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 1,189 records · Page 66Linked to original sources
PubMed2005

Assessment of phonological representations in children with speech impairment.

PURPOSE: This study explored the use of assessment tasks to examine underlying phonological representations in preschool children with speech impairment. The study also investigated the association between performance on phonological representation tasks and phonological awareness development. METHOD: The performance of 9 children (aged 3;09 [years; months] to 5;03) with moderate or severe speech impairment and 17 children of the same age with typical speech development was investigated on a range of novel receptive-based assessment tasks designed to tap underlying phonological representations. RESULTS: Preschool children with speech impairment experienced more difficulty judging correct and incorrect speech productions of familiar multisyllable words and showed inferior performance in the ability to learn nonwords as compared to children without speech impairment. Performance on these tasks was moderately correlated with phonological awareness ability. CLINICAL IMPLICATIONS: Factors such as the precision and accessibility of underlying phonological representations of spoken words may contribute to problems in phonological awareness and subsequent reading development for young children with speech impairment. Receptive-based assessments that examine underlying phonological representations provide clinically relevant information for children with speech impairment.

Child, Preschool↗
PubMed2001

Ockham's razor modeling of the matrisome channels of the basal ganglia thalamocortical loops.

A functional model of the basal ganglia-thalamocortical (BTC) loops is described. In our modeling effort, we try to minimize the complexity of our starting hypotheses. For that reason, we call this type of modeling Ockham's razor modeling. We have the additional constraint that the starting assumptions should not contradict experimental findings about the brain. First assumption: The brain lacks direct representation of paths but represents directions (called speed fields in control theory). Then control should be concerned with speed-field tracking (SFT). Second assumption: Control signals are delivered upon differencing in competing parallel channels of the BTC loops. This is modeled by extending SFT with differencing that gives rise to the robust Static and Dynamic State (SDS) feedback-controlling scheme. Third assumption: Control signals are expressed in terms of a gelatinous medium surrounding the limbs. This is modeled by expressing parameters of motion in parameters of the external space. We show that corollaries of the model fit properties of the BTC loops. The SDS provides proper identification of motion related neuronal groups of the putamen. Local minima arise during the controlling process that works in external space. The model explains the presence of parallel channels as the means to avoiding such local minima. Stability conditions of the SDS predict that the initial phase of learning is mostly concerned with selection of sign for the inverse dynamics. The model provides a scalable controller. State description in external space instead of configurational space reduces the dimensionality problem. Falsifying experiment is suggested. Computer experiments demonstrate the feasibility of the approach. We argue that the resulting scheme has a straightforward connectionist representation exhibiting population coding and Hebbian learning properties.

Basal Ganglia↗
PubMed1988

Coordinates transformation and learning control for visually-guided voluntary movement with iteration: a Newton-like method in a function space.

In order to control visually-guided voluntary movements, the central nervous system (CNS) must solve the following three computational problems at different levels: (1) determination of a desired trajectory in the visual coordinates, (2) transformation of the coordinates of the desired trajectory to the body coordinates and (3) generation of motor command. In this paper, the second and the third problems are treated at computational, representational and hardware levels of Marr. We first study the problems at the computational level, and then propose an iterative learning scheme as a possible algorithm. This is a trial and error type learning such as repetitive training of golf swing. The amount of motor command needed to coordinate activities of many muscles is not determined at once, but in a step-wise, trial and error fashion in the course of a set of repetitions. Actually, the motor command in the (n + 1)-th iteration is a sum of the motor command in the n-th iteration plus two modification terms which are, respectively, proportional to acceleration and speed errors between the desired trajectory and the realized trajectory in the n-th iteration. We mathematically formulate this iterative learning control as a Newton-like method in functional spaces and prove its convergence under appropriate mathematical conditions with use of dynamical system theory and functional analysis. Computer simulations of this iterative learning control of a robotic manipulator in the body or visual coordinates are shown. Finally, we propose that areas 2, 5, and 7 of the sensory association cortex are possible sites of this learning control. Further we propose neural network model which acquires transformation matrices from acceleration or velocity to motor command, which are used in these schemes.

