-dependent gains diminished with training beyond a small number of repetitions. Moreover, the gains in performance could not be accounted for solely by the number of repetitions, as assumed by power-law models of learning; rather, the passage of time was a critical factor. Finally, our results suggest that although both repeated and new words were learned by both declarative and procedural memory mechanisms, even a single repetition of specific words could lead to the establishment of a selective differential representation in memory. The results are compatible with the notion of a repetition-sensitive process, triggered by specific repeating events. This 'repetition counter' may be a critical trigger for the effective formation of procedural as well as some type of declarative memory.

Adolescent↗
PubMed1996

Generalization of form in visual pattern classification.

Human observers were trained to criterion in classifying compound Gabor signals with symmetry relationships, and were then tested with each of 18 blob-only versions of the learning set. Generalization to dark-only and light-only blob versions of the learning signals, as well as to dark-and-light blob versions was found to be excellent, thus implying virtually perfect generalization of the ability to classify mirror-image signals. The hypothesis that the learning signals are internally represented in terms of a 'blob code' with explicit labelling of contrast polarities was tested by predicting observed generalization behaviour in terms of various types of signal representations (pixelwise, Laplacian pyramid, curvature pyramid, ON/OFF, local maxima of Laplacian and curvature operators) and a minimum-distance rule. Most representations could explain generalization for dark-only and light-only blob patterns but not for the high-thresholded versions thereof. This led to the proposal of a structure-oriented blob-code. Whether such a code could be used in conjunction with simple classifiers or should be transformed into a propositional scheme of representation operated upon by a rule-based classification process remains an open question.

Adult↗
PubMed2001

Language and number: a bilingual training study.

Three experiments investigated the role of a specific language in human representations of number. Russian-English bilingual college students were taught new numerical operations (Experiment 1), new arithmetic equations (Experiments 1 and 2), or new geographical or historical facts involving numerical or non-numerical information (Experiment 3). After learning a set of items in each of their two languages, subjects were tested for knowledge of those items, and new items, in both languages. In all the studies, subjects retrieved information about exact numbers more effectively in the language of training, and they solved trained problems more effectively than untrained problems. In contrast, subjects retrieved information about approximate numbers and non-numerical facts with equal efficiency in their two languages, and their training on approximate number facts generalized to new facts of the same type. These findings suggest that a specific, natural language contributes to the representation of large, exact numbers but not to the approximate number representations that humans share with other mammals. Language appears to play a role in learning about exact numbers in a variety of contexts, a finding with implications for practice in bilingual education. The findings prompt more general speculations about the role of language in the development of specifically human cognitive abilities.

Adult↗
PubMed2003

The cognitive neuroscience of category learning.

Recently, a multidisciplinary approach has provided new insights into the mechanisms of category learning. In this article, results from theoretical modeling, experimental psychology, clinical neuropsychology, functional neuroimaging, and single-cell studies are reviewed. Although the results are not conclusive, some general principles have emerged. Areas localized in the sensory neocortex are responsible for the perceptual representation of category exemplars, whereas lateral and anterior prefrontal structures are necessary for the encoding of category boundaries and abstract rules. The prefrontal cortex may influence categorical representation in the sensory neocortex via top-down control. The neostriatum is important in stimulus-response mapping, and the orbitofrontal cortex/ventral striatum are related to stimulus-reward associations accompanying category learning. Many category learning tasks can be performed implicitly. In conclusion, category learning paradigms provide a unique opportunity to investigate cognitive processes such as perception, memory, and attention in a systematic and interactive manner. Category learning tasks are suitable for mapping damaged brain systems in clinical populations.

Animals↗
PubMed2006

Learning your way around town: how virtual taxicab drivers learn to use both layout and landmark information.

By having subjects drive a virtual taxicab through a computer-rendered town, we examined how landmark and layout information interact during spatial navigation. Subject-drivers searched for passengers, and then attempted to take the most efficient route to the requested destinations (one of several target stores). Experiment 1 demonstrated that subjects rapidly learn to find direct paths from random pickup locations to target stores. Experiment 2 varied the degree to which landmark and layout cues were preserved across two successively learned towns. When spatial layout was preserved, transfer was low if only target stores were altered, and high if both target stores and surrounding buildings were altered, even though in the latter case all local views were changed. This suggests that subjects can rapidly acquire a survey representation based on the spatial layout of the town and independent of local views, but that subjects will rely on local views when present, and are harmed when associations between previously learned landmarks are disrupted. We propose that spatial navigation reflects a hierarchical system in which either layout or landmark information is sufficient for orienting and wayfinding; however, when these types of cues conflict, landmarks are preferentially used.

