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 469 records · Page 26Linked to original sources
PubMed2001

Kaliotoxin, a Kv1.1 and Kv1.3 channel blocker, improves associative learning in rats.

Olfactory associative learning was used to investigate the involvement of Kv channels containing Kv1.1 and Kv1.3 alpha-subunits in learning and memory. Kaliotoxin (KTX), a specific inhibitor of these Kv channels, was injected intracerebroventricularly in the rat brain, at a dose of 10 ng that did not disturb the rats' locomotor activity or drinking behaviour. In the first paradigm (odour-reward training), KTX improved learning but not information consolidation. Moreover, KTX increased the long-term retrieval of an odour-reward association tested by a reversal test 1 month after the odour-reward training. The second paradigm (successive odour-pair training) tested reference memory. The first session was an acquisition session where the rats learned a new odour-discrimination problem with the same procedure. The second was a retention session held 24 h later to test retrieval of the learned information. KTX injected before the acquisition or retention session improved performance, but no effect was found when KTX was injected immediately after acquisition. We showed that these effects were not due to the action of KTX on attention processes. Thus, these results suggest that the blockage of Kv1.1 or Kv1.3 channels by KTX facilitates cognitive processes as learning, in particular in a reference representation.

Analysis of Variance↗
PubMedDate not supplied

Degree of adaptability of the somatosensory cortex to change: prospects for integration of bone-mounted dental prostheses.

1. The topographic representation of the body surface in the somatosensory cortex provides an important model system for the in vivo study of neuronal plasticity, induced changes in somatotopy providing a direct measure of plasticity not available in most parts of the central nervous system. 2. Over the past two decades, animal experimentation in a number of laboratories has shown a remarkable degree of adaptability of the cortical representation following peripheral lesions and has had a widespread influence by challenging the once-accepted dogma that the brain is a structurally fixed organ. 3. Although some aspects of original stimulation will be missing, it is likely that receptors stimulated through bone conduction and compression by bone-mounted dental prostheses preserve some of the geometric and temporal relationships of original stimulation. By analogy with data obtained from the forearm representations, it would be expected that many features of the original cortical representations will be recreated. 4. There are also examples in the literature of perceptual learning without gross changes to the cortical representation (some being within a class of adaptability known as gain control) and it is likely that perceptual integration of many dental prostheses occurs within the limits of these neural adaptation mechanisms.

Adaptation, Physiological↗
PubMed1991

Independence of question-answering strategy and searched representation.

Because many studies of question answering, recall, and recognition have examined script-based messages, a spurious link between answering strategy and searched representation has emerged: Direct retrieval has been associated with the episodic representation of a discourse or a learned list, whereas plausibility judgment has been associated with general knowledge structures. Evidence of the independence of answering strategy and searched representation is presented. First, previous research and intuition both provide examples of the application of either strategy to either type of representation. Second, there is empirical evidence of the application of either strategy to both message text bases and general knowledge. Third, the effects of factors influencing strategy selection are independent of the type of searched representation. It is concluded that answering strategy is independent of the representation to which it is applied.

Cognition↗
PubMed2002

Spatial generalization of learning in smooth pursuit eye movements: implications for the coordinate frame and sites of learning.

We have examined the underlying coordinate frame for pursuit learning by testing how broadly learning generalizes to different retinal loci and directions of target motion. Learned changes in pursuit were induced using double steps of target speed. Monkeys tracked a target that stepped obliquely away from the point of fixation, then moved smoothly either leftward or rightward. In each experimental session, we adapted the response to targets moving in one direction across one locus of the visual field by changing target speed during the initial catch-up saccade. Learning occurred in both presaccadic and postsaccadic eye velocity. The changes were specific to the adapted direction and did not generalize to the opposite direction of pursuit. To test the spatial scale of learning, we examined the responses to targets that moved across different parts of the visual field at the same velocity as the learning targets. Learning generalized partially to motion presented at untrained locations in the visual field, even those across the vertical meridian. Experiments with two sets of learning trials showed interference between learning at different sites in the visual field, suggesting that pursuit learning is not capable of spatial specificity. Our findings are consistent with the previous suggestions that pursuit learning is encoded in an intermediate representation that is neither strictly sensory nor strictly motor. Our data add the constraint that the site or sites of pursuit learning must process visual information on a fairly large spatial scale that extends across the horizontal and vertical meridians.

