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 793 records · Page 44Linked to original sources
PubMed1985

Role of timing of visual monitoring and motor rehearsal in observational learning of action patterns.

This experiment tested the hypothesis that observational learning is enhanced by visual monitoring of enactments that is optimally timed for conception-action matching and by motor rehearsal that serves to refine the cognitive representation. Subjects observed a modeled action pattern, after which they enacted it with either concurrent, delayed, or no visual monitoring. They then engaged in motor rehearsal or did not rehearse the action pattern. Development of the cognitive representation of the modeled action was also measured. Concurrent visual monitoring of enactments greatly facilitated observational learning, whereas delayed visual monitoring did not affect the acquisition process. Rehearsal aided cognitive representation and behavioral reproduction. The more accurate the cognitive representation of the modeled action pattern, the more skilled were the subsequent reproductions of it. After gaining proficiency in converting conception to action, subjects showed no decline in reproduction accuracy when modeling and visual monitoring were withdrawn.

Journal Article↗
PubMed2005

Learning-induced modulation of oscillatory activities in the mammalian olfactory system: the role of the centrifugal fibres.

In the mammalian olfactory system, oscillations related to odour representation have been described in field potential activities. Previous results showed that in olfactory bulb (OB) of awake rats engaged in an olfactory learning, odour presentation produced a decrease of oscillations in gamma frequency range (60-90 Hz) associated with a power increase in beta frequency range (15-40 Hz). This response pattern was strongly amplified in trained animals. The aim of this work was twofold: whether learning also induces similar changes in OB target structures and whether such OB response depends on its centrifugal inputs. Local field potentials (LFPs) were recorded through chronically implanted electrodes in the OB, piriform and enthorhinal cortices of freely moving rats performing an olfactory discrimination. Oscillatory activities characteristics (amplitude, frequency and time-course) were extracted in beta and gamma range by a wavelet analysis. First, we found that odour induced beta oscillatory activity was present not only in the OB, but also in the other olfactory structures. In each recording site, characteristics of the beta oscillatory responses were dependent of odour, structure and learning level. Unilateral section of the olfactory peduncle was made before training, and LFPs were symmetrically recorded in the two bulbs all along the acquisition of the learning task. Data showed that deprivation of centrifugal feedback led to an increase of spontaneous gamma activity. Moreover, under this condition olfactory learning was no longer associated with the typical large beta band. As a whole, learning modulation of the beta oscillatory response in olfactory structures may reflect activity of a distributed functional network involved in odour representation.

Action Potentials↗
PubMed2026

scFANCL: Dual contrastive learning with false-negative correction at cell level for single-cell RNA-seq clustering.

BACKGROUND: Single-cell RNA sequencing (scRNA-seq) enables cellular characterization at single-cell resolution. However, its high dimensionality, sparsity, and noise make clustering challenging. Approaches utilizing contrastive learning and data augmentation have been introduced to improve representation quality for scRNA-seq clustering. In particular, dual contrastive frameworks combining instance- and cluster-level objectives can capture both cell-cell similarities and inter-cluster variations. However, existing dual contrastive frameworks focus primarily on discrete cluster boundaries, neglecting the biological continuity inherent in scRNA-seq data. METHODS: We propose scFANCL, a dual contrastive framework designed to capture biological continuity in scRNA data. Rather than treating all non-augmented samples as negatives, scFANCL applies a cosine-similarity-based threshold to exclude cells of the same type from the negative pool, preserving continuous transcriptional relationships among them while maintaining inter-cluster separation. RESULTS: Extensive experiments across seven publicly available scRNA-seq datasets demonstrated that scFANCL achieves competitive clustering performance compared with existing baseline methods, consistently yielding high ARI and NMI scores across datasets of varying size and complexity. Ablation studies further confirmed the contribution of the false negative filtering component, showing measurable improvements over variants without filtering. Downstream analyses further suggest that the learned embeddings may reflect biologically meaningful transcriptional transitions, including continuous differentiation trajectories within related cell types. The source code is available at https://github.com/mjuailab/scFANCL . CONCLUSIONS: scFANCL addresses a key limitation of conventional contrastive learning by applying a cosine-similarity-based threshold to exclude cells of the same type from the negative pool, thereby preserving biological continuity within cell types while maintaining inter-cluster separation. Evaluations across seven benchmark scRNA-seq datasets demonstrate competitive clustering performance, with learned embeddings capturing biologically meaningful transcriptional structure and characteristics of rare cell populations.

