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 397 records · Page 22Linked to original sources
PubMed1989

Aiming error under transformed spatial mappings suggests a structure for visual-motor maps.

Transformed spatial mappings were used to perturb normal visual-motor processes and reveal the structure of internal spatial representations used by the motor control system. In a 2-D discrete aiming task performed under rotated visual-motor mappings, the pattern of spatial movement error was the same for all Ss: peak error between 90 degrees and 135 degrees of rotation and low error for 180 degrees rotation. A two-component spatial representation, based on oriented bidirectional movement axes plus direction of travel along such axes, is hypothesized. Observed reversals of movement direction under rotations greater than 90 degrees are consistent with the hypothesized structure. Aiming error under reflections, unlike rotations, depended on direction of movement relative to the axis of reflection (see Cunningham & Pavel, in press). Reaction time and movement time effects were observed, but a speed-accuracy tradeoff was found only for rotations for which the direction-reversal strategy could be used. Finally, adaptation to rotation operates at all target locations equally but does not alter the relative difficulty of different rotations. Structural properties of the representation are invariant under learning.

Adult↗
PubMed2000

Representation of actions in rats: the role of cerebellum in learning spatial performances by observation.

Experimental evidence demonstrates that cerebellar networks are involved in spatial learning, controlling the acquisition of exploration strategies without blocking motor execution of the task. Action learning by observation has been considered somehow related to motor physiology, because it provides a way of learning performances that is almost as effective as the actual execution of actions. Neuroimaging studies demonstrate that observation of movements performed by others, imagination of actions, and actual execution of motor performances share common neural substrates and that the cerebellum is among these shared areas. The present paper analyzes the effects of observation in learning a spatial task, focusing on the cerebellar role in learning a spatial ability through observation. We allowed normal rats to observe 200 Morris water maze trials performed by companion rats. After this observation training, "observer" rats underwent a hemicerebellectomy and then were tested in the Morris water maze. In spite of the cerebellar lesion, they displayed no spatial defects, exhibiting exploration abilities comparable to controls. When the cerebellar lesion preceded observation training, a complete lack of spatial observational learning was observed. Thus, as demonstrated already for the acquisition of spatial procedures through actual execution, cerebellar circuits appear to play a key role in the acquisition of spatial procedures also through observation. In conclusion, the present results provide strong support for a common neural basis in the observation of actions that are to be reproduced as well as in the actual production of the same actions.

Animals↗
PubMed1996

Tactile agnosia. Casuistic evidence and theoretical remarks on modality-specific meaning representations and sensorimotor integration.

Somaesthetic, motor and cognitive functions were studied in a man with impaired tactile object-recognition (TOR) in his left hand due to a right parietal convexity meningeoma which had been surgically removed. Primary motor and somatosensory functions were not impaired, and discriminative abilities for various tactile aspects and cognitive skills were preserved. Nevertheless, the patient could often not appreciate the object's nature or significance when it was placed in his left hand and was unable to name or to describe or demonstrate the use of these objects. Therefore, he can be regarded as an example of associative tactile agnosia. The view is taken and elaborated that defective modality-specific meaning representations account for associative tactile agnosia. These meaning representations are conceptualized as learned unimodal feature-entity relationships which are thought to be defective in tactile agnosia. In line with this hypothesis, tactile feature analysis and cross-modal matching of features were largely preserved in the investigated patient, while combining features to form entities was defective in the tactile domain. The alternative hypothesis of agnosia as deficit of cross-modal association of features was not supported. The presumed distributed functional network responsible for TOR is thought to involve perception of features, object recognition and related tactile motor behaviour interactively. A deficit leading primarily to impaired combining features to form entities can therefore be expected to result in additional minor impairment of related perceptual-motor processes. Unilaterality of the gnostic deficit can be explained by a lateralized organization of the functional network responsible for tactile recognition of objects.

Agnosia↗
PubMedDate not supplied

Role of parietal cortex and hippocampus in representing spatial information.

Rats with lesions in the parietal cortex or hippocampus as well as cortical lesion and sham-operated controls were tested for acquisition or retention of a cheese board spatial task (dry land version of a water maze task). Results indicated that, relative to controls, rats with hippocampal or parietal cortex lesions were impaired in both acquisition and retention of the spatial task as measured by increased distances traveled to find the correct food location. It is suggested that both the hippocampus and parietal cortex subserve spatial representations required for optimal learning and performance of the cheese board spatial task.

