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.

SourceResearch connectionFrom yearThrough year

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

At least 883 records · Page 49Linked to original sources

PubMed1996

Learning activation rules rather than connection weights.

In the construction of neural networks involving associative recall, information is sometimes best encoded with a local representation. Moreover, a priori knowledge can lead to a natural selection of connection weights for these networks. With predetermined and fixed weights, standard learning algorithms that work by altering connection strengths are unable to train such networks. To address this problem, this paper derives a supervised learning rule based on gradient descent, where connection weights are fixed and a network is trained by changing the activation rule. It incorporates both traditional and competitive activation mechanisms, the latter being an efficient method for instilling competition in a network. The learning rule has been implemented, and the results from several test networks demonstrate that it works effectively.

Algorithms↗

PubMed2001

An investigation of competitive learning for autonomous cluster identification in embedded systems.

Robust signal processing for embedded systems requires the effective identification and representation of features within raw sensory data. This task is inherently difficult due to unavoidable long-term changes in the sensory systems and/or the sensed environment. In this paper we explore four variations of competitive learning and examine their suitability as an unsupervised technique for the automated identification of data clusters within a given input space. The relative performance of the four techniques is evaluated through their ability to effectively represent the structure underlying artificial and real-world data distributions. As a result of this study it was found that frequency sensitive competitive learning provides both reliable and efficient solutions to complex data distributions. As well, frequency sensitive and soft competitive learning are shown to exhibit properties which may permit the evolution of an appropriate network structure through the use of growing or pruning procedures.

Cluster Analysis↗

PubMed1991

Implementations are not conceptualizations: revising the verb learning model.

In a recent issue of this journal, Pinker and Prince (1988) and Lachter and Bever (1988) presented detailed critiques of Rumelhart and McClelland's (1986) connectionist model of the child's learning of the phonological form of the English past tense. In order to address these criticisms, a new connectionist model was constructed using the back-propagation algorithm, a larger input corpus, a fuller paradigm, and a new phonological representation. This new implementation successfully addressed the criticisms of the phonological representation used by Rumelhart and McClelland. It did a much better job of learning the past tense using a fuller input set with realistic frequencies of occurrence. Ancillary simulations using the same network were able to deal with the homonymy problem and the generation of forms like "ated" from "ate". The one feature not provided by the new model was a way of modeling early correct production of irregular forms. The success of the new model can be used to help clarify the extent to which the published critiques apply to a particular connectionist implementation as opposed to fundamental principles underlying the broader connectionist conceptualization.

Concept Formation↗

PubMedDate not supplied

Language representation and working memory with bilinguals.

Working memory (WM) plays a crucial role in learning a second language (L2). The ability to repeat words in an unknown language has been observed to predict success in learning that language. Conversely, decreased digit span and inability to repeat pseudowords have been related with failure in L2 acquisition. Not only digit span, but also "word span" and "semantic span" should be considered in WM analysis. In addition to the phonological system, a semantic system is proposed in the WM model for language. In bilinguals, brain activation patterns during WM tasks have been observed to be more complex when using a L2. Processing information in L2 is more demanding, and WM may be less efficient. It can be conjectured that language understanding defects in L2 are at least partially due to this decreased efficiency of WM in its phonological as well as in its semantic subsystem. The reader will be introduced to the basic assumptions of WM. It will be emphasized that WM is significantly involved in the ability to learn a L2. Cross-linguistic differences in digit span will be analyzed. It will be concluded that despite digit span and word span are affected by different variables, "semantic span" may be similar across languages. Words in a L2 function as low frequency words, and hence, semantic search takes longer and WM is less efficient. It will be concluded that in addition to the "phonological system," the WM model should include a "semantic system," involving a "semantic store" and a "semantic search" process.

Humans↗

PubMedDate not supplied

Learning the topological properties of brain tumors.

