PubMed Health⌕ Search

SEARCH · PubMed Health

Results for “Deep 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 379 records · Page 21Linked to original sources
PubMed2006

Reducing the dimensionality of data with neural networks.

High-dimensional data can be converted to low-dimensional codes by training a multilayer neural network with a small central layer to reconstruct high-dimensional input vectors. Gradient descent can be used for fine-tuning the weights in such "autoencoder" networks, but this works well only if the initial weights are close to a good solution. We describe an effective way of initializing the weights that allows deep autoencoder networks to learn low-dimensional codes that work much better than principal components analysis as a tool to reduce the dimensionality of data.

Journal Article↗
PubMed1990

Relationship between using conceptual comprehension of academic material and thinking abstractly about global life issues.

The study explored whether more frequent use of conceptual comprehension of academic material generalized to greater use of abstract thinking about global life issues, such as death, goal in life, marriage, AIDS, etc. Undergraduate and graduate students (28 men and 61 women) voluntarily completed a questionnaire which assessed their conceptualizations using three indices. These were an intelligence scale and two learning style indices, namely, Deep Processing and Elaborative Processing of R. R. Schmeck. Also assessed were their levels of abstract thinking about Death Issues and about Other Real Life Issues, and their Denial of Death and their Denial of Dying. All three indices of conceptualization correlated with thinking more abstractly about Other Real Life Issues, but only Elaborative Processing correlated with thinking more abstractly about Death Issues. None of the three indices correlated with Denial of Death or Denial of Dying. It appears conceptualization skills were selectively generalized.

Attitude to Death↗
PubMed1990

Nerve transposition for the restoration of elbow flexion following brachial plexus avulsion injuries.

Despite technical advances, the ability to restore motor function following a brachial plexus avulsion is limited. Twenty patients who suffered the loss of elbow flexion following a brachial plexus avulsion injury underwent a neurotization procedure in an attempt to restore that lost function. Of 16 patients who underwent intercostal to musculocutaneous nerve anastomosis, seven obtained good elbow flexion. Four patients who no longer had a viable biceps brachialis muscle underwent an anastomosis between transposed intercostal nerves and a free vascularized gracilis muscle grafted to the position of the biceps. Two of these patients obtained good elbow flexion. Although synkinesis between the biceps brachialis and the inspiratory muscles can be demonstrated during coughing and deep inspiration, the patients learn to flex their reinnervated biceps brachialis muscle and maintain flexion independent of respiration.

Adolescent↗
PubMed1997

[Comparative results of deep sclerectomy transformed to trabeculectomy and classical trabeculectomy].

OBJECTIVE: To study the learning risk of a new surgical technique, the deep sclerectomy (DS), which may be complicated with a perforation of the trabeculodescemetic membrane. To do so, we compared the results and the complications of patients who underwent DS transformed in trabeculectomy (DSt) and those of patients who underwent trabeculectomy (TE). MATERIALS AND METHODS: Between june 1994 and june 1996, we performed 234 DS, 19 of them had to be transformed into a TE, because of a perforation of the trabeculodescemetic membrane during the deep sclerocorneal dissection. Two matched groups of patients were prospectively followed, one group including 19 DSt and a control group including 19 TE. RESULTS: The IOP decrease on the first postoperative day was greater in the DSt-group compared to the TE-group (2.32 +/- 3.89 vs 6.73 +/- 4.31, p = 0.004). Scleral perforations occurred mainly during the learning phase of this new surgical technique. Postoperative complications were similar in both groups, except for hypotony and hyphema which were more frequent in the DSt-group (hypotony: 90% vs 37%, p = 0.0019; hyphema: 68% vs 16%, p = 0.017). The long term success rate with or without medication were comparable in both groups. CONCLUSION: This study shows that, when a DS is complicated with a perforation of the trabeculodescemetic membrane, the long term success rate of the surgery is similar to that of trabeculectomy. However, immediate postoperative complications such as hypotony and hyphema are increased. These results should encourage surgeons to learn this new non perforating filtration surgery.

Aged↗
PubMed1984

Effects of lesions of cerebellar nuclei on conditioned behavioral and hippocampal neuronal responses.