Animals↗
PubMed2000

LTP in the rat basal amygdala induced by perirhinal cortex stimulation in vivo.

The present study examined the effects of ventral perirhinal cortex (vPRC) stimulation on evoked field potentials (EPs) in the basal amygdala (BN), using extracellular recording techniques. Single pulse stimulation of the vPRC reliably evoked a negative field potential in the BN that was missing in the lateral nucleus. Paired-pulses delivered to the vPRC induced a short-lasting facilitation at intervals between 15 and 120 ms. Application of brief theta burst stimulation to the vPRC produced an enduring long-term potentiation (LTP) that reached 150% of control values. The induction of LTP was not accompanied by a decrease in paired-pulse facilitation. These results suggest that emotional learning involving complex stimulus representations may be mediated by cortico-amygdalar projections amenable to LTP.

Amygdala↗
PubMed1993

Molecular scene analysis: the integration of direct-methods and artificial-intelligence strategies for solving protein crystal structure.

A knowledge-based approach to crystal structure determination is presented. The approach integrates direct-methods and artificial-intelligence strategies to rephrase the structure determination process as an exercise in scene analysis. A general joint probability distribution framework, which allows the incorporation of isomorphous replacement, anomalous scattering and a priori structural information, forms the basis of the direct-methods strategies. The accumulated knowledge on crystal and molecular structures is exploited through the use of artificial-intelligence strategies, which include techniques of knowledge representation, search and machine learning.

Journal Article↗
PubMed2005

Kernel pooled local subspaces for classification.

We investigate the use of subspace analysis methods for learning low-dimensional representations for classification. We propose a kernel-pooled local discriminant subspace method and compare it against competing techniques: kernel principal component analysis (KPCA) and generalized discriminant analysis (GDA) in classification problems. We evaluate the classification performance of the nearest-neighbor rule with each subspace representation. The experimental results using several data sets demonstrate the effectiveness and performance superiority of the kernel-pooled subspace method over competing methods such as KPCA and GDA in some classification problems.

Algorithms↗
PubMed2005

Major dissociation between medial and lateral entorhinal input to dorsal hippocampus.

Hippocampal place cells are a model system of how the brain constructs cognitive representations and of how these representations support complex behavior, learning, and memory. There is, however, a lack of detailed knowledge about the properties of hippocampal afferents. We recorded multiple single units from the hippocampus and the medial and lateral entorhinal areas of behaving rats. Although many medial entorhinal neurons had highly specific place fields, lateral entorhinal neurons displayed weak spatial specificity. This finding demonstrates a fundamental dissociation between the information conveyed to the hippocampus by its major input streams, with spatial information represented by the medial and nonspatial information represented by the lateral entorhinal cortex.

Animals↗
PubMedDate not supplied

Background of the demonstrated IMAGIS activities and future expectations.

In The Netherlands a national PACS development programme has been started, supported by the Dutch Society of Radiology and funded by the Dutch Department of Health because of the national character of the project. Three main partners are cooperating in this development: the Utrecht University Hospital (AZU), BAZIS and Philips International (Product Division Medical Systems), with the Delft University of Technology as the main BAZIS subcontractor. The non-profit foundation BAZIS, developing and supporting the ZIS Hospital Information System (in use in some 30 Dutch hospitals, over 15,000 acute beds), initiated its current IMAGe Information System (IMAGIS) projects in 1984. The activities were later integrated into the Dutch PACS project started in 1986. The final goal of the project is to achieve a PACS which is fully integrated with already existing hospital information systems (HIS). The development and operation of a HIS-PACS include many aspects of technical and clinical. The current efforts of BAZIS are concentrated on three main issues: diagnostic image quality evaluation (e.g. effects of data compression); modelling, software simulation and technology assessment of a prototype PACS (both general and detailed aspects); and coupling and integration of PACS and HIS (e.g. the BAZIS ZIS). Philips, Hamburg, is supplying equipment, particularly prototype components. A systematic clinical evaluation will take place at the Utrecht University Hospital.2+ We outline the background of the intermediate results as demonstrated during the 6th EuroPACS Conference:the psychophysical software package for Feature Evaluation And System Inspection By Logged Experiments (FEASIBLE); the modelling and simulation software package for Medical Image Representation, Archiving and Communication, Learned by Extensive Simulation (MIRACLES); and first results of the coupling experiments.(ABSTRACT TRUNCATED AT 250 WORDS)

Computer Communication Networks↗
PubMedDate not supplied

Prediction and analysis of PACS performance with the simulation tool MIRACLES.