Central Nervous System↗
PubMed1997

The evolutionary origin of the language areas in the human brain. A neuroanatomical perspective.

The capacity to learn syntactic rules is a hallmark of the human species, but whether this has been acquired by the process of natural selection has been the subject of controversy. Furthermore, the cortical localization of linguistic capacities has prompted some authors to suggest a modular representation of language in the brain. In this paper, we rather propose that the neural device involved in language is embedded into a large-scale neurocognitive network comprising widespread connections between the temporal, parietal and frontal (especially prefrontal) cortices. This network is involved in the temporal organization of behavior and motor sequences, and in working (active) memory, a sort of short-term memory that participates in immediate cognitive processing. In human evolution, a precondition for language was the establishment of strong cortico-cortical interactions in the postrolandic cortex that enabled the development of multimodal associations. Wernicke's area originated as a converging place in which such associations (concepts) acquired a phonological correlate. We postulate that these phonological representations projected into inferoparietal areas, which were connected to the incipient Broca's area, thus forming a working memory circuit for processing and learning complex vocalizations. As a result of selective pressure for learning capacity and memory storage, this device yielded a sophisticated system able to generate complicated utterances (precursors of syntax) as it became increasingly connected with other brain regions, especially in the prefrontal cortex. This view argues for a gradual origin of the neural substrate for language as required by natural selection.

Animals↗
PubMed2002

Evaluating theories of bird song learning: implications for future directions.

Studies of birdsong learning have stimulated extensive hypotheses at all levels of behavioral and physiological organization. This hypothesis building is valuable for the field and is consistent with the remarkable range of issues that can be rigorously addressed in this system. The traditional instructional (template) theory of song learning has been challenged on multiple fronts, especially at a behavioral level by evidence consistent with selectional hypotheses. In this review I highlight the caveats associated with these theories to better define the limits of our knowledge and identify important experiments for the future. The sites and representational forms of the various conceptual entities posited by the template theory are unknown. The distinction between instruction and selection in vocal learning is not well established at a mechanistic level. There is as yet insufficient neurophysiological data to choose between competing mechanisms of error-driven learning and reinforcement learning. Both may obtain for vocal learning. The possible role of sleep in acoustic or procedural memory consolidation, while supported by some physiological observations, does not yet have support in the behavioral literature. The remarkable expansion of knowledge in the past 20 years and the recent development of new technologies for physiological and behavioral experiments should permit direct tests of these theories in the coming decade.

Animal Communication↗
PubMed1993

Visual and statistical assessment of spatial clustering in mapped data.

Maps have seen increasing use to examine regional variation in health, but there has been little research on the visual perception of spatial patterns in mapped data. Theories of graphical perception suggest that the interpretation of maps is complex relative to other types of graphical material. This paper describes an experiment in which observers assessed a series of maps with respect to their amount of clustering. Maps with various types of spatial pattern were visually distinguishable; comparisons between variants of the same map, however, using different shading and plotting symbols indicated that the method of data representation also had a strong effect on visual perception. There was some evidence for a learning effect in complex maps. The relationship between the visual assessments and a statistical measure of spatial autocorrelation was significant but imperfect.

Cluster Analysis↗
PubMed2006

The roles of various projection areas of the motor cortex in the reorganization of the natural coordination of head and forelimb movements in dogs.

A food-related operant reaction was developed in dogs, in which animals had to maintain tonic elevation of the forelimb to hold a bowl while eating with the head tilted towards the feeder. The acquisition of this reaction involved rearrangement of the natural coordination of head and limb movements which appeared at an early stage of training of the dogs. Forelimb elevation was initially accompanied by anticipatory raising of the head, while lowering of the head led to lowering of the elevated limb. Limb elevation could only be maintained in the posture in which the head was raised. The new coordination required for obtaining food, contrary to the innate coordination and consisting of tonic elevation of the limb with the head lowered, could only be achieved as a result of training. Previous studies have established that lesioning of the primary motor cortex (MI) in the hemisphere contralateral to the working limb leads to stable impairment of the learned coordination, with regression to the initial coordination. The present report describes studies of the effects of local lesions of various projection areas of MI on performance of the learned coordination. Dogs which had acquired the learned operant reaction requiring the new head/limb coordination showed impairment only after lesioning of the representation area of the working limb in the MI; lesioning of the representation area of the head had no such effect.

