SEARCH · PubMed Health

Results for “Contrastive learning”

Explore indexed PubMed citations for clinical trials, systematic reviews and public health research. Read source abstracts and follow each citation to its original PubMed record.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 1,531 records · Page 85Linked to original sources

PubMed2005

Preexisting knowledge versus on-line learning: what do young infants really know about spatial location?

Contemporary knowledge of infant cognition relies heavily on violation-of-expectation experiments. However, there are two ways to conceptualize what occurs in such studies. Babies may react to anomalous test events because of preexisting world knowledge. Alternatively, they may react because they have learned about events during the familiarization period. One way to distinguish these possibilities is to contrast familiarization with everyday versus anomalous events. In the studies we report here, we used this method to probe the nature of 5-month-olds' expectations about the locations of objects hidden in sand and later revealed. In Experiment 1, infants who initially saw everyday events did react to anomalous ones, as found previously, whereas infants who initially saw anomalous events did not react to everyday events. In Experiment 2, two alternative explanations of this pattern were ruled out. We conclude that by the age of 5 months, infants have expectations regarding the location of objects in continuous space.

Attention↗

PubMed2003

Practice makes perfect: the critical role of mixed practice in the acquisition of ECG interpretation skills.

PURPOSE: To examine the effect of instructional format on medical students' learning of ECG diagnosis. METHOD: Two experiments employed different learning and practice methods. In the first, students were randomly allocated to one of two instructional approaches, one organized around features (e.g., QRS voltage) and the other around diagnostic categories (e.g., bundle branch blocks), followed by a practice phase. In the second experiment, the instruction was standardized, and students were randomly allocated to one of two practice phases, either "contrastive" where examples from various categories are mixed together, or "non-contrastive" where all the examples in a single category are practiced in a single block. RESULTS: In the first experiment, there was no significant differences in students' diagnostic accuracy on novel ECG examples. In the second experiment, students exposed to the contrastive approach in the practice phase had superior diagnostic accuracy (46%) compared to 30% accuracy for the non-contrastive session, p < 0.05). CONCLUSION: These experiments highlight two important features in the design of instructional materials. First, learning around the features of the problem (analogous to problem-based learning) may have no advantages over learning the category. Second, the design and organization of deliberate practice can result in significant learning gain.

Canada↗

PubMed2003

Learning to control brain rhythms: making a brain-computer interface possible.

The ability to control electroencephalographic rhythms and to map those changes to the actuation of mechanical devices provides the basis for an assistive brain-computer interface (BCI). In this study, we investigate the ability of subjects to manipulate the sensorimotor mu rhythm (8-12-Hz oscillations recorded over the motor cortex) in the context of a rich visual representation of the feedback signal. Four subjects were trained for approximately 10 h over the course of five weeks to produce similar or differential mu activity over the two hemispheres in order to control left or right movement in a three-dimensional video game. Analysis of the data showed a steep learning curve for producing differential mu activity during the first six training sessions and leveling off during the final four sessions. In contrast, similar mu activity was easily obtained and maintained throughout all the training sessions. The results suggest that an intentional BCI based on a binary signal is possible. During a realistic, interactive, and motivationally engaging task, subjects learned to control levels of mu activity faster when it involves similar activity in both hemispheres. This suggests that while individual control of each hemisphere is possible, it requires more learning time.

Adaptation, Psychological↗

PubMed2007

Impaired spatial learning in the APPSwe + PSEN1DeltaE9 bigenic mouse model of Alzheimer's disease.

Mice co-expressing the Swedish amyloid precursor protein mutation (APP(Swe)) and exon 9 deletion (DeltaE9) of the PSEN1 gene begin to develop amyloid plaques at 6-7 months of age. We demonstrate here a spatial learning deficit in 7-month-old APP(Swe) + PSEN1DeltaE9 bigenic mice using an adaptation of the Barnes maze. Mice were first trained on a cued target followed by a hidden-target condition. Although bigenic mice quickly learned the cued-target version of the task, they were significantly impaired when switched to the hidden-target version. In contrast, a separate group of double-transgenic mice trained first on the spatial hidden-target version of the task were unimpaired relative to wild-type controls. We propose that processes such as general rule learning, context learning and exploratory habituation exert a greater influence when the testing environment is novel and overshadow the spatial memory deficit in naive bigenic mice. However, when cued-target training is conducted first, these processes habituate and the spatial learning deficit is unmasked. Seven-month-old APP(Swe) + PSEN1DeltaE9 mice were unimpaired on tests of memory that did not involve learning the rules governing spatial associations.

