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 631 records · Page 35Linked to original sources
PubMedDate not supplied

Auditory pathway encoding and neural plasticity in children with learning problems.

An inability to process auditory information, especially speech, characterizes many children with learning and attention problems. Our working hypothesis is that these speech-sound perception problems arise, at least in some cases, from faulty representation of the speech signal in central auditory centers. Preconscious neurophysiologic representation of sound structure by central auditory pathway neurons can be reflected by subcortical and cortical aggregate neural responses. These neurophysiologic responses can be modified by perceptual learning. Our research has shown that some children with learning problems demonstrate abnormal perception and neural representation of certain speech sounds. Differences between normal and learning-impaired groups can be attributable to aspects of neural synchrony that are reflected in aggregate neural responses. Deficiencies in neural synchrony in these children are apparent in subcortical (as well as cortical) representations of speech-sound structure, and these timing deficits are related to performance on speech-sound perception and learning measures. Moreover, impaired perception and neurophysiologic encoding of speech sounds can be improved with cue enhancement and can be modified by perceptual learning associated with auditory training.

Child↗
PubMed2003

Chunking during human visuomotor sequence learning.

Motor sequence learning is a process whereby a series of elementary movements is re-coded into an efficient representation for the entire sequence. Here we show that human subjects learn a visuomotor sequence by spontaneously chunking the elementary movements, while each chunk acts as a single memory unit. The subjects learned to press a sequence of 10 sets of two buttons through trial and error. By examining the temporal patterns with which subjects performed a visuomotor sequence, we found that the subjects performed the 10 sets as several clusters of sets, which were separated by long time gaps. While the overall performance time decreased by repeating the same sequence, the clusters became clearer and more consistent. The cluster pattern was uncorrelated with the distance of hand movements and was different across subjects who learned the same sequence. We then split a learned sequence into three segments, while preserving or destroying the clusters in the learned sequence, and shuffled the segments. The performance on the shuffled sequence was more accurate and quicker when the clusters in the original sequence were preserved than when they were destroyed. The results suggest that each cluster is processed as a single memory unit, a chunk, and is necessary for efficient sequence processing. A learned visuomotor sequence is hierarchically represented as chunks that contain several elementary movements. We also found that the temporal patterns of sequence performance transferred from the nondominant to dominant hand, but not vice versa. This may suggest a role of the dominant hemisphere in storage of learned chunks. Together with our previous unit-recording and imaging studies that used the same learning paradigm, we predict specific roles of the dominant parietal area, basal ganglia, and presupplementary motor area in the chunking.

Cluster Analysis↗
PubMed1984

Learning of Blissymbols and manual signs.

Blissymbols and manual signs are two representational systems widely used as alternatives to spoken communication for persons unable to speak understandably. They were compared to determine if there is any advantage of one over the other in ease of learning or in retention. Immediate and 1-day posttest measures were obtained from 20 7- and 8-year-old children engaged in a paired-associate transfer-of-training task using manual signs and Blissymbols. Results showed no overall difference between signs and symbols in number of correct responses on either the immediate or 1-day posttest.

Child↗
PubMed2006

Recurrent neural network architecture with pre-synaptic inhibition for incremental learning.

We propose a recurrent neural network architecture that is capable of incremental learning and test the performance of the network. In incremental learning, the consistency between the existing internal representation and a new sequence is unknown, so it is not appropriate to overwrite the existing internal representation on each new sequence. In the proposed model, the parallel pathways from input to output are preserved as possible, and the pathway which has emitted the wrong output is inhibited by the previously fired pathway. Accordingly, the network begins to try other pathways ad hoc. This modeling approach is based on the concept of the parallel pathways from input to output, instead of the view of the brain as the integration of the state spaces. We discuss the extension of this approach to building a model of the higher functions such as decision making.

Affect↗
PubMed1980

Preoperative overtraining protects against form learning deficits after lateral occipital lesions in Galago senegalensis.

Bushbabies (Galago senegalensis) trained on a form discrimination task in a two-choice apparatus prior to partial disruption of the central field representation of vision in striate cortex were found to be protected from postoperative deficit. These same animals were deficient in learning novel form discriminations. Pre- and postoperative tests of these subjects on discrimination of fine stripe patterns and of small food objects gave no evidence of reduced epicritic visual capacities. Results are discussed in terms of an interpretive role in form learning for areas of central field representation in primary visual cortex.

Animals↗
PubMed2000

Reconstructing mental object representations: a machine vision approach to human visual recognition.

