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 181 records · Page 10Linked to original sources
PubMed2004

Conditioning and cognition.

Animals' abilities to use internal representations of absent objects to guide adaptive behavior and acquire new information, and to represent multiple spatial, temporal, and object properties of complex events and event sequences, may underlie many aspects of human perception, memory, and symbolic thought. In this review, two classes of simple associative learning tasks that address these core cognitive capacities are discussed. The first set, including reinforcer revaluation and mediated learning procedures, address the power of Pavlovian conditioned stimuli to gain access, through learning, to representations of upcoming events. The second set of investigations concern the construction of complex stimulus representations, as illustrated in studies of contextual learning, the conjunction of explicit stimulus elements in configural learning procedures, and recent studies of episodic-like memory. The importance of identifying both cognitive process and brain system bases of performance in animal models is emphasized.

Animals↗
PubMed1999

Central representation of time during motor learning.

This study stemmed from the observation that the brain of human as well as nonhuman primates is capable of forming and memorizing remarkably accurate internal representations of the dynamics of the arm. These dynamics establish a functional relation between applied force and ensuing arm motion, a relation that generally is quite complex and nonlinear. Current evidence shows that the motor control system is capable of adapting to perturbing forces that depend on motion variables such as position, velocity, and acceleration. The experiments we report here were aimed at establishing whether or not the motor system also may adapt to forces that depend explicitly on time rather than on motion variables. Surprisingly, the experiments suggest a negative answer. When asked to compensate for a predictable and repeated time-varying pattern of disturbing forces, subjects learned to counteract the disturbance by producing forces that did not depend on time but on the velocity and the position of the arm. We conclude from this evidence that time and time-dependent dynamics are not explicitly represented within the neural structures that are responsible for motor adaptation. Although our findings are not sufficient to rule out the presence of a timing structure within the central nervous system, they are consistent with other investigations that conspicuously failed to find evidence for such a central clock.

Adaptation, Physiological↗
PubMed2000

Latent learning in medial temporal amnesia: evidence for disrupted representational but preserved attentional processes.

Damage to the hippocampus and medial temporal (MT) structures can lead to anterograde amnesia and may also impair latent learning, in which prior exposure to cues affects their subsequent associability. Normally, latent learning may reflect both representational and attentional mechanisms. Prior work has suggested that individuals with MT amnesia have specific deficits in representational processing; thus, latent learning that invokes primarily representational mechanisms might be especially impaired in MT amnesia. The current results provide preliminary confirmation of this prediction. In Experiment 1, a latent learning paradigm expected to invoke representational mechanisms was impaired in individuals with MT amnesia, whereas in Experiment 2, a paradigm expected to invoke other attentional mechanisms was spared in individuals with MT amnesia. This suggests the representational and attentional components of latent learning are dissociable and differentially affected in anterograde amnesia.

Adult↗
PubMed2001

Praxis lateralization: errors in right and left hemisphere stroke.

Whereas the representations of skilled movements in most right handers are stored in the left hemisphere, the right hemisphere's contribution to action planning remains unclear. We investigated error patterns in left (LHD) and right hemisphere damaged (RHD) subjects as well as normal control subjects (C) to determine if specific components of action programs may be processed by the right hemisphere or bilaterally represented. We had these subjects perform gestures to verbal command with the ipsilesional limb. Although the LHD group made significantly more qualitative errors than the C and RHD groups, the RHD subjects produced a number of apraxic errors. Specifically, the LHD group produced a wide range of spatiotemporal and conceptual errors for both transitive and intransitive gestures, while the RHD group made specific spatial and temporal errors primarily when performing transitive gestures. These findings support the postulate that the left hemisphere stores the spatiotemporal and conceptual representations of learned skilled movements, while several specific components of action programs, such as external configuration (limb orientation) and timing, may have bihemispheric representations.

Aged↗
PubMed1998

Learning viewpoint-invariant face representations from visual experience in an attractor network.

In natural visual experience, different views of an object or face tend to appear in close temporal proximity as an animal manipulates the object or navigates around it, or as a face changes expression or pose. A set of simulations is presented which demonstrate how viewpoint-invariant representations of faces can be developed from visual experience by capturing the temporal relationships among the input patterns. The simulations explored the interaction of temporal smoothing of activity signals with Hebbian learning in both a feedforward layer and a second, recurrent layer of a network. The feedforward connections were trained by competitive Hebbian learning with temporal smoothing of the post-synaptic unit activities. The recurrent layer was a generalization of a Hopfield network with a low-pass temporal filter on all unit activities. The combination of basic Hebbian learning with temporal smoothing of unit activities produced an attractor network learning rule that associated temporally proximal input patterns into basins of attraction. These two mechanisms were demonstrated in a model that took grey-level images of faces as input. Following training on image sequences of faces as they changed pose, multiple views of a given face fell into the same basin of attraction, and the system acquired representations of faces that were approximately viewpoint-invariant.