This work presents a graph-based representation (a.k.a., cell-graph) of histopathological images for automated cancer diagnosis by probabilistically assigning a link between a pair of cells (or cell clusters). Since the node set of a cell-graph can include a cluster of cells as well as individual ones, it enables working with low-cost, low-magnification photomicrographs. The contributions of this work are twofold. First, it is shown that without establishing a pairwise spatial relation between the cells (i.e., the edges of a cell-graph), neither the spatial distribution of the cells nor the texture analysis of the images yields accurate results for tissue level diagnosis of brain cancer called malignant glioma. Second, this work defines a set of global metrics by processing the entire cell-graph to capture tissue level information coded into the histopathological images. In this work, the results are obtained on the photomicrographs of 646 archival brain biopsy samples of 60 different patients. It is shown that the global metrics of cell-graphs distinguish cancerous tissues from noncancerous ones with high accuracy (at least 99 percent accuracy for healthy tissues with lower cellular density level, and at least 92 percent accuracy for benign tissues with similar high cellular density level such as nonneoplastic reactive/inflammatory conditions).

Algorithms↗

PubMed2005

Learning bounds for kernel regression using effective data dimensionality.

Kernel methods can embed finite-dimensional data into infinite-dimensional feature spaces. In spite of the large underlying feature dimensionality, kernel methods can achieve good generalization ability. This observation is often wrongly interpreted, and it has been used to argue that kernel learning can magically avoid the "curse-of-dimensionality" phenomenon encountered in statistical estimation problems. This letter shows that although using kernel representation, one can embed data into an infinite-dimensional feature space; the effective dimensionality of this embedding, which determines the learning complexity of the underlying kernel machine, is usually small. In particular, we introduce an algebraic definition of a scale-sensitive effective dimension associated with a kernel representation. Based on this quantity, we derive upper bounds on the generalization performance of some kernel regression methods. Moreover, we show that the resulting convergent rates are optimal under various circumstances.

Artificial Intelligence↗

PubMed1989

The hippocampal formation is necessary for rats to learn and remember configural discriminations.

A negative patterning discrimination problem was arranged by reinforcing rats for bar pressing when either a light or tone was presented (L+/T+) but not reinforcing the response when the compound stimulus, light and tone, was presented (LT-). To solve this problem, the animal must be able to construct a unique configural representation of the compound that can be distinguished from the representations of the individual elements. We have proposed that the hippocampal formation is essential for the acquisition and retention of associations involving configural representations. Thus, our theory predicts that animals with hippocampal formation damage will not learn the negative patterning problem and that animals who learned this problem before receiving hippocampal formation damage will not retain the solution. These predictions were confirmed by the results of two experiments. The animals with hippocampal formation damage were unable to solve the negative patterning problem. These animals were able to solve a simple discrimination in which responding in the presence of a light was rewarded but responding in the presence of a tone was not rewarded. These results are discussed in relation to several other theories of hippocampal formation function.

Animals↗

PubMed2004

The role of associative history in models of associative learning: a selective review and a hybrid model.

Associative learning theories strive to capture the processes underlying and driving the change in strength of the associations between representations of stimuli that develop as a result of experience of the predictive relationships between those stimuli. Historically, formal models of associative learning have focused on two potential factors underlying associative change, namely processing of the conditioned stimulus (in terms of changes in associability) and processing of the unconditioned stimulus (in terms of changes in error). This review constitutes an analysis of the proper role of these two factors, specifically with regard to the way in which they are influenced by associative history (the prior training undergone by cues). A novel "hybrid" model of associative learning is proposed and is shown to provide a more satisfactory account of the effects of associative history on subsequent learning than any previous single-process theory.

Association Learning↗

PubMed2001

Reorganization in awake rat auditory cortex by local microstimulation and its effect on frequency-discrimination behavior.

