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 217 records · Page 12Linked to original sources
PubMed2006

Object segmentation and recovery via neural oscillators implementing the similarity and prior knowledge gestalt rules.

Object recognition requires the solution of the binding and segmentation problems, i.e., grouping different features to achieve a coherent representation. Synchronization of neural activity in the gamma-band, associated with gestalt perception, has often been proposed as a putative mechanism to solve these problems, not only as to low-level processing, but also in higher cortical functions. In the present work, a network of Wilson-Cowan oscillators is used to segment simultaneous objects, and recover an object from partial or corrupted information, by implementing two gestalt rules: similarity and prior knowledge. The network consists of H different areas, each devoted to representation of a particular feature of the object, according to a topological organization. The similarity law is realized via lateral intra-area connections, arranged as a "Mexican-hat". Prior knowledge is realized via inter-area connections, which link properties belonging to a previously memorized object. A global inhibitor allows segmentation of several objects avoiding interference. Simulation results, performed using three simultaneous input objects, show that the network is able to detect an object even in difficult conditions (i.e., when some features are absent or shifted with respect to the original one). Moreover, the trade-off between sensitivity (capacity to detect true positives) and specificity (capacity to reject false positives) can be controlled acting on the extension of lateral synapses (i.e., on the level of accepted similarity). Finally, the network can also deal with correlated objects, i.e., objects which have some common features. Simulations performed using a different number of objects (2, 3, 4 or 5) suggest that the network is able to segment and recall up to four objects, but the oscillation frequency must increase, the lower the number of objects simultaneously present. The model, although quite simpler compared with neurophysiology, may represent a theoretical framework for the analysis of the relationships between object representation, memory, learning, and gamma-band activity. In particular, it extends previous studies on autoassociative memory since it exploits not only oscillatory dynamics, but also a topological organization of features.

Gestalt Theory↗
PubMed2004

New training strategies for constructive neural networks with application to regression problems.

Regression problem is an important application area for neural networks (NNs). Among a large number of existing NN architectures, the feedforward NN (FNN) paradigm is one of the most widely used structures. Although one-hidden-layer feedforward neural networks (OHL-FNNs) have simple structures, they possess interesting representational and learning capabilities. In this paper, we are interested particularly in incremental constructive training of OHL-FNNs. In the proposed incremental constructive training schemes for an OHL-FNN, input-side training and output-side training may be separated in order to reduce the training time. A new technique is proposed to scale the error signal during the constructive learning process to improve the input-side training efficiency and to obtain better generalization performance. Two pruning methods for removing the input-side redundant connections have also been applied. Numerical simulations demonstrate the potential and advantages of the proposed strategies when compared to other existing techniques in the literature.

Algorithms↗
PubMed2005

The representation of time for motor learning.

We have identified factors that control precise motor timing by studying learning in smooth pursuit eye movements. Monkeys tracked a target that moved horizontally for a fixed time interval before changing direction through the addition of a vertical component of motion. After repeated presentations of the same target trajectory, infrequent probe trials of purely horizontal target motion evoked a vertical eye movement around the time when the change in target direction would have occurred. The pursuit system timed the vertical eye movement by keeping track of the duration of horizontal target motion and by measuring the distance the target traveled before changing direction, but not by learning the position in space where the target changed direction. We conclude that high temporal precision in motor output relies on multiple signals whose contributions to timing vary according to task requirements.

Animals↗
PubMed2006

Predictive neural coding of reward preference involves dissociable responses in human ventral midbrain and ventral striatum.

Food preferences are acquired through experience and can exert strong influence on choice behavior. In order to choose which food to consume, it is necessary to maintain a predictive representation of the subjective value of the associated food stimulus. Here, we explore the neural mechanisms by which such predictive representations are learned through classical conditioning. Human subjects were scanned using fMRI while learning associations between arbitrary visual stimuli and subsequent delivery of one of five different food flavors. Using a temporal difference algorithm to model learning, we found predictive responses in the ventral midbrain and a part of ventral striatum (ventral putamen) that were related directly to subjects' actual behavioral preferences. These brain structures demonstrated divergent response profiles, with the ventral midbrain showing a linear response profile with preference, and the ventral striatum a bivalent response. These results provide insight into the neural mechanisms underlying human preference behavior.