Since the construction of image information systems appears to be extremely difficult in practice, BAZIS has decided to use computer modelling and simulation as decision support tools. In order to support the construction of simulation models, the simulation package and modelling environment MIRACLES (Medical Image Representation, Archiving and Communication Learned from Extensive Simulation) has been developed by BAZIS. This paper describes modelling and simulation techniques in general, as well as the benefits of simulation within the scope of designing Picture Archiving and Communication Systems (PACS). In order to illustrate the theory, results of a concrete yet simple PACS, which has been simulated with MIRACLES, will be described and discussed.

Computer Communication Networks↗
PubMed2001

[Study of the pulmonary heart disease computer-aided diagnosis system based on combining neural network].

We have constructed an expert sub-system for diagnosing pulmonary heart disease with back-propagation (BP) model of typical artificial neural network and BP-Hamming. The system can obtain and represent clinical data, and it has solved the problems such as obtaining automatic knowledge, representation and self-learning. The results of testing demonstrate that the method of neural network can be regarded as an expert sub-system to make intelligent diagnosis of pulmonary heart disease.

Algorithms↗
PubMed2005

Learning invariant object recognition in the visual system with continuous transformations.

The cerebral cortex utilizes spatiotemporal continuity in the world to help build invariant representations. In vision, these might be representations of objects. The temporal continuity typical of objects has been used in an associative learning rule with a short-term memory trace to help build invariant object representations. In this paper, we show that spatial continuity can also provide a basis for helping a system to self-organize invariant representations. We introduce a new learning paradigm "continuous transformation learning" which operates by mapping spatially similar input patterns to the same postsynaptic neurons in a competitive learning system. As the inputs move through the space of possible continuous transforms (e.g. translation, rotation, etc.), the active synapses are modified onto the set of postsynaptic neurons. Because other transforms of the same stimulus overlap with previously learned exemplars, a common set of postsynaptic neurons is activated by the new transforms, and learning of the new active inputs onto the same postsynaptic neurons is facilitated. We demonstrate that a hierarchical model of cortical processing in the ventral visual system can be trained with continuous transform learning, and highlight differences in the learning of invariant representations to those achieved by trace learning.

Computer Simulation↗
PubMed2000

Learning curve of microvascular venous anastomosis: a never ending struggle?

In this study, a simple protocol based on the rat femoral venous anastomosis was established to provide a quantitative representation of the progress. The learning curve is based on the patency rate in each consecutive group of five anastomoses. Two groups of surgeons were observed. The inexperienced group encountered a tough time in the first 25 anastomoses. However, the progress was fast and is represented by the steep slope of the curve. A plateau was reached whereby the avearge patency rate matches that of the experienced group. As expected, there was no learning curve for the experienced group. Despite every effort to attempt to maintain a perfect 100% patency on this model, the best achievable patency was only 88%. The results and its implication are discussed.

Anastomosis, Surgical↗
PubMed2005

The primate amygdala: Neuronal representations of the viscosity, fat texture, temperature, grittiness and taste of foods.

The primate amygdala is implicated in the control of behavioral responses to foods and in stimulus-reinforcement learning, but only its taste representation of oral stimuli has been investigated previously. Of 1416 macaque amygdala neurons recorded, 44 (3.1%) responded to oral stimuli. Of the 44 orally responsive neurons, 17 (39%) represent the viscosity of oral stimuli, tested using carboxymethyl-cellulose in the range 1-10,000 cP. Two neurons (5%) responded to fat in the mouth by encoding its texture (shown by the responses of these neurons to a range of fats, and also to non-fat oils such as silicone oil ((Si(CH(3))(2)O)(n)) and mineral oil (pure hydrocarbon), but no or small responses to the cellulose viscosity series or to the fatty acids linoleic acid and lauric acid). Of the 44 neurons, three (7%) responded to gritty texture (produced by microspheres suspended in cellulose). Eighteen neurons (41%) responded to the temperature of liquid in the mouth. Some amygdala neurons responded to capsaicin, and some to fatty acids (but not to fats in the mouth). Some amygdala neurons respond to taste, texture and temperature unimodally, but others combine these inputs. These results provide fundamental evidence about the information channels used to represent the texture and flavor of food in a part of the brain important in appetitive responses to food and in learning associations to reinforcing oral stimuli, and are relevant to understanding the physiological and pathophysiological processes related to food intake, food selection, and the effects of variety of food texture in combination with taste and other inputs on food intake.