Animals↗
PubMed2003

Learning optimized features for hierarchical models of invariant object recognition.

There is an ongoing debate over the capabilities of hierarchical neural feedforward architectures for performing real-world invariant object recognition. Although a variety of hierarchical models exists, appropriate supervised and unsupervised learning methods are still an issue of intense research. We propose a feedforward model for recognition that shares components like weight sharing, pooling stages, and competitive nonlinearities with earlier approaches but focuses on new methods for learning optimal feature-detecting cells in intermediate stages of the hierarchical network. We show that principles of sparse coding, which were previously mostly applied to the initial feature detection stages, can also be employed to obtain optimized intermediate complex features. We suggest a new approach to optimize the learning of sparse features under the constraints of a weight-sharing or convolutional architecture that uses pooling operations to achieve gradual invariance in the feature hierarchy. The approach explicitly enforces symmetry constraints like translation invariance on the feature set. This leads to a dimension reduction in the search space of optimal features and allows determining more efficiently the basis representatives, which achieve a sparse decomposition of the input. We analyze the quality of the learned feature representation by investigating the recognition performance of the resulting hierarchical network on object and face databases. We show that a hierarchy with features learned on a single object data set can also be applied to face recognition without parameter changes and is competitive with other recent machine learning recognition approaches. To investigate the effect of the interplay between sparse coding and processing nonlinearities, we also consider alternative feedforward pooling nonlinearities such as presynaptic maximum selection and sum-of-squares integration. The comparison shows that a combination of strong competitive nonlinearities with sparse coding offers the best recognition performance in the difficult scenario of segmentation-free recognition in cluttered surround. We demonstrate that for both learning and recognition, a precise segmentation of the objects is not necessary.

Learning↗
PubMed2002

The role of level of representation in the use of paired associate learning for rehabilitation of alexia.

Patients with phonological alexia (difficulty reading pseudowords) frequently have concomitant difficulty reading functor words and verbs compared with concrete nouns. The current study compares two techniques for helping two patients with phonological alexia regain the ability to read functors and verbs. One technique follows the approach of reorganization of function, while the other relies on the stimulation approach. Study 1, employing a reorganization approach, resulted in both patients increasing their reading accuracy from approximately 10 to 90% or greater. Study 2, using a stimulation approach, resulted in significant improvement, however neither patient was able to achieve accuracy greater than 59%. Study 3 reverted back to the reorganization approach using the same words from Study 2. Both patients demonstrated significant success, achieving 90% or greater accuracy. Whereas the reorganization approach meets with far greater success than the stimulation approach, both approaches can be seen as instances of paired associate learning. An explanation of the advantage of the reorganization approach is developed which focuses on the nature of the pairings in the paired associate learning paradigm: it is proposed that pairings within the same level of representation are easier to learn than pairings that cut across levels of representation.

Aged↗
PubMed2000

[Flexibility of mental representations of spatial information as a function of perspective during learning].

View-based theories of the mental representation of spatial information claim that distinct views experienced during learning are represented separately in memory. Networks of such views are considered to be the basis for spatial navigation. Two experiments (N = 56) investigated the role of observer perspective on the resulting mental representation when learning a spatial configuration on the computer. Learning in route perspective, which induced the impression of passive navigation through the configuration, was compared with a survey perspective, which consisted of an overview of the whole configuration from one point of view. In accordance with view-based theories, previously seen views could be identified faster and with less error than new views for both perspectives during learning. Recoding the information into the alternative perspective was also possible. If participants were asked to integrate distinct route views into a survey view during learning, the flexibility of the resulting mental representation was greatly increased. This indicates that conscious processes such as imagery play an important role in the integration of spatial knowledge.

Adult↗
PubMedDate not supplied

When less is more: how infants learn to form an abstract categorical representation of support.