Animals↗
PubMed1990

When representations conflict with reality: the preschooler's problem with false beliefs and "false" photographs.

It has been argued that young preschoolers cannot correctly attribute a false belief to a deceived actor (Wimmer & Perner, 1983). Some researchers claim that the problem lies in the child's inadequate epistemology (Chandler & Boyes, 1982; Wellman, 1988); as such, it is specific to the child's theory of mind and no such problem should appear in reasoning about nonmental representations. This prediction is tested below in the "false photograph" task: here an actor takes a photograph of an object in location X; the object is then moved to location Y. Preschool subjects are asked: "In the picture, where is the object?" Results indicate that photographs are no easier to reason about than are beliefs. Manipulations to boost performance on the photograph task proved ineffective. Further, an explanation of the failure as a processing limitation having nothing to do with the representational nature of beliefs or photographs was ruled out. It is argued that young children's failure on the false belief task is not due to an inadequate epistemology (though they may have one) and is symptomatic of a larger problem with representations.

Child, Preschool↗
PubMed2006

Simulating single word processing in the classic aphasia syndromes based on the Wernicke-Lichtheim-Geschwind theory.

The Wernicke-Lichtheim-Geschwind (WLG) theory of the neurobiological basis of language is of great historical importance, and it continues to exert a substantial influence on most contemporary theories of language in spite of its widely recognized limitations. Here, we suggest that neurobiologically grounded computational models based on the WLG theory can provide a deeper understanding of which of its features are plausible and where the theory fails. As a first step in this direction, we created a model of the interconnected left and right neocortical areas that are most relevant to the WLG theory, and used it to study visual-confrontation naming, auditory repetition, and auditory comprehension performance. No specific functionality is assigned a priori to model cortical regions, other than that implicitly present due to their locations in the cortical network and a higher learning rate in left hemisphere regions. Following learning, the model successfully simulates confrontation naming and word repetition, and acquires a unique internal representation in parietal regions for each named object. Simulated lesions to the language-dominant cortical regions produce patterns of single word processing impairment reminiscent of those postulated historically in the classic aphasia syndromes. These results indicate that WLG theory, instantiated as a simple interconnected network of model neocortical regions familiar to any neuropsychologist/neurologist, captures several fundamental "low-level" aspects of neurobiological word processing and their impairment in aphasia.

Aphasia↗
PubMed2006

Temporal grouping affects the automatic processing of deviant sounds.

The present study investigates the influence of auditory temporal grouping on automatic sound processing. We compared the mismatch negativity (MMN) event-related potential elicited by successive deviant tones in a condition in which sounds were organized in tone pairs with respective MMN obtained in a non-paired sequence. When successive deviants belonged to different tone pairs the second deviant elicited a higher MMN amplitude than the corresponding deviant in the non-paired condition. Moreover, this temporal grouping had immediate effects on the extraction of rules as revealed by the finding that first and single deviants at the second position of a tone pair elicited larger MMNs than corresponding deviants in the non-paired condition or first and single deviants at the first position of a tone pair. Results demonstrate that auditory objects generated by temporal proximity are treated as relatively encapsulated units and that automatic deviance-detection may rely on multiple hierarchically organized representations of sound sequences.

Adult↗
PubMed2001

Short-term functional plasticity in the human auditory cortex: an fMRI study.

Applying functional magnetic resonance imaging (fMRI) techniques, hemodynamic responses elicited by sequences of pure tones of 950 Hz (standard) and deviant tones of 952, 954, and 958 Hz were measured before and 1 week after subjects had been trained at frequency discrimination for five sessions (over 1 week) using an oddball procedure. The task of the subject was to detect deviants differing from the standard stimulus. Frequency discrimination improved during the training session for three subjects (performance gain: T+) but not for three other subjects (no performance gain: T-). Hemodynamic responses in the auditory cortex comprising the planum temporale, planum polare and sulcus temporalis superior significantly decreased during training only for the T+ group. These activation changes were strongest for those stimuli accompanied by the strongest performance gain (958 and 954 Hz). There was no difference with respect to the hemodynamic responses in the auditory cortex for the T- group and the control group (CO) who did not received any pitch discrimination training. The results suggest a plastic reorganization of the cortical representation for the trained frequencies which can be best explained on the basis of 'fast learning' theories.