Animals↗
PubMed1996

A syntactic complexity effect with visual patterns: evidence for the syntactic nature of the memory representation.

In a series of 3 experiments, participants learned visual patterns that contained the same number of visual features but varied in the complexity of the interrelations among the features. The results indicate a large and orderly effect of the pattern's syntactic complexity on recognition speed. Evidence is provided that this effect was not due to physical characteristics, target-foil similarity, speed-accuracy trade-off, or level of pattern learning. A multiple-encoding explanation of the effect is described. According to this framework, there is an initial, automatically generated encoding of the pattern as a short-term pictorial representation that becomes the basis for the construction of a second syntactic-propositional encoding. In this model, the participant's "sense of familiarity" for a particular stimulus is associated only with the syntactic-propositional encoding.

Adult↗
PubMed2001

Learning new words: phonotactic probability in language development.

Though the influences of syntactic and semantic regularity on novel word learning are well documented, considerably less is known about the influence of phonological regularities on lexical acquisition. The influence of phonotactic probability, a measure of the likelihood of occurrence of a sound sequence, on novel word learning is investigated in this study. Thirty-four typically developing children (from ages 3 years 2 months to 6 years 3 months) participated in a multitrial word-learning task involving nonwords of varying phonotactic probability (common vs. rare) paired with unfamiliar object referents. Form and referent learning were tested following increasing numbers of exposures (1 vs. 4 vs. 7) and following a 1-week delay. Correct responses were analyzed to determine whether phonotactic probability affected rate of word learning, and incorrect responses were analyzed to examine whether phonotactic probability affected the formation of semantic representations, lexical representations, or the association between semantic and lexical representations. Results indicated that common sound sequences were learned more rapidly than rare sound sequences across form and referent learning. In addition, phonotactic probability appeared to influence the formation of semantic representations and the association between semantic and lexical representations. These results are integrated with previous findings and theoretical models of language acquisition.

Child↗
PubMed2003

Implicit learning of complex information in amnesia.

Implicit learning abilities of nine amnesic patients were explored by using an artificial grammar learning task in which the test strings were constructed in such a way that grammaticality judgments could not be based on a simple knowledge of bigrams and trigrams (chunks). Results show that amnesic patients and controls performed at the same level during the classification task, whereas amnesic patients performed worse than controls in an explicit generation task. Moreover, there was no correlation between the implicit and explicit measures. These results are compatible with the existence of two kinds of representation intervening in artificial grammar learning. The first one based on processes leading to fragment-specific knowledge (the chunks, which can be accessed explicitly), and the second based on the learning of simple associations and more complex conditional relations between elements.

Adult↗
PubMed2001

The 28th Bartlett Memorial Lecture. Causal learning: an associative analysis.

The concordance between performance and judgements of the causal effectiveness of an instrumental action suggests that such actions are mediated by causal knowledge. Although causal learning exhibits many associative phenomena--blocking, inhibitory or preventative learning, and super-learning--judgements of the causal status of a cue can be changed retrospectively as a result of learning episodes that do not directly involve the cue. In order to explain retrospective revaluation, a modified associative theory is described in which the learning processes for retrieved cue representations are the opposite to those for presented cues, and this theory is evaluated by studies of the role of within-compound associations in retrospective revaluation and blocking. However, this modified theory only applies when the within-compound association represents a contiguous rather than a causal cue relationship.

