Search PubMed⌕ Search

SEARCH · Search PubMed

Results for “sparse 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 73 records · Page 4Linked to original sources
PubMed2001

Stepping out of the box: information processing in the neural networks of the basal ganglia.

The Albin-DeLong 'box and arrow' model has long been the accepted standard model for the basal ganglia network. However, advances in physiological and anatomical research have enabled a more detailed neural network approach. Recent computational models hold that the basal ganglia use reinforcement signals and local competitive learning rules to reduce the dimensionality of sparse cortical information. These models predict a steady-state situation with diminished efficacy of lateral inhibition and low synchronization. In this framework, Parkinson's disease can be characterized as a persistent state of negative reinforcement, inefficient dimensionality reduction, and abnormally synchronized basal ganglia activity.

Animals↗
PubMed1998

An Equivalence Between Sparse Approximation and Support Vector Machines.

This article shows a relationship between two different approximation techniques: the support vector machines (SVM), proposed by V. Vapnik (1995) and a sparse approximation scheme that resembles the basis pursuit denoising algorithm (Chen, 1995; Chen, Donoho, and Saunders, 1995). SVM is a technique that can be derived from the structural risk minimization principle (Vapnik, 1982) and can be used to estimate the parameters of several different approximation schemes, including radial basis functions, algebraic and trigonometric polynomials, B-splines, and some forms of multilayer perceptrons. Basis pursuit denoising is a sparse approximation technique in which a function is reconstructed by using a small number of basis functions chosen from a large set (the dictionary). We show that if the data are noiseless, the modified version of basis pursuit denoising proposed in this article is equivalent to SVM in the following sense: if applied to the same data set, the two techniques give the same solution, which is obtained by solving the same quadratic programming problem. In the appendix, we present a derivation of the SVM technique in one framework of regularization theory, rather than statistical learning theory, establishing a connection between SVM, sparse approximation, and regularization theory.

Journal Article↗
PubMed2006

Are sparse-coding simple cell receptive field models physiologically plausible?

Olshausen and Field (1996) developed a simple cell receptive field model for natural scene processing in V1, based on unsupervised learning and non-orthogonal basis function optimization of an overcomplete representation of visual space. The model was originally tested with an ensemble of whitened natural scenes, simulating pre-cortical filtering in the retinal ganglia and lateral geniculate nucleus, and the basis functions qualitatively resembled the orientation-specific responses of V1 simple cells in the spatial domain. In this study, the quantitative tuning responses of the basis functions in the spectral domain are estimated using a Gaussian model, to determine their goodness-of-fit to the known bandwidths of simple cells in primate V1. Five simulation experiments which examined key features of the model are reported: changing the size of the basis functions; using a complete versus over-complete representation; changing the sparseness factor; using a variable learning rate; and mapping the basis functions with a whitening spatial function. The key finding of this study is that across all image themes, basis function sizes, number of basis functions, sparseness factors and learning rates, the spatial-frequency tuning did not closely resemble that of primate area 17 -- the model results more closely resembled the unclassified cat neurones of area 19 with a single exception, and not area 17 as predicted.

Action Potentials↗
PubMed2003

Learning and inference in the brain.

This article is about how the brain data mines its sensory inputs. There are several architectural principles of functional brain anatomy that have emerged from careful anatomic and physiologic studies over the past century. These principles are considered in the light of representational learning to see if they could have been predicted a priori on the basis of purely theoretical considerations. We first review the organisation of hierarchical sensory cortices, paying special attention to the distinction between forward and backward connections. We then review various approaches to representational learning as special cases of generative models, starting with supervised learning and ending with learning based upon empirical Bayes. The latter predicts many features, such as a hierarchical cortical system, prevalent top-down backward influences and functional asymmetries between forward and backward connections that are seen in the real brain. The key points made in article are: (i). hierarchical generative models enable the learning of empirical priors and eschew prior assumptions about the causes of sensory input that are inherent in non-hierarchical models. These assumptions are necessary for learning schemes based on information theory and efficient or sparse coding, but are not necessary in a hierarchical context. Critically, the anatomical infrastructure that may implement generative models in the brain is hierarchical. Furthermore, learning based on empirical Bayes can proceed in a biologically plausible way. (ii). The second point is that backward connections are essential if the processes generating inputs cannot be inverted, or the inversion cannot be parameterised. Because these processes involve many-to-one mappings, are non-linear and dynamic in nature, they are generally non-invertible. This enforces an explicit parameterisation of generative models (i.e. backward connections) to afford recognition and suggests that forward architectures, on their own, are not sufficient for perception. (iii). Finally, non-linearities in generative models, mediated by backward connections, require these connections to be modulatory, so that representations in higher cortical levels can interact to predict responses in lower levels. This is important in relation to functional asymmetries in forward and backward connections that have been demonstrated empirically.