In common with other sensory cortices, the mammalian primary auditory cortex (AI) demonstrates the capacity for large-scale reorganization following many experimental situations. For example, training animals in frequency-discrimination tasks has been shown to result in an increase in cortical frequency representation. Such central changes-most commonly, an increase in central representation of specific stimulus parameters-have been hypothesized to underlie the improvements in perceptual acuity (perceptual learning) seen in many learning situations. The actual behavioral relevance of central reorganizations, however, remains speculative. Here, we directly examine this issue. We first show that stimulating the AI cortex of the awake rat with a weak electric current (intracortical microstimulation or ICMS) has the effect of inducing central reorganizations similar to those accompanying the traditional plasticity experiments (a result previously noted only in anesthetized preparations). Depending on the site of AI stimulation, ICMS enlarged the cortical representation of certain frequencies. Next we examined the direct perceptual consequences of ICMS-induced AI reorganization for the rat's ability to discriminate frequencies. Over the course of the experiment, we also detailed, and made comparisons between, the frequency-response characteristics of rat AI cortex in the awake and ketamine-anesthetized animal. AI cells that responded to pure tones were divided into two categories--strongly and weakly responsive--based on the strength of their evoked discharge. Individual cells maintained their respective response strengths in both awake and anesthetized conditions. Strongly responsive cells showed at least four different temporal responses and tended to be narrowly tuned. Their responses were stable over the long term. In general frequency-response characteristics were qualitatively similar in the anesthetized and awake animal; bandwidths tended to be broader in awake animals. Although both strong and weak cell populations respond to tones, only the strongly responsive cells fit into a tonotopically organized scheme. By contrast, weakly responsive cells did not exhibit a frequency mapping and may represent a more diffuse input to AI than that underlying strongly responsive cells. In general, the overall frequency organization of AI was found to be equally well expressed in both the awake and anesthetized rat. ICMS reorganization of AI did not alter frequency-discrimination behavior in the rat--either signal detectability or response bias--suggesting that an increase in central representation, by itself, is insufficient to account for perceptual learning. It is likely that cortical reorganizations that accompany perceptual learning are strongly keyed to specific behavioral contexts.

Acoustic Stimulation↗

PubMed1996

Use-dependent alterations of movement representations in primary motor cortex of adult squirrel monkeys.

This study was undertaken to document plastic changes in the functional topography of primary motor cortex (M1) that are generated in motor skill learning in the normal, intact primate. Intracortical microstimulation mapping techniques were used to derive detailed maps of the representation of movements in the distal forelimb zone of M1 of squirrel monkeys, before and after behavioral training on two different tasks that differentially encouraged specific sets of forelimb movements. After training on a small-object retrieval task, which required skilled use of the digits, their evoked-movement digit representations expanded, whereas their evoked-movement wrist/forearm representational zones contracted. These changes were progressive and reversible. In a second motor skill exercise, a monkey pronated and supinated the forearm in a key (eyebolt)-turning task. In this case, the representation of the forearm expanded, whereas the digit representational zones contracted. These results show that M1 is alterable by use throughout the life of an animal. These studies also revealed that after digit training there was an areal expansion of dual-response representations, that is, cortical sectors over which stimulation produced movements about two or more joints. Movement combinations that were used more frequently after training were selectively magnified in their cortical representations. This close correspondence between changes in behavioral performance and electrophysiologically defined motor representations indicates that a neurophysiological correlate of a motor skill resides in M1 for at least several days after acquisition. The finding that cocontracting muscles in the behavior come to be represented together in the cortex argues that, as in sensory cortices, temporal correlations drive emergent changes in distributed motor cortex representations.

Animals↗

PubMed1994

On the cognitive processes underlying contextual interference and observational learning.

The main goal of the present study was to determine whether observation of an unskilled model learning a timing task enables the observer to develop a cognitive representation of the task similar to the one acquired through physical practice (Adams, 1986; Bandura, 1977; Lee & White, 1990). To reach that goal, we tested whether a contextual interference effect would be obtained in a retention test of subjects who had observed an individual practicing three variations of a timing task under a random or a blocked schedule of practice. Similar patterns of results in an immediate retention test were found following observation and physical practice. This suggests that observation indeed engaged the observers in the same type of cognitive activities as did physical practice. Moreover, a schedule of practice made up of 100% physical practice led to improved learning compared with a schedule of practice made up of 50% observation followed by 50% physical practice. This suggests that learning is enhanced more by numerous implementations of a motor program than by its mere construction or retrieval.