Adult↗
PubMedDate not supplied

The integrity of the syllable in developmental apraxia of speech.

UNLABELLED: Three children with developmental apraxia of speech (DAS) identified the number of syllables in words, judged intrasyllabic sound positions, and constructed syllable shapes within monosyllabic frames. Results suggest that DAS children demonstrate an apparent breakdown in the ability to perceive "syllableness" and to access and compare syllable representations with regard to position and structure. Based on these findings, DAS is viewed as a disorder characterized by an impoverished phonological representation system. LEARNING OUTCOMES: On the basis of this article, the reader will be able to (1) describe the deficits in syllabic perception demonstrated by some participants with DAS; (2) examine the utility of metalinguistic tasks in the differential diagnosis of DAS; and (3) evaluate the findings with respect to competing theoretical perspectives on DAS.

Apraxias↗
PubMed2002

The planum temporale as a computational hub.

It is increasingly recognized that the human planum temporale is not a dedicated language processor, but is in fact engaged in the analysis of many types of complex sound. We propose a model of the human planum temporale as a computational engine for the segregation and matching of spectrotemporal patterns. The model is based on segregating the components of the acoustic world and matching these components with learned spectrotemporal representations. Spectrotemporal information derived from such a 'computational hub' would be gated to higher-order cortical areas for further processing, leading to object recognition and the perception of auditory space. We review the evidence for the model and specific predictions that follow from it.

Algorithms↗
PubMed2000

A learning rule for dynamic recruitment and decorrelation.

The interest in neuronal networks originates for a good part in the option not to construct, but to train them. The mechanisms governing synaptic modifications during such training are assumed to depend on signals locally available at the synapses. In contrast, the performance of a network is suitably measured on a global scale. Here we propose a learning rule that addresses this conflict. It is inspired by recent physiological experiments and exploits the interaction of inhibitory input and backpropagating action potentials in pyramidal neurons. This mechanism makes information on the global scale available as a local signal. As a result, several desirable features can be combined: the learning rule allows fast synaptic modifications approaching one-shot learning. Nevertheless, it leads to stable representations during ongoing learning. Furthermore, the response properties of the neurons are not globally correlated, but cover the whole stimulus space.

Animals↗
PubMed2002

Comparing the brain areas supporting nondeclarative categorization and recognition memory.

Brain areas associated with both nondeclarative categorization and recognition memory were identified and contrasted using functional magnetic resonance imaging (fMRI) of healthy volunteers. Activity during dot-pattern categorization and recognition were compared with a control task (counting dots) in two separate groups of participants (n=5 each). The network of areas associated with nondeclarative categorization was found to include bilateral inferior prefrontal and parietal cortical areas that have been implicated in several other studies of categorization. During recognition, increased activity was found in posterior visual areas, the precuneus, posterior cingulate and right prefrontal cortex. Using the common control condition as a reference, recognition and categorization were contrasted and recognition was found to evoke more activity in posterior early visual cortex, the precuneus, right medial temporal lobe and right dorso-lateral prefrontal cortex. Previous research has implicated changes in visual representation in learning a category of dot-pattern by comparing activity evoked by categorical and non-categorical stimuli. The current findings support those results and additionally identify brain areas active during categorization that are involved in expressing this category knowledge.

Brain↗
PubMed2005

Error-driven knowledge restructuring in categorization.

Knowledge restructuring occurs when people shift to a new strategy or representation during learning. Although knowledge restructuring can frequently be experimentally encouraged, there are instances in which people resist restructuring and continue to use an expedient but imperfect initial strategy. The authors report 3 category learning experiments that reconciled those conflicting outcomes by postulating that, for restructuring to occur, learners must be dissatisfied with their knowledge and a usable alternative must be available. In line with expectation, restructuring was elicited only when an alternative strategy was pointed out and when people's initial expedient strategy entailed performance error. Neither error nor information about the alternative strategy by itself was sufficient to induce restructuring.

Attention↗
PubMed1999

Webs, cell assemblies, and chunking in neural nets: introduction.