Two experiments explored how infants learn to form an abstract categorical representation of support (i.e., on) when habituated to few (i.e., 2) or many (i.e., 6) examples of the relation. When habituated to 2 pairs of objects in a support relation, 14-month-olds, but not 10-month-olds, formed the abstract spatial category (i.e., generalized the relation to novel objects). When habituated to 6 object pairs in a support relation, infants did not attend to the relation. The results indicate that infants learn to form an abstract spatial category of support between 10 and 14 months and that having fewer object pairs depicting this relation facilitates their acquisition of the abstract categorical representation.

Cognition↗
PubMed2004

Temporally asymmetric learning supports sequence processing in multi-winner self-organizing maps.

We examine the extent to which modified Kohonen self-organizing maps (SOMs) can learn unique representations of temporal sequences while still supporting map formation. Two biologically inspired extensions are made to traditional SOMs: selection of multiple simultaneous rather than single "winners" and the use of local intramap connections that are trained according to a temporally asymmetric Hebbian learning rule. The extended SOM is then trained with variable-length temporal sequences that are composed of phoneme feature vectors, with each sequence corresponding to the phonetic transcription of a noun. The model transforms each input sequence into a spatial representation (final activation pattern on the map). Training improves this transformation by, for example, increasing the uniqueness of the spatial representations of distinct sequences, while still retaining map formation based on input patterns. The closeness of the spatial representations of two sequences is found to correlate significantly with the sequences' similarity. The extended model presented here raises the possibility that SOMs may ultimately prove useful as visualization tools for temporal sequences and as preprocessors for sequence pattern recognition systems.

Computer Simulation↗
PubMed1988

Organization of the zebra finch song control system: I. Representation of syringeal muscles in the hypoglossal nucleus.

Understanding the representation of learned skills in the brain requires that one know the neural substrate for those skills. The avian song control system uses auditory information to establish and modify motor programs, which provide patterns for the excitation of individual muscles. In the present study, a combination of neurophysiological and anatomical techniques was used to map the representation of syringeal muscles in the tracheosyringeal part of the hypoglossal nucleus of adult male zebra finches. Microstimulation revealed that control zones for individual muscles are arranged along the rostrocaudal axis of the nucleus. The ventralis and dorsalis muscles have the largest domains, located at the rostral and caudal ends of the nucleus, respectively. The retrograde tracer fluorogold was applied to the muscles and confirmed this pattern. The muscle map obtained will provide a useful tool for further study of the convergence of muscle representation and sound representation in the more central portions of the song control pathway. This knowledge is essential for understanding how learned sounds are perceived and produced.

Animals↗
PubMed2001

Birdsong: models and mechanisms.

Recent studies have provided important information concerning the neural signals that subserve vocal learning in songbirds: advanced signal processing techniques are beginning to clarify the behavioral trajectories followed by developing birds; single-unit physiology in behaving animals is providing important clues about sensory and motor representations during learning; in vitro whole-cell recordings are revealing patterns of synaptic communication; and experimental alterations in song behavior have advanced our understanding of specific structure-function relationships. The construction of theoretical and computational models will be crucial in integrating such disparate experimental results.

Animals↗
PubMed1997

Dissociating types of mental computation.

A fundamental issue in the study of cognition and the brain is the nature of mental computation. How far does this depend on internally represented systems of rules, expressed as strings of symbols with a syntax, as opposed to more distributed neural systems, operating subsymbolically and without syntax? The mental representation of the regular and irregular past tense of the English verb has become a crucial test case for this debate. Single-mechanism approaches argue that current multilayer connectionist networks can account for the learning and representation both of regular and of irregular forms. Dual-mechanism approaches, although accepting connectionist accounts for the irregular forms, argue that a symbolic, rule-based system is required to explain the properties of the regular past tense and, by extension, the properties of language and cognition in general. We show here that the regular and irregular past tense are supported by different neural systems, which can become dissociated by damage to the brain. This is evidence for functional and neurological distinctions in the types of mental computation that support these different aspects of linguistic and cognitive performance.

Adult↗
PubMedDate not supplied

Scopolamine enhances generalization between odor representations in rat olfactory cortex.