Journal Article↗

PubMedDate not supplied

An audit of the processes involved in identifying and assessing bilingual learners suspected of being dyslexic: a Scottish study.

The Commission for Racial Equality (Special Educational Needs Assessment in Strathclyde: Report of a Formal Investigation, CRE, London, 1996) highlighted the significant under-representation of bilingual children among pupils assessed as having specific learning difficulties/dyslexia. In this present study an audit was undertaken in order to explore issues arising from the Commission's report, initially using 53 schools from one education authority. This revealed an extremely low incidence of suspected dyslexia among bilingual pupils. A second study was carried out in a further nine education authorities, surveying 91 schools with bilingual pupils. The incidence of suspected dyslexia in bilingual pupils was found to be extremely low. Twenty-seven cases were examined. Most cases concerned pupils aged 7:0-9:0. Difficulties associated with conventional indicators of dyslexia are discussed. A wide variety of assessment approaches were reported and the use of first language (L1) assessment varied. The process of assessment tended to be lengthy and inconclusive. However, this report suggests that caution is necessary when considering dyslexia in the early stages of second language (L2) development.

Adolescent↗

PubMed1979

Visual cortex ablation and thresholds for successively presented stimuli in rhesus monkeys: II. Hue.

Rhesus monkeys were trained to discriminate successively presented hues. The smallest difference they could reliably detect was determined before and after either inferotemporal ablation, or a lesion intended to remove as much as possible of prestriate area V4 (Zeki, 1973). As a group, the animals with lesions of V4 showed good but not perfect retention of their preoperative performance, and their thresholds were unaltered. The inferotemporal group showed no retention of the simplest successive task, red versus green, but after relearning their thresholds too were unaltered. It appears that animals without inferotemporal cortex can form precise internal representations of hues, and that the basis of the inferotemporal learning impairment may depend upon the nature of the stimuli to be discriminated.

Animals↗

PubMed2026

DNA methylation biomarkers for early detection of ovarian cancer.

Ovarian cancer (OC) remains difficult to detect at an early stage, and current screening approaches using CA125 and transvaginal ultrasonography have not demonstrated sufficient benefit for population screening. DNA methylation is a promising biomarker class because epigenetic alterations may arise early in tumourigenesis, can be detected in circulating cell-free DNA (cfDNA), and may provide tissue-of-origin information. This review critically evaluates recent evidence on DNA methylation biomarkers for early OC detection. PubMed/MEDLINE, Web of Science, and Scopus were searched for studies published between January 2020 and September 2025, supplemented by selected earlier studies of biological or methodological relevance. Evidence was synthesised across single-gene biomarkers, multi-locus panels, genome-wide signatures, assay platforms, and machine-learning classifiers, with emphasis on early-stage performance, histological representation, comparator populations, analytical methodology, and validation design. Single-gene markers such as BRCA1, RASSF1A, OPCML, HOXA9, and HIC1 show variable performance, while multi-gene and classifier-based approaches generally provide stronger discrimination. However, many studies remain limited by retrospective case-control designs, small FIGO stage I-II subsets, predominance of serous disease, and insufficient prospective validation. Integration with CA125 may improve sensitivity but can reduce specificity, which is critical in low-prevalence screening. Clinical translation will therefore require minimal and reproducible methylation signatures, standardised low-input cfDNA workflows, rigorous external validation, and prospective longitudinal evaluation in intended-use populations.

Humans↗

PubMed2006

Category-specificity can emerge from bottom-up visual characteristics: evidence from a modular neural network.