Acoustic Stimulation↗
PubMed1997

DHEA-S selectively impairs contextual-fear conditioning: support for the antiglucocorticoid hypothesis.

The authors had reported that glucocorticoids play a selective role in fear conditioning. The adrenal steroid dehydroepiandrosterone (DHEA) has been reported to act as a functional antiglucocorticoid. If DHEA has antiglucocorticoid properties, then its effects on fear conditioning might resemble those produced by adrenalectomy. The authors now report that chronic exposure to high levels of dehydroepiandrosterone sulfate (DHEA-S; converted in vivo to DHEA) produced the same pattern of results as adrenalectomy. Specifically, treatment with DHEA-S impaired contextual fear conditioning 24 hr after conditioning but not immediately after conditioning, and like adrenalectomy, DHEA-S had no effect on auditory-cue fear conditioning. Preexposure to the context before drug treatment eliminated the amnestic effects of DHEA-S, suggesting that, like adrenalectomy, DHEA-S exerted its effect by interfering with the construction of a contextual memory representation. Thus, DHEA appears to act as a functional antiglucocorticoid in the processes that mediate learning and memory.

Acoustic Stimulation↗
PubMed2000

Reward processing in primate orbitofrontal cortex and basal ganglia.

This article reviews and interprets neuronal activities related to the expectation and delivery of reward in the primate orbitofrontal cortex, in comparison with slowly discharging neurons in the striatum (caudate, putamen and ventral striatum, including nucleus accumbens) and midbrain dopamine neurons. Orbitofrontal neurons showed three principal forms of reward-related activity during the performance of delayed response tasks, namely responses to reward-predicting instructions, activations during the expectation period immediately preceding reward and responses following reward. These activations discriminated between different rewards, often on the basis of the animals' preferences. Neurons in the striatum were also activated in relation to the expectation and detection of reward but in addition showed activities related to the preparation, initiation and execution of movements which reflected the expected reward. Dopamine neurons responded to rewards and reward-predicting stimuli, and coded an error in the prediction of reward. Thus, the investigated cortical and basal ganglia structures showed multiple, heterogeneous, partly simultaneous activations which were related to specific aspects of rewards. These activations may represent the neuronal substrates of rewards during learning and established behavioral performance. The processing of reward expectations suggests an access to central representations of rewards which may be used for the neuronal control of goaldirected behavior.

Animals↗
PubMed2004

Cocaine makes actions insensitive to outcomes but not extinction: implications for altered orbitofrontal-amygdalar function.

Addiction is characterized by persistent drug-seeking despite adverse consequences or outcomes. Such persistent behavior may result from drug-induced brain changes that increase the control of behavior by associations between antecedent cues and responses. However, it is equally plausible that brain changes cause a decrease in the control of behavior by the value of likely outcomes. To test whether drug exposure can cause persistent behavior, and to distinguish between these two accounts of such behavior, we tested cocaine-experienced rats in a Pavlovian 'reinforcer devaluation' task, which provides independent assessments of the control of behavior by antecedent cues and outcome representations. We found that cocaine exposure caused persistent responding in this setting a month after the last drug treatment, and that this deficit resulted from an inability to use representations of outcome value to guide behavior rather than from changes in stimulus-response learning or response inhibition.

Amygdala↗
PubMed2004

Analysis and synthesis of textured motion: particles and waves.