Bayes Theorem↗
PubMed2005

Recognition of a familiar place by the honeybee (Apis mellifera).

Recent work shows that at any one place bees detect a limited variety of simple cues in parallel. At each choice point, they recognize a few cues in the range of positions where the cues occurred during the learning process. There is no need to postulate that they re-assemble the surrounding panorama in memory; only that they retain memories of the coincidences of cues in the expected retinotopic directions. The cues could be stimuli that excite groups of peripheral visual neurons. All the experimentally known cues are described, including modulation of the receptors, the locations of areas of black or colour, the nearness, size, averaged edge orientation, and radial and tangential edges. Cues of each type are separately summed within large fields, the size of which varies with the cue. Local orientation cues from edges at right angles cancel each other within each field, which also suggests that the discrimination of shape and texture is limited. Resolution depends on lateral interactions and the number of ommatidia required for each cue. To identify a new place, a few sparse cues, together with their directions, are learned in orientation flights. When the bee returns, the cues in the panorama are progressively matched as they coincide with the cues in memory. The limited number of cues, though economical for memory, may restrict the foraging behaviour and lead to flower constancy. This kind of a visual system is a candidate model for other animals or machines with economical processing systems.

Animals↗
PubMed2006

Tikhonov training of the CMAC neural network.

The architecture of the cerebellar model articulation controller (CMAC) presents a rigid compromise between learning and generalization. In the presence of a sparse training dataset, this limitation manifestly causes overfitting, a drawback that is not overcome by current training algorithms. This paper proposes a novel training framework founded on the Tikhonov regularization, which relates to the minimization of the power of the sigma-order derivative. This smoothness criterion yields to an internal cell-interaction mechanism that increases the generalization beyond the degree hardcoded in the CMAC architecture while preserving the potential CMAC learning capabilities. The resulting training mechanism, which proves to be simple and computationally efficient, is deduced from a rigorous theoretical study. The performance of the new training framework is validated against comparative benchmarks from the DELVE environment.

Algorithms↗
PubMed2006

Microinfusions of neurotensin antagonist SR 48692 within the nucleus accumbens core impair spatial learning in rats.

The involvement of neurotensin (NT) within the nucleus accumbens core (NAC) in behavior has been sparsely investigated. Moreover, little is known of what role NT within the ventral striatum has on spatial learning. The present study investigated whether NT receptors in the NAC are implicated in learning of spatial information. Male Long-Evans rats were trained on a food search spatial learning task. Rats were microinfused with either NT antagonist SR 48692 (50 nM/0.5 =L) or saline in the NAC before each training session. Rats treated with SR 48692 made more reference and working memory errors during the acquisition of spatial learning than did rats infused with saline. These results suggest that NT receptors contribute to NAC-mediated spatial learning.

Animals↗
PubMed1987

Evidence for the inhibitory neurotransmitter gamma-aminobutyric acid in aspiny and sparsely spiny nonpyramidal neurons of the turtle dorsal cortex.