Rabbits were overtrained using classical conditioning of the rabbit nictitating membrane (NM)/eyelid response. Unilateral electrolytic lesions were then made through electrodes previously implanted in dentate and interpositus cerebellar nuclei ipsilateral to the trained (left) eye. Lesions caused a complete or near-complete abolition of conditioned behavioral responses on the ipsilateral side, but had no effect on unconditioned responses to corneal airpuff. When training was switched to the contralateral (right) side, animals learned within the first few trials, but did not relearn when training was returned to the left (lesioned) side. Control animals in which lesions spared the deep nuclei showed no such learning deficits. Lesions of cerebellar nuclei also abolished conditioned increases in hippocampal CAl neural activity evoked by the tone conditioned stimulus in this paradigm. As with the behavior, training on the right (non-lesioned) side reinstated the conditioned neuronal response within the first few trials of training, even though behavioral responding on lesioned side showed little or no improvement. These results indicate that the cerebellum is an essential structure for the behavioral expression of learning, and plays an important role in the generation of conditioned hippocampal responses observed in this paradigm.

Animals↗
PubMed2004

How long do the advantages of learning to read a transparent orthography last? An investigation of the reading skills and reading impairment of Welsh children at 10 years of age.

Spencer and Hanley (2003) showed that Welsh-speaking children aged between 5 and 7 years who were learning to read Welsh (a transparent orthography) performed significantly better at reading both real words and nonwords than did English-speaking children living in Wales who were learning to read English (a deep orthography). In this study, the reading skills of these children were reexamined three years later, during their sixth year of formal reading instruction. The children learning to read English continued to perform poorly at reading low- and medium-frequency irregular words but no differences were observed in reading regular words or nonwords. These findings emphasize how long it takes to acquire a large sight vocabulary in English, but indicated that the reading skills of the majority of the English-speaking children had caught up with those of their Welsh-speaking counterparts. However, the poorest 25% of the English readers continued to perform much worse than the lowest performing 25% of Welsh readers on both words and nonwords. An underachieving tail of this kind was not observed in the reading performance of the Welsh-speaking group. Overall, these findings suggest that in the long term the detrimental effects of an opaque orthography are most damaging to the poorest readers.

Child↗
PubMed2003

Morphological and numerical analysis of synaptic interactions between neurons in deep and superficial layers of the entorhinal cortex of the rat.

Neurons providing connections between the deep and superficial layers of the entorhinal cortex (EC) constitute a pivotal link in the network underlying reverberation and gating of neuronal activity in the entorhinal-hippocampal system. To learn more of these deep-to-superficial neurons and their targets, we applied the tracer Neurobiotin pericellularly in layer V of the medial EC of 12 rats. Labeled axons in the superficial layers were studied with light and electron microscopy, and their synaptic organization recorded. Neurobiotin-labeled layer V neurons displayed "Golgi-like" staining. Two major cell types were distinguished among these neurons: (1) pyramidal neurons with apical spiny dendrites traversing all layers and ramifying in layer I, and (2) horizontal neurons with dendrites confined to the deep layers. Labeled axons ramified profusely in layer III, superficially in layer II and deep in layer I. Analysis of labeled axon terminals in layers I-II and III showed that most synapses (95%) were asymmetrical. Of these synapses, 56% occurred with spines (presumably belonging to principal neurons) and 44% with dendritic shafts (presumably interneurons). A small fraction of the synapses (5%) was of the symmetrical type. Such synapses were mainly seen on dendritic shafts. We found in two sections a symmetrical synapse on a spine. These findings suggest that the deep to superficial projection is mainly excitatory in nature, and that these fibers subserve both excitation and feed-forward inhibition. There is an additional, much weaker, inhibitory component in this projection, which may have a disinhibitory effect on the entorhinal network in the superficial layers.

Animals↗
PubMed1987

Incidental and deliberate memory for words and faces after focal cerebral lesions.

It was predicted that right-hemisphere lesions cause a more marked deficit of shallow incidental memory than deep incidental or deliberate memory, the degree of the deficits being reversed after left-hemisphere lesions. This hypothesis was studied by measuring memory for words and faces in brain-damaged patients and nine controls after: shallow incidental learning, i.e. detecting words containing the letter 'A' and counting male and female faces; deep incidental learning, i.e. detecting words that mean man-made things and searching for well-known faces; and deliberate learning of words and faces. The results failed to support the hypothesis, but were consistent with the well-established finding that left- and right-hemisphere lesions disturb memory for words and faces, respectively.

Adult↗
PubMed2006

Twenty-first century learning in schools: A case study of New Technology High School in Napa, California.