Clustering Algorithms↗
PubMed2002

Fear and power-dominance drive motivation: neural representations and pathways mediating sensory and mnemonic inputs, and outputs to premotor structures.

Based on the available literature on activation of brain structures by fear- and anger-inducing stimuli, on the effects of electrical and chemical stimulation and lesions of candidate structures, and on connectional data, we propose that both the fear and power-dominance drives are represented in four distinct locations: the medial hypothalamus, lateral/dorsolateral periaqueductal gray, midline thalamic nuclei, and medial prefrontal cortex. The hypothalamic fear representation is located in the dorsomedial and posterior hypothalamic nuclei, the midbrain representation in the caudal part of the lateral/dorsolateral periaqueductal gray, the thalamic representation primarily in parts of the paraventricular and reuniens thalamic nuclei, and the cortical representation in prelimbic cortex. The hypothalamic power-dominance representation is located in the anterior hypothalamic nucleus, dorsomedial aspect of the ventromedial nucleus, and in adjacent parts of the medial preoptic area. The corresponding midbrain representation occurs in rostral part of the lateral/dorsolateral periaqueductal gray, and the thalamic representation in parts of the paraventricular, parataenial, and reuniens thalamic nuclei. We discuss sensory/mnemonic inputs to these representations, and outputs to premotor structures in the medulla, caudate-putamen, and cortex, and their differential contributions to involuntary, learned sequential, and voluntary motor acts. We examine potential contributions of neuronal activities in these representations to the subjective awareness of fear and anger.

Animals↗
PubMed2003

Dissociating explicit and implicit category knowledge with fMRI.

Neuroimaging of healthy volunteers identified separate neural systems supporting the expression of category knowledge depending on whether the learning mode was intentional or incidental. The same visual category was learned either intentionally or implicitly by two separate groupsof participants. During a categorization test, functional magnetic resonance imaging (fMRI) was used to compare brain activity evoked by category members and nonmembers. After implicit learning, when participants had learned the category incidentally, decreased occipital activity was observed for novel categorical stimuli compared with noncategorical stimuli. In contrast, after intentional learning, novel categorical stimuli evoked increased activity in the hippocampus, right prefrontal cortex, left inferior temporal cortex, precuneus, and posterior cingulate. Even though the categorization test was identical in the two conditions, the differences in brain activity indicate differing representations of category knowledge depending on whether the category had been learned intentionally or implicitly.

Brain Mapping↗
PubMed1982

Simultaneous second-order conditioning produces S-S learning in conditioned suppression.

Four experiments explored the consequences of simultaneous second-order conditioning of one stimulus (S2) by another (S1) in a conditioned suppression preparation. In two experiments, that mode of presentation produced second-order conditioning of S2 which was attenuated by subsequent extinction of S1. Sequential presentation of the same stimuli produced similar levels of conditioning of S2 which were not affected by extinction of S1, results that replicate previous findings. Two additional experiments found tht the sensitivity of a simultaneous S2 to changes in S1 also depended on that S1 receiving few separate reinforced or nonreinforced presentations prior to second-order conditioning . The results suggest circumstances under which the typical stimulus-response learning observed in second-order conditioned suppression can be converted into stimulus-response learning. An interpretation is suggested in terms of simultaneous presentation generating a unitary representation involving both S2 and S1 rather than an association between them.

Acoustic Stimulation↗
PubMed2004

Paleoenvironmental basis of cognitive evolution in great apes.