Action Potentials↗
PubMed2005

Effects of an ontology display with history representation on organizational memory information systems.

Ontologies, as a possible element of organizational memory information systems, appear to support organizational learning. Ontology tools can be used to share knowledge among the members of an organization. However, current ontology-viewing user interfaces of ontology tools do not fully support organizational learning, because most of them lack proper history representation in their display. In this study, a conceptual model was developed that emphasized the role of ontology in the organizational learning cycle and explored the integration of history representation in the ontology display. Based on the experimental results from a split-plot design with 30 participants, two conclusions were derived: first, appropriately selected history representations in the ontology display help users to identify changes in the ontologies; and second, compatibility between types of ontology display and history representation is more important than ontology display and history representation in themselves.

Female↗
PubMed2004

From parallel sequence representations to calligraphic control: a conspiracy of neural circuits.

Calligraphic writing presents many challenges for motor control, including: learning and recall of stroke sequences; critical timing of stroke onsets and durations; fine control of grip and contact forces; and letterform invariance under size scaling, which entails fine control of stroke directions and amplitudes during recruitment and derecruitment of musculoskeletal degrees of freedom. Experimental and computational studies in behavioral neuroscience have progressed toward explaining the learning, planning, and control exercised in tasks that share features with calligraphic writing and drawing. This article highlights component operations ranging from parallel sequence representations to fine force control. Treated in succession are: competitive queuing models of sequence representation, performance, learning, and recall; letter size scaling and motor equivalence; cursive handwriting models in which sensory-motor transformations are performed by circuits that learn inverse differential kinematic mappings; and fine-grained control of timing and transient forces by circuit models that learn to solve inverse dynamics problems.

Biomechanical Phenomena↗
PubMed2002

An ultra-sparse code underlies the generation of neural sequences in a songbird.

Sequences of motor activity are encoded in many vertebrate brains by complex spatio-temporal patterns of neural activity; however, the neural circuit mechanisms underlying the generation of these pre-motor patterns are poorly understood. In songbirds, one prominent site of pre-motor activity is the forebrain robust nucleus of the archistriatum (RA), which generates stereotyped sequences of spike bursts during song and recapitulates these sequences during sleep. We show that the stereotyped sequences in RA are driven from nucleus HVC (high vocal centre), the principal pre-motor input to RA. Recordings of identified HVC neurons in sleeping and singing birds show that individual HVC neurons projecting onto RA neurons produce bursts sparsely, at a single, precise time during the RA sequence. These HVC neurons burst sequentially with respect to one another. We suggest that at each time in the RA sequence, the ensemble of active RA neurons is driven by a subpopulation of RA-projecting HVC neurons that is active only at that time. As a population, these HVC neurons may form an explicit representation of time in the sequence. Such a sparse representation, a temporal analogue of the 'grandmother cell' concept for object recognition, eliminates the problem of temporal interference during sequence generation and learning attributed to more distributed representations.

Action Potentials↗
PubMed1996

Suppression of synaptic transmission may allow combination of associative feedback and self-organizing feedforward connections in the neocortex.

Selective suppression of synaptic transmission during learning is proposed as a physiological mechanism for combining associative memory function at feedback synapses with self-organization of feedforward synapses in neocortical structures. A computational model demonstrates how selective suppression of feedback transmission allows this combination of synaptic function. During learning, sensory stimuli and the desired response are simultaneously presented as input to the network. Feedforward connections form self-organized representations of input, while suppressed feedback connections learn the transpose of the feedforward connectivity. During recall, suppression of transmission is removed, input activates the self-organized representation, and activity settles into a learned solution to the problem. This computational model can be used for learning of problems which are not linearly separable, including the negative patterning task (the XOR problem). Experiments in brain slice preparations of the rat somatosensory cortex tested whether the combination of self-organization and associative memory function could be provided by cholinergic suppression selective for feedback versus feedforward synapses. The cholinergic agonist carbachol selectively suppressed synaptic potentials elicited by stimulation of layer I (which contains a high percentage of feedback synapses), while having no effect on synaptic potentials elicited by stimulation of layer IV (with a high percentage of afferent and feedforward synapses).

Animals↗