This paper introduces a new approach to assess visual representations underlying the recognition of objects. Human performance is modeled by CLARET, a machine learning and matching system, based on inductive logic programming and graph matching principles. The model is applied to data of a learning experiment addressing the role of prior experience in the ontogenesis of mental object representations. Prior experience was varied in terms of sensory modality, i.e. visual versus haptic versus visuohaptic. The analysis revealed distinct differences between the representational formats used by subjects with haptic versus those with no prior object experience. These differences suggest that prior haptic exploration stimulates the evolution of object representations which are characterized by an increased differentiation between attribute values and a pronounced structural encoding.

Computer Simulation↗
PubMed2005

Coherent theta-band EEG activity predicts item-context binding during encoding.

Episodic memories consist of semantic information coupled with a rich array of contextual detail. Here, we investigate the neural processes by which information about the sensory context of a learning event is "bound" to the semantic representation of the to-be-encoded item. We present evidence that item-context binding during encoding is mediated by frontoposterior electroencephalographic (EEG) phase locking within and between hemispheres in the theta (4-8 Hz) band. During a task in which subjects encoded words in different font colors, later memory for the word was associated with sustained frontal theta activity and frontoposterior theta-band coherence, primarily within the left hemisphere. When the word-color association was later successfully retrieved, however, neurons synchronized their theta-band responses bilaterally in a more sustained fashion, particularly during the latter part of the stimulus epoch (>800 ms). Our results confirm the importance of functional coupling between frontal and posterior regions for successful encoding. One interpretation of these data is hemispheric contributions to item and context encoding may be asymmetric, with left hemisphere coherence facilitating semantic processing of an item and right hemisphere coherence facilitating processing of sensory context. Theta-band coherence may be an important mechanism by which brain networks exchange information during learning.

Adult↗
PubMed2006

The relation between children's conceptual functioning with color and color term acquisition.

Young children experience considerable difficulty in learning their first few color terms. One explanation for this difficulty is that initially they lack a conceptual representation of color sufficiently abstract to support word meaning. This hypothesis, that prior to learning color terms children do not represent color as an abstraction, was tested in two experiments using samples of 25- to 39-month-olds and 20- to 32-month-olds. Children's ability to conceptually represent color and their knowledge of color terms were assessed, and a strong association was found between the ability to make inferences based on color and the comprehension of color words. Children who did not comprehend color terms were unsuccessful at a conceptual task that required them to represent color as a property independent of the particular objects that displayed it. The results suggest that the initial absence of an abstract representation of color contributes to the difficulty that young children encounter when first learning color words.

Association Learning↗
PubMed1997

Long short-term memory.

Learning to store information over extended time intervals by recurrent backpropagation takes a very long time, mostly because of insufficient, decaying error backflow. We briefly review Hochreiter's (1991) analysis of this problem, then address it by introducing a novel, efficient, gradient-based method called long short-term memory (LSTM). Truncating the gradient where this does not do harm, LSTM can learn to bridge minimal time lags in excess of 1000 discrete-time steps by enforcing constant error flow through constant error carousels within special units. Multiplicative gate units learn to open and close access to the constant error flow. LSTM is local in space and time; its computational complexity per time step and weight is O(1). Our experiments with artificial data involve local, distributed, real-valued, and noisy pattern representations. In comparisons with real-time recurrent learning, back propagation through time, recurrent cascade correlation, Elman nets, and neural sequence chunking, LSTM leads to many more successful runs, and learns much faster. LSTM also solves complex, artificial long-time-lag tasks that have never been solved by previous recurrent network algorithms.

Algorithms↗
PubMed2003

Neural circuits subserving the retrieval and maintenance of abstract rules.

Behavior is often governed by abstract rules or instructions for behavior that can be abstracted from one context and applied to another. Prefrontal cortex (PFC) is thought to be important for representing rules, although the contributions of ventrolateral (VLPFC) and dorsolateral (DLPFC) regions remain under-specified. In the present study, event-related fMRI was used to examine abstract rule representation in humans. Prior to scanning, subjects learned to associate unfamiliar shapes and nonwords with particular rules. During each fMRI trial, presentation of one of these cues was followed by a delay and then by sample and probe stimuli. Match and non-match rules required subjects to indicate whether or not the sample and probe matched; go rules required subjects to make a response that was not contingent on the sample/probe relation. Left VLPFC, parietal cortex, and pre-SMA exhibited sensitivity to rule type during the cue and delay periods. Delay-period activation in these regions, but not DLPFC, was greater when subjects had to maintain response contingencies (match, non-match) relative to when the cue signaled a specific response (go). In contrast, left middle temporal cortex exhibited rule sensitivity during the cue but not delay period. These results support the hypothesis that VLPFC interacts with temporal cortex to retrieve semantic information associated with a cue and with parietal cortex to retrieve and maintain relevant response contingencies across delays. Future investigations of cross-regional interactions will enable full assessment of this account. Collectively, these results demonstrate that multiple, neurally separable processes are recruited during abstract rule representation.