A bias favoring tree-dominated habitats and ripe-fruit frugivory has persisted in great ape evolution since the early Miocene. This bias is indicated by fossil ape paleoenvironments, molar morphology, dental microwear, the geographic pattern of extinctions, and extant apes' reliance on wooded settings. The ephemeral aspect of high-quality fruit has placed a premium on cognitive and social means of finding and defending food sources, and appears related to great apes' affinity since the Miocene for wooded, fruit-rich environments. These habitats have, however, undergone a severe withdrawal toward the low latitudes of Africa and Southeast Asia since the late Miocene, corresponding to a decline in the diversity of great apes beginning 9.5 million years ago. Plio-Pleistocene records imply that wooded settings of Africa and SE Asia were prone to substantial fragmentation and coalescence. Once apes were confined to equatorial settings, therefore, habitat instability heightened the spatial/temporal uncertainty of ripe-fruit sources. Prolonged learning, the assignment of attributes to distant places, mental representation, and reliance on fallback foods were all favored in this dynamic environmental context. These abilities helped sustain forest frugivory in most lineages. Fluid social grouping afforded the animals opportunities to locate ephemeral foods in continuous and fragmented forests. Fission-fusion grouping also magnified the problems of object impermanence (of individuals) and dispersion manifested by food sources in the ecological realm. Thus the spatial and temporal dynamics of fruit and wooded habitats since the Miocene are reflected in important components of great ape cognition, foraging, and sociality. In contrast to great apes, cercopithecoid monkeys have increased their plant dietary options and diversified in seasonal environments since the late Miocene. Early hominins eventually severed the habitat bias that characterized the evolution of great apes, and later expanded into diverse environments.

Animals↗
PubMed2003

Learning input correlations through nonlinear temporally asymmetric Hebbian plasticity.

Triggered by recent experimental results, temporally asymmetric Hebbian (TAH) plasticity is considered as a candidate model for the biological implementation of competitive synaptic learning, a key concept for the experience-based development of cortical circuitry. However, because of the well known positive feedback instability of correlation-based plasticity, the stability of the resulting learning process has remained a central problem. Plagued by either a runaway of the synaptic efficacies or a greatly reduced sensitivity to input correlations, the learning performance of current models is limited. Here we introduce a novel generalized nonlinear TAH learning rule that allows a balance between stability and sensitivity of learning. Using this rule, we study the capacity of the system to learn patterns of correlations between afferent spike trains. Specifically, we address the question of under which conditions learning induces spontaneous symmetry breaking and leads to inhomogeneous synaptic distributions that capture the structure of the input correlations. To study the efficiency of learning temporal relationships between afferent spike trains through TAH plasticity, we introduce a novel sensitivity measure that quantifies the amount of information about the correlation structure in the input, a learning rule capable of storing in the synaptic weights. We demonstrate that by adjusting the weight dependence of the synaptic changes in TAH plasticity, it is possible to enhance the synaptic representation of temporal input correlations while maintaining the system in a stable learning regime. Indeed, for a given distribution of inputs, the learning efficiency can be optimized.

Action Potentials↗
PubMed2004

Anticipatory spatial representation of 3D regions explored by sighted observers and a deaf-and-blind-observer.

Viewers who study photographs of scenes tend to remember having seen beyond the boundaries of the view [boundary extension; J. Exp. Psychol. Learn. Mem. Cogn. 15 (1989) 179]. Is this a fundamental aspect of scene representation? Forty undergraduates explored bounded regions of six common (3D) scenes, visually or haptically (while blindfolded) and then the delimiting borders were removed. Minutes later they reconstructed boundary placement. Boundary extension occurred: mean areas were increased by 53% (vision) and by 17% (haptics). A deaf-and-blind woman (KC) haptically explored the same regions. Although a "haptic expert", she too remembered having explored beyond the boundaries, with performance similar to that of the blindfolded-sighted. Boundary extension appears to be a fundamental aspect of spatial cognition. Possibly constrained by the "scope" of the input modality (vision>haptics), this anticipatory spatial representation may facilitate integration of successively perceived regions of the world irrespective of modality and the perceiver's sensory history.

Adult↗
PubMed1997

The neural basis of cognitive development: a constructivist manifesto.

How do minds emerge from developing brains? According to "neural constructivism," the representational features of cortex are built from the dynamic interaction between neural growth mechanisms and environmentally derived neural activity. Contrary to popular selectionist models that emphasize regressive mechanisms, the neurobiological evidence suggests that this growth is a progressive increase in the representational properties of cortex. The interaction between the environment and neural growth results in a flexible type of learning: "constructive learning" minimizes the need for prespecification in accordance with recent neurobiological evidence that the developing cerebral cortex is largely free of domain-specific structure. Instead, the representational properties of cortex are built by the nature of the problem domain confronting it. This uniquely powerful and general learning strategy undermines the central assumption of classical learnability theory, that the learning properties of a system can be deduced from a fixed computational architecture. Neural constructivism suggests that the evolutionary emergence of neocortex in mammals is a progression toward more flexible representational structures, in contrast to the popular view of cortical evolution as an increase in innate, specialized circuits. Human cortical postnatal development is also more extensive and protracted than generally supposed, suggesting that cortex has evolved so as to maximize the capacity of environmental structure to shape its structure and function through constructive learning.