Natural scenes contain a wide range of textured motion phenomena which are characterized by the movement of a large amount of particle and wave elements, such as falling snow, wavy water, and dancing grass. In this paper, we present a generative model for representing these motion patterns and study a Markov chain Monte Carlo algorithm for inferring the generative representation from observed video sequences. Our generative model consists of three components. The first is a photometric model which represents an image as a linear superposition of image bases selected from a generic and overcomplete dictionary. The dictionary contains Gabor and LoG bases for point/particle elements and Fourier bases for wave elements. These bases compete to explain the input images and transfer them to a token (base) representation with an O(10(2))-fold dimension reduction. The second component is a geometric model which groups spatially adjacent tokens (bases) and their motion trajectories into a number of moving elements--called "motons." A moton is a deformable template in time-space representing a moving element, such as a falling snowflake or a flying bird. The third component is a dynamic model which characterizes the motion of particles, waves, and their interactions. For example, the motion of particle objects floating in a river, such as leaves and balls, should be coupled with the motion of waves. The trajectories of these moving elements are represented by coupled Markov chains. The dynamic model also includes probabilistic representations for the birth/death (source/sink) of the motons. We adopt a stochastic gradient algorithm for learning and inference. Given an input video sequence, the algorithm iterates two steps: 1) computing the motons and their trajectories by a number of reversible Markov chain jumps, and 2) learning the parameters that govern the geometric deformations and motion dynamics. Novel video sequences are synthesized from the learned models and, by editing the model parameters, we demonstrate the controllability of the generative model.

Algorithms↗
PubMed1992

A quantitative study of the Ca2+/calmodulin sensitivity of adenylyl cyclase in Aplysia, Drosophila, and rat.

Studies in Aplysia and Drosophila have suggested that Ca2+/calmodulin-sensitive adenylyl cyclase may act as a site of convergence for the cellular representations of the conditioned stimulus (Ca2+ influx) and unconditioned stimulus (facilitatory transmitter) during elementary associative learning. This hypothesis predicts that the rise in intracellular free Ca2+ concentration produced by spike activity during the conditioned stimulus will cause an increase in the activity of adenylyl cyclase. However, published values for the Ca2+ sensitivity of Ca2+/calmodulin-sensitive adenylyl cyclase in mammals and in Drosophila vary widely. The difficulty in evaluating whether adenylyl cyclase would be activated by physiological elevations in intracellular Ca2+ levels is in part a consequence of the use of Ca2+/EGTA buffers, which are prone to several types of errors. Using a procedure that minimizes these errors, we have quantified the Ca2+ sensitivity of adenylyl cyclase in membranes from Aplysia, Drosophila, and rat brain with purified species-specific calmodulins. In all three species, adenylyl cyclase was activated by an increase in free Ca2+ concentration in the range caused by spike activity. Ca2+ sensitivity was dependent on both calmodulin concentration and Mg2+ concentration. Mg2+ raised the threshold for adenylyl cyclase activation by Ca2+ but also acted synergistically with Ca2+ to activate maximally adenylyl cyclase.

Adenylyl Cyclases↗
PubMed2000

Nonlinear dimensionality reduction by locally linear embedding.

Many areas of science depend on exploratory data analysis and visualization. The need to analyze large amounts of multivariate data raises the fundamental problem of dimensionality reduction: how to discover compact representations of high-dimensional data. Here, we introduce locally linear embedding (LLE), an unsupervised learning algorithm that computes low-dimensional, neighborhood-preserving embeddings of high-dimensional inputs. Unlike clustering methods for local dimensionality reduction, LLE maps its inputs into a single global coordinate system of lower dimensionality, and its optimizations do not involve local minima. By exploiting the local symmetries of linear reconstructions, LLE is able to learn the global structure of nonlinear manifolds, such as those generated by images of faces or documents of text.

Algorithms↗
PubMed2000

Influence of feature training on the formation of exemplar-specific representations in children with mental retardation.

Whether exemplar-specific or prototype representations were primarily used by children with mild mental retardation in a categorization task consisting of two groups of make-believe animals whose features were ill-defined was examined. Categorization performance of familiar, prototypic, and novel test phase exemplars indicated that these children were able to acquire and use both types of representations. A feature-labeling training procedure, expected to promote the formation of exemplar-specific representations prior to exemplar classification, was also incorporated within the current study's design. It was unclear whether such training had any impact on the formation of exemplar-specific representations.

Attention↗