Alzheimer Disease↗

PubMed2004

Long-term potentiation in an avian basal ganglia nucleus essential for vocal learning.

Vocal learning in songbirds provides an excellent model for sensorimotor learning in vertebrates, with an accessible, well-defined behavior and discrete neural substrate. The rich behavioral plasticity exhibited by songbirds, however, contrasts starkly with the scarcity of candidate cellular mechanisms. Here, we report for the first time on an activity-dependent form of synaptic plasticity in area X, a component of the song system required for song learning and song maintenance. In slice preparations of zebra finch area X, pairing of high-frequency presynaptic stimulation with postsynaptic depolarization induces Hebbian long-term potentiation (LTP) of the glutamatergic inputs to spiny neurons. This form of LTP requires activation of NMDA receptors and D1-like dopamine receptors. In addition, LTP is observed in birds as young as 47 d after hatching and also in adult birds but not in younger birds, providing evidence of developmental regulation of the onset of synaptic plasticity. These properties make this form of LTP the best known candidate mechanism for reinforcement-based vocal learning in juveniles and song maintenance in adult birds.

Animals↗

PubMed2002

Dose-dependent effects of prenatal ethanol exposure on synaptic plasticity and learning in mature offspring.

BACKGROUND: We have observed profound deficits in hippocampal synaptic plasticity and one-trial learning in offspring whose mothers drank moderate quantities of ethanol during pregnancy. In the present study, we examined the question of whether lower maternal blood ethanol concentrations (BECs) could produce functional deficits in offspring. METHODS: Rat dams consumed either a 2%, 3%, or 5% ethanol liquid diet throughout gestation. Three other groups of dams were pair-fed a 0% ethanol liquid diet, and a seventh group consumed lab chow ad libitum. Adult offspring from each diet group were assigned either to studies of evoked [3H]-D-aspartate (D-ASP) release from hippocampal slices or spatial learning studies using the Morris Water Task. RESULTS: Consumption of the 2%, 3%, and 5% ethanol liquid diets produced mean peak maternal BECs of 7, 30 and 83 mg/dL, respectively. Consumption of these ethanol diets had no effect on offspring birthweight, litter size or neonatal mortality. Likewise, evoked D-ASP release from hippocampal slices and performance on a standard version of the Morris Water Task were not affected by prenatal ethanol exposure. By contrast, activity-dependent potentiation of evoked D-ASP release from slices and one-trial learning on a "moving platform" version of the Morris Water Task were markedly reduced in offspring whose mothers consumed the 5% ethanol liquid diet. Intermediate deficits in these two parameters were observed in offspring from the 3% ethanol diet group, whereas offspring from the 2% ethanol diet group were not statistically different than controls. CONCLUSIONS: We conclude that the threshold for eliciting subtle, yet significant learning deficits in offspring prenatally exposed to ethanol is less than 30 mg/dL. This BEC is roughly equivalent to drinking 1 to 1.5 ounces of ethanol per day.

Age Factors↗

PubMed2005

Appearance of amyloid beta-like substances and delayed-type apoptosis in rat hippocampus CA1 region through aging and oxidative stress.

To elucidate whether oxidative stress induces cognitive deficit, and whether nerve cells in the hippocampus, which modulates learning and memory functions in the brain, are damaged by oxidative stress and during aging, the influence of hyperoxia as oxidative stress on either the cognitive function of rats or the oxidative damage of nerve cells was investigated. Young rats showed better learning ability than both old rats and vitamin E-deficient young rats. Vitamin E- supplemented young rats showed similar ability to young control rats. After they learned the location of the platform in the Morris water maze test, the young rats and vitamin E-supplemented young rats were subjected to oxidative stress for 48 h, and the old rats and vitamin E-deficient young rats were kept in normal atmosphere. The memory function of the old rats and vitamin E-deficient young rats declined even when they were not subjected to oxidative stress for 48 h. In contrast, the young rats maintained their memory function for 4 days after the oxidative stress. However, their learning abilities suddenly declined toward that of the normal old rats after 5 days. At this point, nerve cell loss and apoptosis were observed in the hippocampal CA 1 region of young rats. Vitamin E-supplementation in the young rats prevented either memory deficit or the induction of delayed-type apoptosis. The old rats and vitamin E-deficient young rats kept in normal atmosphere for 48 h also showed apoptosis in the hippocampus. Also, 10 days after oxidative stress, amyloid beta-like substances appeared in the CA-1 region of control young rats; these substances were also observed in the CA-1 region of the old rats and vitamin E- deficient young rats. These results suggest that reactive oxygen species (ROS) generated by oxidative stress induced amyloid beta-like substances and delayed-type apoptosis in the rat hippocampus, resulting in cognitive deficit. Since amyloid beta in Alzheimer's disease characterized by cognitive deficit induces neuronal cell death, it is reasonable to consider that amyloid beta deposition in the brain may be associated with memory dysfunction. The results of this study imply that age-related hippocampal neuronal damage is prevented by vitamin E supplementation due to the antioxidant effect of vitamin E.