Axons↗
PubMed2004

Developmental regulation of basal ganglia circuitry during the sensitive period for vocal learning in songbirds.

A hallmark of sensitive periods of development is an enhanced capacity for learning, such that experience exerts a profound effect on the brain resulting in the establishment of behaviors and underlying neural circuitry that can last a lifetime. Songbirds, like humans, have a sensitive period for vocal learning: they acquire the sounds used for vocal communication during a restricted period of development. In principle, any organism that undertakes vocal learning is faced with the same challenge: to form some representation of target vocal sounds based on auditory experience, and then to translate that auditory target into a motor program that reproduces the sound. Both birds and humans achieve this translation by using auditory (and other) feedback resulting from incipient vocalizations ("babbling" in humans, "subsong" in birds) to adjust motor commands until vocal output produces a good copy of the target sounds. Similarities between vocal learning in birds and humans suggest that many aspects of the learning process have evolved to meet demands imposed by vocal communication. Thus songbirds provide a valuable animal model in which to study the physiological basis of learned vocal communication and the nature of sensitive periods in general. In this article, I describe aspects of both behavioral and neural frameworks that currently inform our thinking about mechanisms underlying vocal learning and behavior in songbirds, and highlight ideas that may need re-examination.

Animals↗
PubMedDate not supplied

Cortical plasticity during three-week motor skill learning.

The authors studied motor behavior and primary motor (M1) and somatosensory (S1) cortical representations of movement during hand motor skill acquisition over 3 weeks. During four functional MRI sessions 1 week apart, subjects performed simple movements of single fingers and wrist, and a sequential movement of the middle three fingers, contrasted with rest. Half of the subjects practiced the sequential movement, whereas the other half practiced a gross motor task (squeezing a sponge). In both groups, motor performance gradually improved both on the practiced sequence and also on unrelated tests of general hand performance. Similarly, gradual expansion of active M1 and S1 areas was observed with the practiced sequence and with the unpracticed single-finger and wrist movements. Motor practice enhanced overlaps and sharing of cortex (significantly more so in the fine-skill group), preserving somatotopy and the overall extent of the hand representations. Even a limited amount of practice on a complex task can thus lead to both specific and nonspecific improvements in behavior and to expansion of M1 and S1 movement representations. Finally, the extent of active M1 and S1 was significantly correlated with out-of-scanner performance on sequential finger movement and may reflect the current motor ability of the individual.

Adult↗
PubMed2002

The influence of exposure to phonological neighbours on preschoolers' novel word production.

We investigated the influence of exposure to phonologically similar words on four-year-olds' acquisition of novel object words. In Experiment 1, hearing phonological neighbours before learning a new word did not influence children's novel word productions. In Experiment 2, when children heard the phonological neighbours of a novel word after learning a new word, they correctly produced the target word more often than children who did not receive this exposure. These findings suggest that exposing children to similar sounding words after a novel word was introduced may have helped maintain a representation of that word in working memory, leading to enhanced word learning.

Child Language↗
PubMed2004

Slower implicit categorical learning in adult poor readers.

We investigated the relationship between reading and explicit and implicit categorical learning by comparing university students with poor reading to students with normal reading abilities on two categorical learning tasks. One categorical learning task involved sorting simple geometric shapes into two groups according to a unidimensional rule. The sorting rule was easily stated by the participants, consistent with explicit learning, and all participants attained criterion levels of performance. The second task involved the integration of features on different dimensions with a more complex rule that could not be described by participants, even though most could attain criterion levels of performance consistent with implicit learning. Poor readers performed as well as those without reading problems in explicit learning but not in implicit learning. Implicit learning was correlated with word reading, phonological decoding, and orthographic skill, independent of verbal ability. We consider the role of implicit learning in reading, and how a deficit could impair phonological and orthographic representation and processing.