The most pertinent question concerning teaching and learning in the twenty-first century is not what knowledge and skills students need--that laundry list was identified over a decade ago--but rather how to foster twenty-first century learning. What curricula, experiences, assessments, environments, and technology best support twenty-first century learning? New Technology High School (NTHS) in Napa, California, is one example of a successful twenty-first century school. In this chapter, the author describes the components of this exemplary high school, illustrating an environment that will cultivate twenty-first century student learning. New Technology High School began by defining eight learning outcomes, aligned with the standards of the Partnership for 21st Century Skills; to graduate, students demonstrate mastery of these outcomes through an online portfolio. To help students achieve the outcomes, NTHS employs project- and problem-based learning. Whereas in traditional classrooms students work alone on short-term assignments that do not lend themselves to deep understanding, the project-based learning approach has students working in teams on long-term, in-depth, rigorous projects. Students' work is supported by the school's workplace-like environment and effectiv use of technology. Meaningful assessment is essential to project-based learning; students receive continuous feedback, helping them become self-directed learners. In fact, NTHS uses outcome-based grading through which students constantly know how they are performing on the twenty-first century outcomes. Research has shown that NTHS graduates are better prepared for postsecondary education, careers, and citizenship than their peers from other schools. To facilitate twenty-first century learning, all schools need to rethink their approach to teaching and learning. New Technology High School is one way to do so.

Adolescent↗
PubMed2000

Cut DVT (deep vein thrombosis) treatment costs with outpatient program for low-molecular-weight heparin.

Learn how to set up an outpatient program for deep venous thrombosis. Although many therapies can be provided more cheaply on an outpatient basis, physicians often fear abandoning hospital care will put their patients at risk. In an innovative program, Harvard Vanguard Medical Associates in Boston was able to shift treatment of its cases for deep venous thrombosis to a clinic-based program, reducing overall costs by 32%. The secret, it says, was building a reliable infrastructure to assure superior patient follow up.

Ambulatory Care↗
PubMed2002

Role of the nuclei in eyeblink conditioning.

Evidence to date supports the strong conclusion that the cerebellum learns. Classical conditioning of the eyeblink response is critically dependent upon the cerebellum. The issue addressed here is whether cerebellar cortex or deep nuclei form the basic association. Learning occurs with large cerebellar cortical aspirations in rabbits and with a Purkinje-cell-deficient mutation in mice. The learned response is poorly timed, small in amplitude, and inconsistent in its occurrence. Learning nevertheless occurs. Lesions of the interpositus, on the other hand, prevent new learning and abolish previously learned conditioned responses. Small electrolytic lesions, kainic acid lesions, and temporary inactivation (cooling, muscimol, anisomycin) localize learning to the dorsolateral anterior interpositus nucleus. Learning-related unit activity-the signature of the engram-recorded throughout the brain depends on the interpositus. Electrical stimulation of interpositus afferents are needed for conditioning, and the conditioned interpositus has a lowered threshold. Finally, a recent anatomical study with electron microscopy shows synaptic changes in the excitatory inputs to the interpositus with conditioning. The interpositus is responsible for making the basic association between conditioned and unconditioned stimuli, which in turn allows ancillary learning to occur in cerebellar cortex, and possibly brainstem and forebrain.

Animals↗
PubMed1996

Impaired classical eyeblink conditioning in cerebellar-lesioned and Purkinje cell degeneration (pcd) mutant mice.

Converging lines of evidence from rabbits, rats, and humans argue for the crucial involvement of the cerebellum in classical conditioning of the eyeblink/nictitating membrane response in mammals. For example, selective lesions (permanent or reversible) of the cerebellum block both acquisition and retention of eyeblink conditioning. Correspondingly, electrophysiological and brain-imaging studies indicate learning-related plasticity in the cerebellum. The involvement of the cerebellum in eyeblink conditioning is also supported by stimulation studies showing that direct stimulation of the two major afferents to the cerebellum (the mossy fibers emanating from the pontine nucleus and climbing fibers originating from the inferior olive) can substitute for the peripheral conditioned stimulus (CS) and unconditioned stimulus (US), respectively, to yield normal behavioral learning. In the present study, we examined the relative contribution of the cerebellar cortex versus deep nuclei (specifically the interpositus nucleus) in eyeblink learning by using mutant mice deficient of Purkinje cells, the exclusive output neurons of the cerebellar cortex. We report that Purkinje cell degeneration (pcd) mice exhibit a profound impairment in the acquisition of delay eyeblink conditioning in comparison with their wild-type littermates. Nevertheless, the pcd animals did acquire a subnormal level of conditioned eyeblink responses. In contrast, wild-type mice with lesions of the interpositus nucleus were completely unable to learn the conditioned eyeblink response. These results suggest that both cerebellar cortex and deep nuclei are important for normal eyeblink conditioning.