Animals↗
PubMed1997

Non-linear Feature Extraction by Redundancy Reduction in an Unsupervised Stochastic Neural Network.

Unsupervised feature extraction by a stochastic neural network can be defined as a minimization of the redundancy between the elements of the output layer, given complete information transfer from input to output. Redundancy minimization can be achieved by minimization of the mutual information between the units of the output layer. Complete information transfer is enforced by maximizing the mutual information of the input and output. With these two conditions we define a novel learning algorithm for stochastic recurrent networks. The minimum of redundancy corresponds to the extraction of statistically independent features, leading to a factorial representation of the environment. The resulting learning rule includes Hebbian and anti-Hebbian learning terms. These two terms are weighted by the amount of information transmitted in the learning synapse minus the grade of redundant information in the corresponding output neuron, giving thus, an information-theoretic interpretation of the proportionality constant of Hebb's biological rule. Simulations demonstrate the performance of this method. When a retina is simulated, the learning algorithm forms decorrelated receptive fields. This represents the first experiment that extends the results of the linear principle component analysis to the nonlinear case by a direct implementation of Barlow's principle of redundancy reduction for unsupervised features extraction by receptive fields formation in a retina model. Copyright 1997 Elsevier Science Ltd.

Journal Article↗
PubMed2005

Different neural correlates of reward expectation and reward expectation error in the putamen and caudate nucleus during stimulus-action-reward association learning.

To select appropriate behaviors leading to rewards, the brain needs to learn associations among sensory stimuli, selected behaviors, and rewards. Recent imaging and neural-recording studies have revealed that the dorsal striatum plays an important role in learning such stimulus-action-reward associations. However, the putamen and caudate nucleus are embedded in distinct cortico-striatal loop circuits, predominantly connected to motor-related cerebral cortical areas and frontal association areas, respectively. This difference in their cortical connections suggests that the putamen and caudate nucleus are engaged in different functional aspects of stimulus-action-reward association learning. To determine whether this is the case, we conducted an event-related and computational model-based functional MRI (fMRI) study with a stochastic decision-making task in which a stimulus-action-reward association must be learned. A simple reinforcement learning model not only reproduced the subject's action selections reasonably well but also allowed us to quantitatively estimate each subject's temporal profiles of stimulus-action-reward association and reward-prediction error during learning trials. These two internal representations were used in the fMRI correlation analysis. The results revealed that neural correlates of the stimulus-action-reward association reside in the putamen, whereas a correlation with reward-prediction error was found largely in the caudate nucleus and ventral striatum. These nonuniform spatiotemporal distributions of neural correlates within the dorsal striatum were maintained consistently at various levels of task difficulty, suggesting a functional difference in the dorsal striatum between the putamen and caudate nucleus during stimulus-action-reward association learning.

Adult↗
PubMed2004

The performance of young deaf children in spatial and temporal number tasks.

Deaf children tend to fall behind in mathematics at school. This problem may be a direct result of particular experiences in the classroom; for example, deaf children may find it hard to follow teachers' presentations of basic, but nevertheless quite abstract, mathematical ideas. Another possibility is that the problem starts before school: They may either be worse than hearing children at early, nonlinguistic number representations, they may be behind in learning the culturally transmitted number string, or both. This may result in deaf children failing to develop informal problem-solving strategies, which prepare most children for the more formal learning of number and arithmetic that they will have to do at school. We compared 3- and 4-year-old deaf and hearing children's ability to remember and to reproduce the number of items in a set of objects. In one condition, we presented all the items together in a spatial array; in another, we presented them one at a time in a temporal sequence. Deaf children performed as well as the hearing children in the temporal tasks, but outperformed their hearing counterparts in the spatial task. These results suggest that preschool deaf children's number representation is at least as advanced as that of hearing children, and that they are actually better than hearing children at representing the number of objects in spatial arrays. We conclude that deaf children's difficulties with mathematical learning are not a consequence of a delay in number representation. We also conclude that deaf children should benefit from mathematical instruction that emphasizes spatial representation.