In order to learn more about the anatomical substrate for gamma-aminobutyric acid (GABA)-mediated inhibition in cortical structures, the intrinsic neuronal organization of turtle dorsal cortex was studied by using Golgi impregnation, immunohistochemical localization of GABA and its synthetic enzyme glutamic acid decarboxylase (GAD), and histochemical localization of the presynaptic GABA-degrading enzyme GABA-transaminase (GABA-T). GABAergic markers are found in neurons identical in morphology and distribution to Golgi-impregnated aspiny and sparsely spiny nonpyramidal neurons with locally arborizing axons and appear to label most if not all of the nonpyramidal neurons. In addition, the GABAergic markers are found in punctate structures in a distribution characteristic of presumed inhibitory terminals. The spine-laden pyramidal neurons, the principal projecting cell type in the dorsal cortex, are devoid of labelling for GABAergic markers but are surrounded by presumed GABAergic terminals. The data complement previous physiological and ultrastructural studies that implicate aspiny and sparsely spiny nonpyramidal neurons as mediators of intrinsic inhibition of pyramidal neurons in turtle cortex. The results also suggest similarities in the functional organization of intrinsic inhibitory elements in turtle and mammalian cortex.

4-Aminobutyrate Transaminase↗
PubMed2026

Comparative effectiveness of game-based learning modalities in nursing and medical education: a systematic review and Bayesian network meta-analysis.

BACKGROUND: Game-based learning (GBL) is increasingly used in healthcare education, but educators must choose among diverse modalities (e.g., quiz platforms, apps, serious games and metaverse environments). Comparative evidence on which modalities perform best across learning domains (knowledge, attitudes, and practice) remains limited. AIM: To compare the effects of distinct GBL modalities on knowledge, attitudes, and practice outcomes in nursing and medical education and to explore whether comparative effects differ by learner group (pre-licensure students and in-service professionals). DESIGN: PRISMA-NMA-aligned systematic review and Bayesian network meta-analysis. METHODS: We searched eight databases and trial registries through September 2, 2024, for randomized controlled trials comparing GBL with traditional teaching (TT). Outcomes were transformed to a 0-100 scale and analysed as change from baseline in Bayesian consistency models; random-effects models were selected using deviance information criterion (DIC). Risk of bias was assessed using RoB 2. We report mean differences (MDs) with 95% credible intervals (CrIs) versus TT, ranking probabilities, and subgroup NMAs by learner group. RESULTS: Thirty-one RCTs (n = 3439) were included; 15 contributed complete data to the network. Risk of bias was low in 15 trials and raised some concerns in 16. The network was modest for knowledge (11 trials) and sparse for attitudes (3) and practice (4). Compared with TT, metaverse-based learning showed improved attitudes (MD 15; 95% CrI 12 to 18), based on a single trial. For knowledge and practice, Kahoot-based quizzes (MD 9.1; 95% CrI -8.9 to 27) and app-based learning (MD 4.6; 95% CrI -4.4 to 14) had the highest estimated mean improvements, but credible intervals were wide and included the null for most comparisons. Subgroup rankings differed by learner group, but several comparisons were imprecise and uncertainty was substantial, particularly in sparse networks. CONCLUSIONS: GBL modalities may improve learning outcomes compared with TT, but relative effects appear domain-specific and the certainty of rankings is limited by sparse evidence and imprecision. Future trials should prioritise head-to-head comparisons, robust outcome measurement, and longer-term retention and transfer outcomes in both student and in-service populations.

Humans↗
PubMed2006

Associative learning in hierarchical self-organizing learning arrays.

In this paper, we introduce feedback-based associative learning in self-organized learning arrays (SOLAR). SOLAR structures are hierarchically organized networks of sparsely connected neurons that define their own functions and select their interconnections locally. This paper provides a description of neuron self-organization and signal processing. Feedforward processing is used to make necessary correlations and learn the input patterns. Discovered associations between neuron inputs are used to generate feedback signals. These feedback signals, when propagated to the primary inputs, can establish the expected input values. This can be used for heteroassociative (HA) and autoassociative (AA) learning and pattern recognition. Example applications in HA learning are given.

Algorithms↗
PubMed1994

Influence of treatment with L-deprenyl on the structure of the cerebellar cortex of aged rats.