Adult↗
PubMedDate not supplied

Substructure mining using elaborate chemical representation.

Substructure mining algorithms are important drug discovery tools since they can find substructures that affect physicochemical and biological properties. Current methods, however, only consider a part of all chemical information that is present within a data set of compounds. Therefore, the overall aim of our study was to enable more exhaustive data mining by designing methods that detect all substructures of any size, shape, and level of chemical detail. A means of chemical representation was developed that uses atomic hierarchies, thus enabling substructure mining to consider general and/or highly specific features. As a proof-of-concept, the efficient, multipurpose graph mining system Gaston learned substructures of any size and shape from a mutagenicity data set that was represented in this manner. From these substructures, we extracted a set of only six nonredundant, discriminative substructures that represent relevant biochemical knowledge. Our results demonstrate the individual and synergistic importance of elaborate chemical representation and mining for nonlinear substructures. We conclude that the combination of elaborate chemical representation and Gaston provides an excellent method for 2D substructure mining as this recipe systematically explores all substructures in different levels of chemical detail.

Databases as Topic↗
PubMed2006

Similarity-based online feature selection in content-based image retrieval.

Content-based image retrieval (CBIR) has been more and more important in the last decade, and the gap between high-level semantic concepts and low-level visual features hinders further performance improvement. The problem of online feature selection is critical to really bridge this gap. In this paper, we investigate online feature selection in the relevance feedback learning process to improve the retrieval performance of the region-based image retrieval system. Our contributions are mainly in three areas. 1) A novel feature selection criterion is proposed, which is based on the psychological similarity between the positive and negative training sets. 2) An effective online feature selection algorithm is implemented in a boosting manner to select the most representative features for the current query concept and combine classifiers constructed over the selected features to retrieve images. 3) To apply the proposed feature selection method in region-based image retrieval systems, we propose a novel region-based representation to describe images in a uniform feature space with real-valued fuzzy features. Our system is suitable for online relevance feedback learning in CBIR by meeting the three requirements: learning with small size training set, the intrinsic asymmetry property of training samples, and the fast response requirement. Extensive experiments, including comparisons with many state-of-the-arts, show the effectiveness of our algorithm in improving the retrieval performance and saving the processing time.

Algorithms↗
PubMedDate not supplied

Nonlinear and higher-order approaches to the encoding of natural scenes.

Linear operations can only partially exploit the statistical redundancies of natural scenes, and nonlinear operations are ubiquitous in visual cortex. However, neither the detailed function of the nonlinearities nor the higher-order image statistics are yet fully understood. We suggest that these complicated issues can not be tackled by one single approach, but require a range of methods, and the understanding of the crosslinks between the results. We consider three basic approaches: (i) State space descriptions can theoretically provide complete information about statistical properties and nonlinear operations, but their practical usage is confined to very low-dimensional settings. We discuss the use of representation-related state-space coordinates (multivariate wavelet statistics) and of basic nonlinear coordinate transformations of the state space (e.g., a polar transform). (ii) Indirect methods, like unsupervised learning in multi-layer networks, provide complete optimization results, but no direct information on the statistical properties, and no simple model structures. (iii) Approximation by lower-order terms of power-series expansions is a classical strategy that has not yet received broad attention. On the statistical side, this approximation amounts to cumulant functions and higher-order spectra (polyspectra), on the processing side to Volterra Wiener systems. In this context we suggest that an important concept for the understanding of natural scene statistics, of nonlinear neurons, and of biological pattern recognition can be found in AND-like combinations of frequency components. We investigate how the different approaches can be related to each other, how they can contribute to the understanding of cortical nonlinearities such as complex cells, cortical gain control, end-stopping and other extraclassical receptive field properties, and how we can obtain a nonlinear perspective on overcomplete representations and invariant coding in visual cortex.

Animals↗
PubMed1985

Auditory representation of autogenous song in the song system of white-crowned sparrows.