Acetylcholine (ACh) has a critical, modulatory role in plasticity in many sensory systems. In the rat olfactory system, both behavioral and physiological data indicate that ACh may be required for normal odor memory and synaptic plasticity. Based on these data, neural network models have hypothesized that ACh muscarinic receptors reduce interference between learned cortical representations of odors within the piriform cortex. In this study, odor receptive fields of rat anterior piriform cortex (aPCX) single-units for alkane odors were mapped before and after either a systemic injection of the muscarinic receptor antagonist scopolamine (0.5 mg/kg) or aPCX surface application of 500 microM scopolamine (or saline/ACSF controls). Cross-habituation between alkanes differing by two to four carbons was then examined following a 50-sec habituating stimulus. The results demonstrate that neither aPCX spontaneous activity nor odor-evoked activity (receptive field) was affected by scopolamine, but that cross-habituation in aPCX neurons was enhanced significantly by either systemic or cortical scopolamine. These results indicate that scopolamine selectively enhances generalization between odor representations in aPCX in a simple memory task. Given that ACh primarily affects intracortical association fibers in the aPCX, the results support a role for the association system in odor memory and discrimination and indicate an important ACh modulatory control over this basic sensory process.

Animals↗
PubMed1995

Remodelling of hand representation in adult cortex determined by timing of tactile stimulation.

The primate somatosensory cortex, which processes tactile stimuli, contains a topographic representation of the signals it receives, but the way in which such maps are maintained is poorly understood. Previous studies of cortical plasticity indicated that changes in cortical representation during learning arise largely as a result of hebbian synaptic change mechanisms. Here we show, using owl monkeys trained to respond to specific stimulus sequence events, that serial application of stimuli to the fingers results in changes to the neuronal response specificity and maps of the hand surfaces in the true primary somatosensory cortical field (S1 area 3b). In this representational remodelling stimuli applied asychronously to the fingers resulted in these fingers being integrated in their representation, whereas fingers to which stimuli were applied asynchronously were segregated in their representation. Ventroposterior thalamus response maps derived in these monkeys were not equivalently reorganized. This representational plasticity appears to be cortical in origin.

Animals↗
PubMed2004

Reward representations and reward-related learning in the human brain: insights from neuroimaging.

This review outlines recent findings from human neuroimaging concerning the role of a highly interconnected network of brain areas including orbital and medial prefrontal cortex, amygdala, striatum and dopaminergic mid-brain in reward processing. Distinct reward-related functions can be attributed to different components of this network. Orbitofrontal cortex is involved in coding stimulus reward value and in concert with the amygdala and ventral striatum is implicated in representing predicted future reward. Such representations can be used to guide action selection for reward, a process that depends, at least in part, on orbital and medial prefrontal cortex as well as dorsal striatum.

Animals↗
PubMed2025

CASTER-DTA: Equivariant Graph Neural Networks for Predicting Drug-Target Affinity.

Accurately determining the binding affinity of a ligand with a protein is important for drug design, development, and screening. With the advent of accessible protein structure prediction methods such as AlphaFold, predicted protein 3D structures are readily available; however, methods for predicting binding affinity currently do not take full advantage of 3D protein information. Here, we present CASTER-DTA (Cross-Attention with Structural Target Equivariant Representations for Drug-Target Affinity), which uses an equivariant graph neural network to learn more robust protein representations alongside a standard graph neural network to learn molecular representations to predict drug-target affinity. We augment these representations by incorporating an attention-based mechanism between protein residues and drug atoms to improve interpretability. We show that CASTER-DTA represents a state-of-the-art improvement on multiple benchmarks for predicting drug-target affinity and that it generates novel insights for several related tasks. We then apply CASTER-DTA to create a large resource of the binding affinities of every FDA-approved drug against every protein in the human proteome and make these predictions freely available for download. We also make available a web server for researchers to apply a pretrained CASTER-DTA model for predicting binding affinities between arbitrary proteins and drugs.

deep learning↗
PubMed2001

Invariant object recognition in the visual system with error correction and temporal difference learning.

It has been proposed that invariant pattern recognition might be implemented using a learning rule that utilizes a trace of previous neural activity which, given the spatio-temporal continuity of the statistics of sensory input, is likely to be about the same object though with differing transforms in the short time scale. Recently, it has been demonstrated that a modified Hebbian rule which incorporates a trace of previous activity but no contribution from the current activity can offer substantially improved performance. In this paper we show how this rule can be related to error correction rules, and explore a number of error correction rules that can be applied to and can produce good invariant pattern recognition. An explicit relationship to temporal difference learning is then demonstrated, and from this further learning rules related to temporal difference learning are developed. This relationship to temporal difference learning allows us to begin to exploit established analyses of temporal difference learning to provide a theoretical framework for better understanding the operation and convergence properties of these learning rules, and more generally, of rules useful for learning invariant representations. The efficacy of these different rules for invariant object recognition is compared using VisNet, a hierarchical competitive network model of the operation of the visual system.

Learning↗