Analysis of Variance↗
PubMed2006

Essential role of the hippocampal formation in rapid learning of higher-order sequential associations.

Evidence from studies of amnesia and functional imaging in humans suggest that the medial temporal lobe is involved in the learning of higher-order sequences during acquisition of serial reaction time (SRT) tasks. However, these studies have not clarified whether the hippocampal formation specifically is essential to this type of learning. Here, we developed a rodent model of the SRT task to examine the specific roles of the hippocampal formation in learning first- and second-order conditional sequences. Selective hippocampal formation lesions speeded performance and impaired accuracy on both first- and second-order sequences. Additionally, whereas controls distinguished the sequences based on their complexity alone, animals with hippocampal lesions initially differentiated sequences only by their length. Over multiple sessions, hippocampal rats gradually differentiated sequences by their complexity and not length, similar to control subjects. These findings indicate that the hippocampal formation itself plays an essential role in rapid acquisition of higher-order sequence representations. Extrahippocampal systems can also acquire complex sequential representations, albeit via a gradual learning mechanism.

Animals↗
PubMed1996

Reduced perceptual dimensionality in extrafoveal vision.

The classification behaviour of human observers with respect to compound Gabor signals is tested at foveal and extrafoveal retinal positions. Classification performance is analysed in terms of a probabilistic classification model recently proposed by Rentschler, Jüttner and Caelli [(1994) Vision Research, 34, 669-687]. The analysis allows inferences about structure and dimensionality of the individual internal representations underlying the classification task and their temporal evolution during the learning process. Using this technique it is found that the internal representations of direct and eccentric viewing are intrinsically incommensurable, in the sense that extrafoveal pattern representations are characterized by a lower perceptual dimension in feature space relative to the corresponding physical input signals, whereas foveal representations are not. The observed deficits cannot be renormalized by size scaling (cortical magnification); however, they can be partially reduced by learning although the learning progress strongly depends on the observer's practice. The structural incommensurability between foveal and extrafoveal representations poses constraints on possible forms of foveal-extrafoveal interaction, which might have implications on related perceptual phenomena such as visual stability across saccadic eye movements.

Adult↗
PubMed1999

The learning of first and second person pronouns in English: network models and analysis.

Although most English-speaking children master the correct use of first and second person pronouns by three years, some children show persistent reversal errors in which they refer to themselves as you and to others as me. Recently, such differences have been attributed to the relative availability of overheard speech during the learning process. The present study tested this proposal with feed-forward neural networks learning these pronouns. Network learning speed and analysis of their knowledge representations confirmed the importance of exposure to shifting reference provided by overheard speech. Errorless pronoun learning was linked to the amount of overheard speech, interactions with a greater number of speakers, and prior knowledge of the basic-level kind PERSON.

Child↗
PubMed2025

Deep learning-based cell-specific gene regulatory networks inferred from single-cell multiome data.

Gene regulatory networks (GRNs) provide a global representation of how genetic/genomic information is transferred in living systems and are a key component in understanding genome regulation. Single-cell multiome data provide unprecedented opportunities to reconstruct GRNs at fine-grained resolution. However, the inference of GRNs is hindered by insufficient single omic profiles due to the characteristic high loss rate of single-cell sequencing data. In this study, we developed scMultiomeGRN, a deep learning framework to infer transcription factor (TF) regulatory networks via unique integration of single-cell genomic (single-cell RNA sequencing) and epigenomic (single-cell ATAC sequencing) data. We create scMultiomeGRN to elucidate these networks by conceptualizing TF network graph structures. Specifically, we build modality-specific neighbor aggregators and cross-modal attention modules to learn latent representations of TFs from single-cell multi-omics. We demonstrate that scMultiomeGRN outperforms state-of-the-art models on multiple benchmark datasets involved in diseases and health. Via scMultiomeGRN, we identified Alzheimer's disease-relevant regulatory network of SPI1 and RUNX1 for microglia. In summary, scMultiomeGRN offers a deep learning framework to identify cell type-specific gene regulatory network from single-cell multiome data.