The role of bottom-up visual processes in category-specific object recognition has been largely unexplored. We examined the role of low-level visual characteristics in category specific recognition using a modular neural network comprising both unsupervised and supervised components. One hundred standardised pictures from ten different categories (five living and five nonliving, including body parts and musical instruments) were presented to a Kohonen self-organising map (SOM) which re-represents the visual stimuli by clustering them within a smaller number of dimensions. The SOM representations were then used to train an attractor network to learn the superordinate category of each item. The ease with which the model acquired the category mappings was investigated with respect to emerging category effects. We found that the superordinates could be separated by very low-level visual factors (as extracted by the SOM). The model also accounted for the well documented atypicality of body parts and musical instrument superordinates. The model has clear relevance to human object recognition since the model was quicker to learn more typical category exemplars and finally the model also accounted for more than 20% of the naming variance in a sample of 57 brain injured subjects. We conclude that purely bottom-up visual characteristics can explain some important features of category-specific phenomena.

Female↗

PubMed2007

Tuning the engine of cognition: a focus on NMDA/D1 receptor interactions in prefrontal cortex.

The prefrontal cortex of the primate frontal lobes provides the capacity for judgment which can constantly adapt behavior in order to optimize its outcome. Adjudicating between long-term memory programs and prepotent responses, this capacity reviews all incoming information and provides an interpretation dependent on the events that have just occurred, the events that are predicted to happen, and the alternative response strategies that are available in the given situation. It has been theorized that this function requires two essential integrated components, a central executive which guides selective attention based on mechanisms of associative memory, as well as the second component, working memory buffers, in which information is held online, abstracted, and translated on a mental sketchpad of work in progress. In this review, we critically outline the evidence that the integration of these processes and, in particular, the induction and maintenance of persistent activity in prefrontal cortex and related networks, is dependent upon the interaction of dopamine D1 and glutamate NMDA receptor signaling at critical nodes within local circuits and distributed networks. We argue that this interaction is not only essential for representational memory, but also core to mechanisms of neuroadaptation and learning. Understanding its functional significance promises to reveal major new insights into prefrontal dysfunction in schizophrenia and, hence, to target a new generation of drugs designed to ameliorate the debilitating working memory deficits in this disorder.

Animals↗

PubMed2005

Relational language and the development of relational mapping.

We test the claim that learning and using language for spatial relations can influence spatial representation and reasoning. Preschool children were given a mapping task in which they were asked to find a "winner" placed in a three-tiered box after seeing one placed in a virtually identical box. The correct choice was determined by finding the corresponding relative location in the test box, making it a difficult task for preschool children. We found that hearing language for spatial relations facilitated children's mapping performance. We found effects at younger ages on easier tasks (Experiments 1 and 2) and at older ages on harder tasks (Experiment 3). The effects of spatial relational language differed predictably according to the semantics of the terms children heard (Experiment 4). Finally, the effects of spatial language were maintained over time (Experiment 5): children given one initial exposure to the spatial terms maintained their advantage over baseline children when they again carried out the mapping task 2 days later, with no further exposure to the spatial terms. The evidence is consistent with the explanation that language bolsters children's spatial encodings, which in turn supports their mapping performance.

Age Factors↗

PubMedDate not supplied

Self-organization of distributedly represented multiple behavior schemata in a mirror system: reviews of robot experiments using RNNPB.

The current paper reviews a connectionist model, the recurrent neural network with parametric biases (RNNPB), in which multiple behavior schemata can be learned by the network in a distributed manner. The parametric biases in the network play an essential role in both generating and recognizing behavior patterns. They act as a mirror system by means of self-organizing adequate memory structures. Three different robot experiments are reviewed: robot and user interactions; learning and generating different types of dynamic patterns; and linguistic-behavior binding. The hallmark of this study is explaining how self-organizing internal structures can contribute to generalization in learning, and diversity in behavior generation, in the proposed distributed representation scheme.

Artificial Intelligence↗