Aging↗

PubMed2004

Memory for objects and their locations: the role of the hippocampus in retention of object-place associations.

Computational models of hippocampal function have suggested that the hippocampus is involved in the formation and storage of arbitrary associations. Previous studies have shown that rats with hippocampal lesions are impaired in object-place associative learning. However, few studies have examined the role of the hippocampus in the retention of previously learned arbitrary associations. In the present study, male Long-Evans rats with either cortical control or hippocampal lesions were tested on a task measuring the retention of previously learned arbitrary associations using an object-place paired-associate task. To assess retention, each animal was trained on the paired-associate task for 360 trials, then received a lesion, and was retested to examine retention of the previously learned associations. The results indicate that all rats learned the task prior to surgery. Following surgery, rats with cortical control lesions were not impaired in the retention of object-place associations. In contrast, hippocampal lesions resulted in an initial deficit in retention of the paired-associate task followed by recovery. Therefore, the hippocampus may play a role in the retrieval of previously learned arbitrary association.

Animals↗

PubMed1989

A characteristic pattern of active avoidance behavior in SART-stressed rats.

In SART-stressed (repeated cold-stressed) rats, shuttle-avoidance response was examined. The rats exposed to SART stress prior to training showed a high avoidance rate and no change in the intertrial response. Upon exposure to SART stress after completion of learning, the rats showed no changes in the avoidance rate and an increase in the intertrial response in the retention test. These results are in contrast to the previous observations of passive avoidance response, and the abnormal behavior in such animals may be based on excessive emotionality and/or hyperreactivity rather than alterations in the process of learning and memory.

Animals↗

PubMed2006

High levels of new neuron addition persist when the sensitive period for song learning is experimentally prolonged.

Socially reared zebra finch males imitate a song they hear during posthatching days 30-65; during this time, many new neurons are added to the high vocal center (HVC), a forebrain nucleus necessary for the production of learned song. New neuron addition drops sharply after day 65, and no new songs are imitated. In contrast, male zebra finches reared in isolation from other males have more variable songs at day 65 and thereafter can still imitate new sounds (Eales, 1985). We show that, in isolate birds, a greater number of new neurons continues to be added to HVC during the next 85 d, and this number correlates with syllable variability. We suggest that new neuron addition and turnover facilitate song change and that this effect lingers when an expected learning event is delayed.

Animals↗

PubMed1981

Modification of motor activity, passive avoidance conditioning and evoked potentials by microinjections of strychnine in both caudate nuclei in cats.

Passive avoidance conditioning (PAC) seems to depend on inhibitory actions within the caudate nucleus (CN). Thus topical application of strychnine could block these influences. In cats, with permanently implanted cannulae in the head of the CN, bilateral microinjections of 100 micrograms of strychnine produced a stereotyped behavior of grooming and running with the extremities flexed to dark places. When placed into a two compartment chamber for a PAC acquisition they would not move from the dark compartment to the illuminated one. After decreasing the illumination some cats crossed and the acquisition test was performed. Twenty four hours later they showed the learned response. A second retention test was not modified by another strychnine application. These results contrast with those obtained by picrotoxin microinjections which disrupt the retention of a PAC. However, strychnine affects the evoked potentials (EP) recorded in CN by n. centralis medialis (NCM) stimulation by decreasing a positive peak, which probably represents post-synaptic inhibition and increasing the late positive component. In the lateral geniculate body strychnine affected a late positive wave and in the occipital cortex it increased the initial negative peaks and decreased the late positive ones of the potentials evoked by flash stimulation. These findings suggest a more suitable role of the CN in the regulation of visual information, which is probably related with the attention processes.

Animals↗

PubMed1996

Generalization of form in visual pattern classification.