This introduction to Wickelgren (1992), describes a theory of idea representation and learning in the cerebral cortex and seven properties of Hebb's (1949) formulation of cell assemblies that have played a major role in all such neural net models. Ideas are represented in the cerebral cortex by webs (innate cell assemblies), using sparse coding with sparse, all-or-none, innate linking. Recruiting a web to represent a new idea is called chunking. The innate links that bind the neurons of a web are basal dendritic synapses. Learning modifies the apical dendritic synapses that associate neurons in one web to neurons in another web.

Brain↗
PubMed2006

Long-term motor cortex plasticity induced by an electronic neural implant.

It has been proposed that the efficacy of neuronal connections is strengthened when there is a persistent causal relationship between presynaptic and postsynaptic activity. Such activity-dependent plasticity may underlie the reorganization of cortical representations during learning, although direct in vivo evidence is lacking. Here we show that stable reorganization of motor output can be induced by an artificial connection between two sites in the motor cortex of freely behaving primates. An autonomously operating electronic implant used action potentials recorded on one electrode to trigger electrical stimuli delivered at another location. Over one or more days of continuous operation, the output evoked from the recording site shifted to resemble the output from the corresponding stimulation site, in a manner consistent with the potentiation of synaptic connections between the artificially synchronized populations of neurons. Changes persisted in some cases for more than one week, whereas the output from sites not incorporated in the connection was unaffected. This method for inducing functional reorganization in vivo by using physiologically derived stimulus trains may have practical application in neurorehabilitation after injury.

Action Potentials↗
PubMed1993

Control of Three- and Four-Joint Arm Movement: Strategies for a Manipulator With Redundant Degrees of Freedom.

Control of arm movements when the number of joints exceeds the degrees of freedom necessary for the task requires a strategy for selecting specific arm configurations out of an infinite number of possibilities. This report reviews strategies used by human subjects to control the shoulder, elbow, and wrist (three degrees of freedom) while moving a pointer to positions in a horizontal plane (two degrees of freedom). Analysis of final arm configurations assumed when the pointer was at the target showed the following: (a) Final arm configurations were virtually independent of the configuration at the start of the pointing movement, (b) subjects avoided configurations subjectively felt to be uncomfortable (e.g., those with extreme flexion or extension of the wrist), and (c) the results could be simulated by assigning hypothetical cost functions to each joint and selecting the arm configuration that minimized the sum of the costs. The fitted cost functions qualitatively agreed with psychophysically determined comfort; they appeared to depend on joint angle and on muscular effort. Simple neural networks can learn implicit representations of these cost functions and use them to specify final arm configurations. The minimum cost principle can be extended to movements that use the fingers as a fourth movable segment. For this condition, however, experiments showed that final configurations of the arm depended upon initial configurations. Analysis of movement trajectories for arms with three degrees of freedom led to a control model in which the minimum cost principle is augmented by a mechanism that distributes required joint movements economically among the three joints and a mechanism that implements a degree of mass-spring control.

Journal Article↗
PubMed2001

A knowledge-based system for patient image pre-fetching in heterogeneous database environments--modeling, design, and evaluation.

When performing primary reading on a newly taken radiological examination, a radiologist often needs to reference relevant prior images of the same patient for confirmation or comparison purposes. Support of such image references is of clinical importance and may have significant effects on radiologists' examination reading efficiency, service quality, and work satisfaction. To effectively support such image reference needs, we proposed and developed a knowledge-based patient image pre-fetching system, addressing several challenging requirements of the application that include representation and learning of image reference heuristics and management of data-intensive knowledge inferencing. Moreover, the system demands an extensible and maintainable architecture design capable of effectively adapting to a dynamic environment characterized by heterogeneous and autonomous data source systems. In this paper, we developed a synthesized object-oriented entity- relationship model, a conceptual model appropriate for representing radiologists' prior image reference heuristics that are heuristic oriented and data intensive. We detailed the system architecture and design of the knowledge-based patient image pre-fetching system. Our architecture design is based on a client-mediator-server framework, capable of coping with a dynamic environment characterized by distributed, heterogeneous, and highly autonomous data source systems. To adapt to changes in radiologists' patient prior image reference heuristics, ID3-based multidecision-tree induction and CN2-based multidecision induction learning techniques were developed and evaluated. Experimentally, we examined effects of the pre-fetching system we created on radiologists' examination readings. Preliminary results show that the knowledge-based patient image pre-fetching system more accurately supports radiologists' patient prior image reference needs than the current practice adopted at the study site and that radiologists may become more efficient, consultatively effective, and better satisfied when supported by the pre-fetching system than when relying on the study site's pre-fetching practice.