Adult↗
PubMed2002

On the neural correlates of object recognition awareness: relationship to computational activities and activities mediating perceptual awareness.

Based on theoretical considerations of Aurell (1979) and Block (1995), we argue that object recognition awareness is distinct from purely sensory awareness and that the former is mediated by neuronal activities in areas that are separate and distinct from cortical sensory areas. We propose that two of the principal functions of neuronal activities in sensory cortex, which are to provide sensory awareness and to effect the computations that are necessary for object recognition, are dissociated. We provide examples of how this dissociation might be achieved and argue that the components of the neuronal activities which carry the computations do not directly enter the awareness of the subject. The results of these computations are sparse representations (i.e., vector or distributed codes) which are activated by the presentation of particular sensory objects and are essentially engrams for the recognition of objects. These final representations occur in the highest order areas of sensory cortex; in the visual analyzer, the areas include the anterior part of the inferior temporal cortex and the perirhinal cortex. We propose, based on lesion and connectional data, that the two areas in which activities provide recognition awareness are the temporopolar cortex and the medial orbitofrontal cortex. Activities in the temporopolar cortex provide the recognition awareness of objects learned in the remote past (consolidated object recognition), and those in the medial orbitofrontal cortex provide the recognition awareness of objects learned in the recent past. The activation of the sparse representation for a particular sensory object in turn activates neurons in one or both of these regions of cortex, and it is the activities of these neurons that provide the awareness of recognition of the object in question. The neural circuitry involved in the activation of these representations is discussed.

Awareness↗
PubMed2002

Mechanisms of knowledge learning and acquisition.

The mechanism by which knowledge enters into memory has been a source of debate for some time. Theorists have proposed several models that aim at explaining the sequence of events from the perception of a stimulus, to its entrance into long-term storage. Much of this work was prompted by early research into the nuances of classical conditioning where it was first firmly established that organisms are capable of detecting covariations of stimuli within their environment. Subsequent work in the field has shown that these covariations form the basis for the mental representation of our surroundings, as well as the basis of learning. Work within the field of classical conditioning, along with the advance of computer technology and neuroscience has made these architectural models even more complex. Furthermore, experiments designed to support some of these proposed models have revealed that there are several conditions that can either aid or inhibit the transition of information into permanent storage. In this review we explore a number of these models, along with some classic critiques that have been levied against them. We also provide some history into the form of knowledge, termed 'implicit knowledge', as well as some of the proposed mechanisms of implicit knowledge acquisition. We conclude by exploring the newly proposed theoretical framework within which implicit learning theory operates.

Animals↗
PubMed1995

Neural pattern formation via a competitive Hebbian mechanism.

In this contribution we investigate a simple pattern formation process [9,10] based on Hebbian learning and competitive interactions within cortex. This process generates spatial representations of afferent (sensory) information which strongly resemble patterns of response properties of neurons commonly called brain maps. For one of the most thoroughly studied phenomena in cortical development, the formation of topographic maps, orientation and ocular dominance columns in macaque striate cortex, the process, for example, generates the observed patterns of receptive field properties including the recently described correlations between orientation preference and ocular dominance. Competitive Hebbian learning has not only proven to be a useful concept in the understanding of development and plasticity in several brain areas, but the underlying principles have have been successfully applied to problems in machine learning [22]. The model's universality, simplicity, predictive power, and usefulness warrants a closer investigation.

Animals↗
PubMed2001

Superordinate categorization via learned stimulus equivalence: quantity of reinforcement, hedonic value, and the nature of the mediator.

Three experiments examined superordinate categorization via stimulus equivalence training in pigeons. Experiment 1 established superordinate categories by association with a common number of food pellet reinforcers, plus it established generalization to novel photographic stimuli. Experiment 2 documented generalization of choice responding from stimuli signaling different numbers of food pellets to stimuli signaling different delays to food reinforcement. Experiment 3 indicated that different numbers of food pellets did not substitute as discriminative stimuli for the photographic stimuli with which the food pellets had been paired. The collective results suggest that the effective mediator of superordinate categories that are established via learned stimulus equivalence is not likely to be an accurate representation of the reinforcer, neither is it likely to be a distinctive response that is made to the discriminative stimulus. Motivational or emotional mediation is a more likely account.

Affect↗