Human observers were trained to criterion in classifying compound Gabor signals with symmetry relationships, and were then tested with each of 18 blob-only versions of the learning set. Generalization to dark-only and light-only blob versions of the learning signals, as well as to dark-and-light blob versions was found to be excellent, thus implying virtually perfect generalization of the ability to classify mirror-image signals. The hypothesis that the learning signals are internally represented in terms of a 'blob code' with explicit labelling of contrast polarities was tested by predicting observed generalization behaviour in terms of various types of signal representations (pixelwise, Laplacian pyramid, curvature pyramid, ON/OFF, local maxima of Laplacian and curvature operators) and a minimum-distance rule. Most representations could explain generalization for dark-only and light-only blob patterns but not for the high-thresholded versions thereof. This led to the proposal of a structure-oriented blob-code. Whether such a code could be used in conjunction with simple classifiers or should be transformed into a propositional scheme of representation operated upon by a rule-based classification process remains an open question.

Adult↗

PubMed2002

Thought before language: how deaf and hearing children express motion events across cultures.

Do children come to the language-learning situation with a predetermined set of ideas about motion events that they want to communicate? If so, is the expression of these ideas modified by exposure to a language model within a particular cultural context? We explored these questions by comparing the gestures produced by Chinese and American deaf children who had not been exposed to a usable conventional language model with the speech of hearing children learning Mandarin or English. We found that, even in the absence of any conventional language model, deaf children conveyed the central elements of a motion event in their communications. More surprisingly, deaf children growing up in an American culture used their gestures to express motion events in precisely the same ways as deaf children growing up in a Chinese culture. In contrast, hearing children in the two cultures expressed motion events differently, in accordance with the languages they were learning. The American children obeyed the patterns of English and rarely omitted words for figures or agents. The Chinese children had more flexibility as Mandarin permits (but does not demand) deletion. Interestingly, the Chinese hearing children's descriptions of motion events resembled the deaf children's descriptions more closely than did the American hearing children's. The thoughts that deaf children convey in their gestures thus may serve as the starting point and perhaps a default for all children as they begin the process of grammaticization--thoughts that have not yet been filtered through a language model.

Child, Preschool↗

PubMed2004

Reparative mechanisms in the cerebellar cortex.

In the adult brain, different neuronal populations display different degrees of plasticity. Here, we describe the highly different plastic properties of inferior olivary neurones and Purkinje cells. Olivary neurones show a basal expression of growth-associated proteins, such as GAP-43 and Krox24/EGR-1, and remarkable remodelling capabilities of their terminal arbour. They also regenerate their transected neurites into growth-permissive territories and may reinnervate the lost target. Sprouting and regrowing olivary axons are able to follow specific positional information cues to establish new connections according to the original projection map. In addition, they set a strong cell body reaction to injury, which in specific olivary subsets is regulated by inhibitory target-derived cues. In contrast, Purkinje cells do not have a constitutive level of growth-associated genes, and show little cell body reaction, no axonal regeneration after axotomy, and weak sprouting capabilities. Block of myelin-derived signals allows terminal arbour remodelling, but not regeneration, while selective over-expression of GAP-43 induces axonal sprouting along the axonal surface and at the level of the lesion. We suggest that the high constitutive intrinsic plasticity of the inferior olive neurones allows their terminal arbour to sustain the activity-dependent ongoing competition with the parallel fibres in order to maintain the post-synaptic territory, and possibly underlies mechanisms of learning and memory. Such a plasticity is used also as a reparative mechanism following axotomy. In contrast, in Purkinje cells, poor intrinsic regenerative capabilities and myelin-derived signals stabilise the mature connectivity and prevent axonal regeneration after lesion.

Animals↗

PubMed1997

Effect of age on pentylenetetrazol-kindling and kindling-induced impairments of learning performance.

Epileptogenesis during ontogeny may not be linearly related to time. It is known that the behavioral manifestations of seizures are age-dependent, but more research is needed to clarify ontogenetic aspects of epilepsies and related alterations, including cognitive deficits. Kindling is an accepted animal model for the study of the convulsive component of epilepsy and its consequences on behavior. Recently, we demonstrated an impairment in acquisition of a conditioned reaction in young adult kindled rats, using pentylenetetrazol (PTZ) as the kindling stimulus. The present study was undertaken to investigate the dependence on age of alterations in the induction of PTZ kindling in rats. We started the kindling protocol in 4-, 6-, and 8-week- and 6-, 12-, 18-, and 24-month-old rats. The PTZ kindling showed an age-dependent decrease in expression of convulsions. The diminished kindling capacity was already seen in 6-month-old rats. In contrast, kindling-related impairment effects on cognitive functions increased with age. Thus, the correlation between learning impairment and occurrence of tonic-clonic seizures that we had demonstrated in 8-week-old rats was abolished in older rats. On the other hand, when the kindling procedure was started in 6-week-old rats, no impairment was found in fully kindled rats.