Animals↗
PubMed2003

Learning networks in health and Parkinson's disease: reproducibility and treatment effects.

In a previous H(2) (15)O/PET study of motor sequence learning, we used principal components analysis (PCA) of region of interest (ROI) data to identify performance-related activation patterns in normal subjects and patients with Parkinson's disease (PD). In the present study, we determined whether these patterns predicted learning performance in subsequent normal and untreated PD cohorts. Using a voxel-based PCA approach, we correlated the changes in network activity that occurred during antiparkinsonian treatment and their relationship to learning performance. We found that the previously identified ROI-based patterns correlated with learning performance in the prospective normal (P < 0.01) and untreated PD (P < 0.05) cohorts. Voxel analysis revealed that target retrieval was related to a network characterized by bilateral activation of the dorsolateral prefrontal, premotor and anterior cingulate cortex, the precuneus, and the occipital association areas as well as the right ventral prefrontal and inferior parietal regions. Target acquisition was associated with a different network involving activation of the caudate, putamen, and right dentate nucleus, as well as the left ventral prefrontal and inferior parietal areas. Antiparkinsonian therapy gave rise to changes in retrieval performance that correlated with network modulation (P < 0.01). Increases in network activation and learning performance occurred with internal pallidal deep brain stimulation (GPi DBS); decrements in these measures were present with levodopa. Our findings suggest that network analysis of activation data can provide stable descriptors of learning performance. Network quantification can provide an objective means of assessing the effects of therapy on cognitive functioning in neurodegenerative disorders.

Adult↗
PubMed2003

Event-related brain potentials in memory: correlates of episodic, semantic and implicit memory.

OBJECTIVE: To study cognitive evoked potentials, recorded from scalp EEG and foramen ovale electrodes, during activation of explicit and implicit memory. The subgroups of explicit memory, episodic and semantic memory, are looked at separately. METHODS: A word-learning task was used, which has been shown to activate hippocampus in H(2)(15)O positron emission tomography studies. Subjects had to study and remember word pairs using different learning strategies: (i) associative word learning (AWL), which activates the episodic memory, (ii) deep single word encoding (DSWE), which activates the semantic memory, and (iii) shallow single word encoding (SSWE), which activates the implicit memory and serves as a baseline. The test included the 'remember/know' paradigm as a behavioural learning control. During the task condition, a 10-20 scalp EEG with additional electrodes in both temporal lobes regions was recorded from 11 healthy volunteers. In one patient with mesiotemporal lobe epilepsy, the EEG was recorded from bilateral foramen ovale electrodes directly from mesial temporal lobe structures. Event-related potentials (ERPs) were calculated off-line and visual and statistical analyses were made. RESULTS: Associative learning strategy produced the best memory performance and the best noetic awareness experience, whereas shallow single word encoding produced the worst performance and the smallest noetic awareness. Deep single word encoding performance was in between. ERPs differed according to the test condition, during both encoding and retrieval, from both the scalp EEG and the foramen ovale electrode recordings. Encoding showed significant differences between the shallow single word encoding (SSWE), which is mainly a function of graphical characteristics, and the other two strategies, deep single word (DSWE) and associative learning (AWL), in which there is a semantic processing of the meaning. ERPs generated by these two categories, which are both functions of explicit memory, differed as well, indicating the presence or the absence of associative binding. Retrieval showed a significant test effect between the word pairs learned by association (AWL) and the ones learned by encoding the words in isolation of each other (DSWE and SSWE). The comparison of the ERPs generated by autonoetic awareness ('remember') and noetic awareness ('know') exhibited a significant test effect as well. CONCLUSIONS: The results of behavioural data, in particular that of the 'remember/know' procedure, are evidence that the task paradigm was efficient in activating different kinds of memory. Associative word learning generated a high degree of autonoetic awareness, which is a result of the episodic memory, whereas both kinds of single word learning generated less. AWL, DSWE and SSWE resulted in different electrophysiological correlates, both for encoding as well as retrieval, indicating that different brain structures were activated in different temporal sequence.

Adult↗