Acoustic Stimulation↗
PubMed2005

Protein structure and fold prediction using Tree-Augmented naïve Bayesian classifier.

Due to the large volume of protein sequence data, computational methods to determine the structure class and the fold class of a protein sequence have become essential. Several techniques based on sequence similarity, Neural Networks, Support Vector Machines (SVMs), etc. have been applied. Since most of these classifiers use binary classifiers for multi-classification, there may be (N) c2 classifiers required. This paper presents a framework using the Tree-Augmented Bayesian Networks (TAN) which performs multi-classification based on the theory of learning Bayesian Networks and using improved feature vector representation of (Ding et al., 2001). In order to enhance TAN's performance, pre-processing of data is done by feature discretization and post-processing is done by using Mean Probability Voting (MPV) scheme. The advantage of using Bayesian approach over other learning methods is that the network structure is intuitive. In addition, one can read off the TAN structure probabilities to determine the significance of each feature (say, hydrophobicity) for each class, which helps to further understand the complexity in protein structure. The experiments on the datasets used in three prominent recent works show that our approach is more accurate than other discriminative methods. The framework is implemented on the BAYESPROT web server and it is available at http://www-appn.comp.nus.edu.sg/~bioinfo/bayesprot/Default.htm. More detailed results are also available on the above website.

Algorithms↗
PubMed2004

Experience-dependent changes in cortical whisker representation in the adult mouse: a 2-deoxyglucose study.

Sensory experience and learning can modify cortical body maps. We have previously reported that 3 days of classical conditioning, in which stimulation of a row of whiskers was paired with tail shock, produced an expansion of the cortical representation of the "trained row" labeled with 2-deoxyglucose (2DG), in layer IIIb and IV of the barrel cortex. The present study examined plastic remodelling of the vibrissal cortical representation after pairing whisker stimulation with a drop of sweet water. Cortical representations of rows of whiskers were mapped by 2DG autoradiography after 3 days and 2 months of training. The training resulted in enlargement of the cortical representation of vibrissae involved in the stimulus pairing compared with the contralateral representation of a row of whiskers, that were not touched during the training. This modification of whisker representation was different after short-term and long-term appetitive training. After three pairing sessions, changes in the width of cortical representation were visible in layers II/IIIa (29%) and layers V/VI (28%). After 2 months of training, significant changes in the width of cortical representation row B were found only in layer IV (41%). The changes were not observed in animals, that received whisker stimulation alone or in those who were subjected to training with unpaired stimuli. The results demonstrate that stimulus-pairing-induced changes in cortical whisker representation appeared with different time courses at different levels of cortical columnar information processing.

Animals↗
PubMed1996

Landmark stability: further studies pointing to a role in spatial learning.

Two experiments were conducted to investigate the possible role of landmark stability in spatial learning. Rats were trained to search in a large arena for food hidden at a consistent distance and direction from either a single radially symmetric landmark or an array of two landmarks. We varied the relative degree to which the landmark array and/or the cues of the training context predicted the location of food, without varying the conditional probability of food being available given either cue. Experiment 1 used vestibular disorientation to ensure control of search location by experimenter-controlled cues. The results showed that making either a single landmark or a cluster of two adjacent landmarks the sole spatial predictor of reward location reduced the accuracy of search compared to a condition where both the landmark array and context cues were reliable spatial predictors. Varying global landmark stability had no effect when training was conducted using an array of two landmarks located some distance from each other. Context cues, when tested alone, triggered very little searching in appropriate locations, and the absolute magnitude of control over search was insufficient to account for the superiority of stable landmarks. The better learning with a stable landmark, and the dependence of this effect on the geometrical arrangement of landmarks, points to the conditions of spatial learning involving additional principles to those of simple associative conditioning. Experiment 2 examined landmark stability using a single landmark and fixed directional cues in the absence of vestibular disorientation. This also revealed a relative advantage of landmark stability, but animals with a landmark that moved from trial to trial did show some evidence of learning. Context cues when tested alone had minimal influence. Parametric manipulation of landmark stability offers a novel way of influencing spatial learning and thus understanding better the process through which egocentric representations of perceived space are transformed into allocentric representations of the real world.

Animals↗
PubMed2003

Representation and discrimination on an artificial dimension.