The HVc (hyperstriatum ventrale, pars caudale) is a forebrain nucleus in the motor pathway for the control of song. Neurons in the HVc also exhibit auditory responses. A subset of these auditory neurons in the white-crowned sparrow (Zonotrichia leucophrys) have been shown to be highly selective for the individual bird's own (autogenous) song. By using multiunit recording techniques to sample from a large population, we demonstrate that the entire population of auditory neurons in the HVc is selective for autogenous song. The selectivity of these neurons must reflect the song-learning process, for the acoustic parameters of a sparrow's song are acquired by learning. By testing with laboratory-reared birds, we show that HVc auditory neurons prefer autogenous song over the tutor model to which the birds were exposed early in life. Thus, these neurons must be specified at or after the time song crystallizes. Since song is learned by reference to auditory feedback, HVc auditory neurons may guide the development of the motor program for song. The maintenance of a precise auditory representation of autogenous song into adulthood can contribute to the ability to distinguish the fine differences among conspecific songs.

Journal Article↗
PubMed2003

The interaction between discriminative stimuli and outcomes during instrumental learning.

Rats were trained on a biconditional discrimination in which the delivery of a food pellet stimulus signalled that pressing on one of two levers would be reinforced, whereas the delivery of a sucrose solution stimulus signalled that the reward was contingent on pressing the other lever. The outcome was the same food type as the discriminative stimulus in the congruent group but the other food type in the incongruent group. Both responses were rewarded with the same outcome in the same group. All the three groups learned the discrimination at statistically indistinguishable rates. Prefeeding one of the outcomes selectively reduced the associated response thereby demonstrating that responding was mediated by a representation of the outcome. Moreover, the outcome of one trial controlled responding on the next trial in accord with the stimulus function of the food type. These results are discussed in relation to the associative structures mediating the discriminative control of instrumental performance.

Animals↗
PubMed2005

Single neurons in the medial prefrontal cortex of the rat exhibit tonic and phasic coding during trace fear conditioning.

Trace fear conditioning is a learning task that requires the association of an auditory conditioned stimulus (CS) and a shock unconditioned stimulus (US) that are separated by a 20-s trace interval. Single neuron activity was recorded from the prelimbic and infralimbic areas of the medial prefrontal cortex in rats during trace fear conditioning or nonassociative unpaired training. Prelimbic neurons showed learning-related increases in activity to the CS and US, whereas infralimbic neurons showed learning-related decreases in activity to these stimuli. A subset of prelimbic neurons exhibited sustained increases in activity during the trace interval. These sustained prelimbic responses may provide a bridging code that allows for overlapping representations of CS and US information within the trace fear conditioning circuit.

Acoustic Stimulation↗
PubMed2004

Language and phonological skills in children at high risk of reading difficulties.

BACKGROUND: Dyslexia is now generally acknowledged to involve difficulties in phonological processing. However, the links between reading difficulties and speech difficulties remain unclear. METHOD: In the present study, 17 children with speech difficulties between the ages of four and six were compared to children with a family history of dyslexia and normally developing controls on phonological processing, phonological learning, phonological awareness and literacy tasks. RESULTS: The two groups of children at risk of reading difficulties showed very similar patterns of impairment, with average vocabulary but poor input and output speech processing, phonological learning, phonological awareness and reading development. CONCLUSIONS: It is concluded that the antecedents of reading difficulty are similar in these two groups of children, with both groups showing deficits in the development of phonological representations.

Articulation Disorders↗
PubMed2006

A disengagement deficit in representational space.

Some patients with hemispatial neglect show a deficit of "disengagement", i.e., their left-sided inattention is largely a consequence of an inability to move the spotlight of attention away from right-sided stimuli. We report a neglect patient with a failure of disengagement in imagined space, a feature not previously described. The patient was repeatedly moved along a hallway and had to memorize 20 objects placed alongside the walls (alternating starting points). Each learning run was followed by a recall run, in which objects had to be named in their correct sequence from one imaginary starting point. Initially, when performance was still poor, only right-sided items were named, a response pattern mimicking a neglect of representational space. However, as recall improved over successive runs, left-sided objects were as well memorized as right-sided, but the latter were named before the former. By contrast, if photographs of single objects were presented in the center of a screen for laterality decisions, neither accuracy nor latency of the patient's decisions differentiated between left-sided and right-sided items. We interpret the sticking to the right side during initial periods of free recall, in the absence of side-differences during cued recognition, as a failure to disengage from the right side of a mental image. In view of the extensive cortical and subcortical lesions in our patient the current debate about the functional neuroanatomy of this deficit cannot be resolved. However, the present report adds to our understanding of the heterogeneous nature of deficiencies in the representation of space.

Adult↗