Deep Learning↗
PubMed1997

Symmetry breaking and training from incomplete data with Radial Basis Boltzmann Machines.

A Radial Basis Boltzmann Machine (RBBM) is a specialized Boltzmann Machine architecture that combines feed-forward mapping with probability estimation in the input space, and for which very efficient learning rules exist. The hidden representation of the network displays symmetry breaking as a function of the noise in the dynamics. Thus, generalization can be studied as a function of the noise in the neuron dynamics instead of as a function of the number of hidden units. We show that the RBBM can be seen as an elegant alternative of k-nearest neighbor, leading to comparable performance without the need to store all data. We show that the RBBM has good classification performance compared to the MLP. The main advantage of the RBBM is that simultaneously with the input-output mapping, a model of the input space is obtained which can be used for learning with missing values. We derive learning rules for the case of incomplete data, and show that they perform better on incomplete data than the traditional learning rules on a 'repaired' data set.

Computer Simulation↗
PubMedDate not supplied

BYY harmony learning, structural RPCL, and topological self-organizing on mixture models.

The Bayesian Ying-Yang (BYY) harmony learning acts as a general statistical learning framework, featured by not only new regularization techniques for parameter learning but also a new mechanism that implements model selection either automatically during parameter learning or via a new class of model selection criteria used after parameter learning. In this paper, further advances on BYY harmony learning by considering modular inner representations are presented in three parts. One consists of results on unsupervisedmixture models, ranging from Gaussian mixture based Mean Square Error (MSE) clustering, elliptic clustering, subspace clustering to NonGaussian mixture based clustering not only with each cluster represented via either Bernoulli-Gaussian mixtures or independent real factor models, but also with independent component analysis implicitly made on each cluster. The second consists of results on supervised mixture-of-experts (ME) models, including Gaussian ME, Radial Basis Function nets, and Kernel regressions. The third consists of two strategies for extending the above structural mixtures into self-organized topological maps. All these advances are introduced with details on three issues, namely, (a) adaptive learning algorithms, especially elliptic, subspace, and structural rival penalized competitive learning algorithms, with model selection made automatically during learning; (b) model selection criteria for being used after parameter learning, and (c) how these learning algorithms and criteria are obtained from typical special cases of BYY harmony learning.

Bayes Theorem↗
PubMed2026

Conditional Diffusion Model-Based Method for Annotation of Antibiotic Resistance Gene Properties.

The crisis of bacterial antibiotic resistance, which has led to a decline in the effectiveness of antibiotics originally used to combat bacterial infections, has emerged as an urgent challenge for public health. Antibiotic resistance genes (ARGs) are one of the key reasons for bacteria to develop resistance to antibiotics. Therefore, accurately identifying and annotating the critical properties of ARGs is of great importance for addressing the antibiotic resistance emergency. Although existing deep learning models demonstrate remarkable effectiveness in extracting local features from sequence data, they still face limitations in the capacity to further gain the enriched latent representations within the data. To address the critical challenge of extracting higher-quality representations from ARGs sequence data, we propose a novel ARGs properties annotation method based on the conditional diffusion model which is used to learn latent representations through domain-specific knowledge injection. Specifically, during the conditional information integration phase, we systematically incorporate ARGs' domain knowledge to guide the diffusion process in generating high-quality latent representations. To overcome information redundancy caused by direct concatenation of conditional information and intermediate features, we design a cross-attention mechanism that enables feature fusion between heterogeneous information sources, thereby enhancing further the quality of obtained representations. Experimental results on widely used data sets demonstrate the framework's effectiveness in achieving superior prediction performance compared to existing methods.

Anti-Bacterial Agents↗