Aging↗

PubMed2026

Deep learning-based cross-attention fusion of multimodal MRI for survival prediction and risk stratification in IDH-wildtype glioblastoma: a multicenter study.

BACKGROUND: Glioblastoma (GBM) exhibits profound molecular and spatial heterogeneity, complicating prognostic evaluations. While multiparametric MRI provides crucial multidimensional biological information, conventional end-to-end deep learning integration strategies, such as early or late fusion, often fail to capture complex nonlinear cross-modal interactions. We aimed to systematically evaluate a cross-attention fusion (CAF) architecture for GBM survival prediction and quantify its incremental prognostic value relative to existing clinical tools. METHODS: In this multicenter retrospective study, 386 adults with IDH-wildtype, WHO grade 4 GBM were assembled from an institutional cohort (n = 226), the Chinese Glioma Genome Atlas (CGGA, n = 62), and The Cancer Genome Atlas (TCGA, n = 98). Using a unified 3D ResNet-18 backbone, we compared single-modality models, early fusion, late fusion, and CAF on preoperative T1-weighted, contrast-enhanced T1-weighted (T1CE), and T2-weighted MRI, and integrated the resulting deep learning risk score with routine clinical variables through multivariable Cox regression. Performance was assessed using Harrell's C-index, time-dependent AUC, and decision curve analysis. RESULTS: CAF showed numerically higher, more consistent C-index trends than early fusion, late fusion, and single-modality models (pooled C-index 0.629, 95% CI 0.594-0.664), although pairwise differences in time-dependent AUC were not statistically significant. Integrating clinical variables raised the pooled C-index to 0.691 (95% CI 0.660-0.721) in the treatment-era model, with comparable performance across the three cohorts (Local 0.688; CGGA 0.716; TCGA 0.689); a pre-treatment configuration excluding adjuvant therapy yielded a pooled C-index of 0.642. Under leave-one-cohort-out external validation, the combined model retained significant risk stratification in all held-out cohorts (C-index 0.63-0.71; all log-rank P < 0.01), albeit with attenuated discrimination. The deep learning risk score remained independent after multivariable adjustment (HR 1.41 per SD, 95% CI 1.26-1.57; P < 0.001). Kaplan-Meier analysis confirmed significant high- versus low-risk separation in all cohorts, and decision curve analysis showed greater net benefit than clinical-only and deep-learning-only models. CONCLUSION: The CAF-derived risk score offers prognostic information complementary to routine clinical variables, representing a promising noninvasive tool for individualized risk stratification when molecular profiling is incomplete or unavailable; these findings warrant prospective external validation before clinical use.

cross-attention fusion↗

PubMed1989

Rules and exemplars in categorization, identification, and recognition.

Subjects learned to classify perceptual stimuli varying along continuous, separable dimensions into rule-described categories. The categories were designed to contrast the predictions of a selective-attention exemplar model and a simple rule-based model formalizing an economy-of-description view. Converging evidence about categorization strategies was obtained by also collecting identification and recognition data and by manipulating strategies via instructions. In free-strategy conditions, the exemplar model generally provided an accurate quantitative account of identification, categorization, and recognition performance, and it allowed for the interrelationship of these paradigms within a unified framework. Analyses of individual subject data also provided some evidence for the use of rules, but in general, the rules seemed to have a great deal in common with exemplar storage processes. Classification and recognition performance for subjects given explicit instructions to use specific rules contrasted dramatically with performance in the free-strategy conditions and could not be predicted by the exemplar model.

Adult↗

PubMedDate not supplied

An integrated model of meaning and sense activation and disambiguation.

A simulation of meaning and sense selection is presented. In the model, distributed representations are used for both unrelated meanings and related meaning senses; however, related meaning senses are assumed to be correlated. Related meaning senses are assumed to be learned associatively in similarly correlated contexts. Our model is sufficient to account for the contrasting results obtained in priming paradigms by Whitney, McKay, Kellas, and Emerson (1985) and Tabossi (1988).

Cognition↗