How we represent stimuli that are drawn from either natural (e.g., hue) or artificial (e.g., morphed face) dimensions is an issue of great significance for human learning. In this paper we outline a model of human dimensional representation in conjunction with some supporting empirical evidence for transfer along a continuum in humans (following Lawrence, 1952) and the first recorded case of transfer after outcome reversal with human subjects (following Mackintosh & Little, 1970). Our results support an elemental representation for dimensional stimuli in conjunction with algorithms that modulate both the salience and the associability of those representations.

Adolescent↗
PubMed1998

Temporal and amplitude generalization in motor learning.

A fundamental feature of human motor control is the ability to vary effortlessly over a substantial range, both the duration and amplitude of our movements. We used a three-dimensional robotic interface, which generated novel velocity dependent forces on the hand, to investigate how adaptation to these altered dynamics experienced only for movements at one temporal rate and amplitude generalizes to movements made at a different rate or amplitude. After subjects had learned to make a single point-to-point movement in a novel velocity-dependent force field, we examined the generalization of this learning to movements of both half the duration or twice the amplitude. Such movements explore a state-space not experienced during learning-any changes in behavior are due to generalization of the learning, the form of which was used to probe the intrinsic constraints on the motor control process. The generalization was assessed by determining the force field in which subjects produced kinematically normal movements. We found substantial generalization of the motor learning to the new movements supporting a nonlocal representation of the control process. Of the fields tested, the form of the generalization was best characterized by linear extrapolation in a state-space representation of the controller. Such an intrinsic constraint on the motor control process can facilitate the scaling of natural movements.

Adult↗
PubMedDate not supplied

Nonlinear prediction of quantitative structure-activity relationships.

Predicting the log of the partition coefficient P is a long-standing benchmark problem in Quantitative Structure-Activity Relationships (QSAR). In this paper we show that a relatively simple molecular representation (using 14 variables) can be combined with leading edge machine learning algorithms to predict logP on new compounds more accurately than existing benchmark algorithms which use complex molecular representations.

Journal Article↗
PubMed2000

Representing and reasoning about protein families using generative and discriminative methods.

This work addresses the issues of data representation and incorporation of domain knowledge into the design of learning systems for reasoning about protein families. Given the limited expressive capacity of a particular method, a mixture of protein annotation and fold recognition experts, each implementing a different underlying representation, should provide a robust method for assigning sequences to families. These ideas are illustrated using two data-driven learning methods that make use of different prior information and employ independent, yet complementary, projections of a family: hidden Markov models (HMMs) based on a multiple sequence alignment and neural networks (NNs) based on global sequence descriptors of proteins. Examination of seven protein families indicates that combining a generative (HMM) and a discriminative (NN) method is better than either method on its own. Biologically, human 4-hydroxyphenylpyruvic acid dioxygenase, involved in tyrosinemia type 3, is predicted to be structurally and functionally related to the glyoxalase I family.

Amino Acid Sequence↗
PubMed2003

[Plasticity of motor maps in primates: recent advances and therapeutical perspectives].

In the past decade, there have been considerable advances in understanding the neuronal bases of sensory and motor map reorganisation in adults and it is now clear that cortical representations are not invariant and stable, but rather, are dynamic and can continuously be modified. In human subjects, substantial advances in this field have been possible because of the spectacular development of non-invasive imaging and brain stimulation techniques. This review addresses specific questions about the capacity of motor maps in adult primates, including man, to change in response to behaviourally relevant experiences or as a result of central or peripheral lesion. The first part of this review deals with recent progress in understanding the role of the primary motor cortex (M1) in both motor control and cognition. The organisation and function of multiple "non-primary" motor areas located rostrally to the primary motor cortex and in the cingulate cortex are also discussed. This review then focuses on advances made in understanding motor cortex plasticity in different conditions. Firstly, since representations in M1 have been shown to change after motor learning, the contribution of M1 in motor learning has been insinuated; arguments against and in favour of this view are discussed. In addition, data suggesting that intracortical circuitry of M1 may play a role in map reorganisation following motor learning are also evaluated. Secondly, a large body of evidence from both animal and human observations is reviewed that confirms that M1 representations can also be altered as a result of changes in availability of effectors or following sensory deprivation. The mechanisms underlying such a plasticity of cortical maps following peripheral lesions are increasingly well understood. Thirdly, we discuss data showing that a corticospinal system lesion can lead to a complete reorganisation of the area allocated to the hand representation in the primary motor cortex or to a reorganization of the whole network of motor areas responsible for voluntary movements. As a conclusion, therapeutical perspectives that result from a better understanding of those various mechanisms responsible for motor map plasticity are briefly discussed.

Animals↗