Artificial Intelligence↗
PubMed2005

Visual learning by imitation with motor representations.

We propose a general architecture for action (mimicking) and program (gesture) level visual imitation. Action-level imitation involves two modules. The viewpoint Transformation (VPT) performs a "rotation" to align the demonstrator's body to that of the learner. The Visuo-Motor Map (VMM) maps this visual information to motor data. For program-level (gesture) imitation, there is an additional module that allows the system to recognize and generate its own interpretation of observed gestures to produce similar gestures/goals at a later stage. Besides the holistic approach to the problem, our approach differs from traditional work in i) the use of motor information for gesture recognition; ii) usage of context (e.g., object affordances) to focus the attention of the recognition system and reduce ambiguities, and iii) use iconic image representations for the hand, as opposed to fitting kinematic models to the video sequence. This approach is motivated by the finding of visuomotor neurons in the F5 area of the macaque brain that suggest that gesture recognition/imitation is performed in motor terms (mirror) and rely on the use of object affordances (canonical) to handle ambiguous actions. Our results show that this approach can outperform more conventional (e.g., pure visual) methods.

Algorithms↗
PubMed2004

Modeling developmental transitions in adaptive resonance theory.

Neural networks are applied to a theoretical subject in developmental psychology: modeling developmental transitions. Two issues that are involved will be discussed: discontinuities and acquiring qualitatively new knowledge. We will argue that by the appearance of a bifurcation, a neural network can show discontinuities and may acquire qualitatively new knowledge. First, it is shown that biological principles of neurite outgrowth result in self-organization in a neural network, which is strongly dependent on a bifurcation in the activity dynamics. Second, the effect of a bifurcation due to morphological change is investigated in an Adaptive Resonance Theory (ART) network. Exact ART networks with quantitative differences in network structure at the category level show qualitatively different dynamical regimes, which are separated by bifurcations. These qualitative differences in dynamics affect the cognitive function of Exact ART: Representations of learned categories are local or distributed.

Adaptation, Biological↗
PubMed2000

Vision as temporal trace.

Most current models of human and animal vision assume that the processes of vision involve 2D (or even 3D) internal representations of the external world--an iconic representation. Within these models, recognition involves some form of lineal, areal or volumetric comparison of these internal representations (either learned or innate) with current sensory inputs. However, this view has recently come under criticism. In this paper, a neural model of vision is explored in which this iconic world view is replaced by a temporally ordered trace of (essentially) local features. The model employs hierarchical, recurrently linked, self-organizing topological maps.

Animals↗
PubMed2005

The neural correlates of motor skill automaticity.

Acquisition of a new skill is generally associated with a decrease in the need for effortful control over performance, leading to the development of automaticity. Automaticity by definition has been achieved when performance of a primary task is minimally affected by other ongoing tasks. The neural basis of automaticity was examined by testing subjects in a serial reaction time (SRT) task under both single-task and dual-task conditions. The diminishing cost of dual-task performance was used as an index for automaticity. Subjects performed the SRT task during two functional magnetic imaging sessions separated by 3 h of behavioral training over multiple days. Behavioral data showed that, by the end of testing, subjects had automated performance of the SRT task. Before behavioral training, performance of the SRT task concurrently with the secondary task elicited activation in a wide network of frontal and striatal regions, as well as parietal lobe. After extensive behavioral training, dual-task performance showed comparatively less activity in bilateral ventral premotor regions, right middle frontal gyrus, and right caudate body; activity in other prefrontal and striatal regions decreased equally for single-task and dual-task conditions. These data suggest that lateral and dorsolateral prefrontal regions, and their corresponding striatal targets, subserve the executive processes involved in novice dual-task performance. The results also showed that supplementary motor area and putamen/globus pallidus regions showed training-related decreases for sequence conditions but not for random conditions, confirming the role of these regions in the representation of learned motor sequences.

Adult↗