Treatment with L-deprenyl increases mean and maximum life span in the rat and reverses memory and learning deficits associated with old age. Since only sparse information is available concerning the influence of L-deprenyl administration on the aging brain microanatomy, we have investigated the effect of long-term treatment with L-deprenyl on the structure of the cerebellar cortex in the aged rat. The cerebellar cortex was used since it represents a useful model for assessing age-related changes in nervous system anatomy and function. Male Sprague-Dawley rats were treated from the 19th to the 24th month of age with a daily oral dose of 1.25 mg/kg or 5 mg/kg L-deprenyl. Age-matched rats were left untreated and used as a control group. Eleven-month-old untreated rats were used as an adult reference group. The density of Purkinje and granule neuron profiles as well as the intensity of Nissl's staining within the cytoplasm of Purkinje neurons were reduced in 24-month in comparison with 11-month rats. Moreover, an increased accumulation of lipofuscin was noticeable in the cytoplasm of Purkinje neurons of old rats as well as an increase in MAO-B activity in the molecular layer of the cerebellar cortex. The two doses of L-deprenyl increased the density of both Purkinje and granule neuron profiles and the intensity of Nissl's staining in the cytoplasm of Purkinje neurons and reduced lipofuscin deposition within Purkinje neurons. The lower dose of L-deprenyl caused only a slight decrease in MAO-B activity, whereas the 5-mg/kg/day dose remarkably reduced it. These results suggest that long-term treatment with L-deprenyl counters the expression of some age-related microanatomical changes in the rat cerebellar cortex. The possible independence of the effects of the compound on age-related microanatomical changes of the cerebellar cortex and on MAO-B inhibitory activity is discussed.

Aging↗
PubMed2002

On the neural correlates of object recognition awareness: relationship to computational activities and activities mediating perceptual awareness.

Based on theoretical considerations of Aurell (1979) and Block (1995), we argue that object recognition awareness is distinct from purely sensory awareness and that the former is mediated by neuronal activities in areas that are separate and distinct from cortical sensory areas. We propose that two of the principal functions of neuronal activities in sensory cortex, which are to provide sensory awareness and to effect the computations that are necessary for object recognition, are dissociated. We provide examples of how this dissociation might be achieved and argue that the components of the neuronal activities which carry the computations do not directly enter the awareness of the subject. The results of these computations are sparse representations (i.e., vector or distributed codes) which are activated by the presentation of particular sensory objects and are essentially engrams for the recognition of objects. These final representations occur in the highest order areas of sensory cortex; in the visual analyzer, the areas include the anterior part of the inferior temporal cortex and the perirhinal cortex. We propose, based on lesion and connectional data, that the two areas in which activities provide recognition awareness are the temporopolar cortex and the medial orbitofrontal cortex. Activities in the temporopolar cortex provide the recognition awareness of objects learned in the remote past (consolidated object recognition), and those in the medial orbitofrontal cortex provide the recognition awareness of objects learned in the recent past. The activation of the sparse representation for a particular sensory object in turn activates neurons in one or both of these regions of cortex, and it is the activities of these neurons that provide the awareness of recognition of the object in question. The neural circuitry involved in the activation of these representations is discussed.

Awareness↗
PubMed2000

Controlling activity fluctuations in large, sparsely connected random networks.

Controlling activity in recurrent neural network models of brain regions is essential both to enable effective learning and to reproduce the low activities that exist in some cortical regions such as hippocampal region CA3. Previous studies of sparse, random, recurrent networks constructed with McCulloch-Pitts neurons used probabilistic arguments to set the parameters that control activity. Here, we extend this work by adding an additional, biologically appropriate, parameter to control the magnitude and stability of activity oscillations. The new constant can be considered to be the rest conductance in a shunting model or the threshold when subtractive inhibition is used. This new parameter is critical for large networks run at low activity levels. Importantly, extreme activity fluctuations that act to turn large networks totally on or totally off can now be avoided. We also show how the size of external input activity interacts with this parameter to affect network activity. Then the model based on fixed weights is extended to estimate activities in networks with distributed weights. Because the theory provides accurate control of activity fluctuations, the approach can be used to design a predictable amount of pseudorandomness into deterministic networks. Such nonminimal fluctuations improve learning in simulations trained on the transitive inference problem.

Computer Simulation↗
PubMed2004

Cortical rewiring and information storage.

Current thinking about long-term memory in the cortex is focused on changes in the strengths of connections between neurons. But ongoing structural plasticity in the adult brain, including synapse formation/elimination and remodelling of axons and dendrites, suggests that memory could also depend on learning-induced changes in the cortical 'wiring diagram'. Given that the cortex is sparsely connected, wiring plasticity could provide a substantial boost in storage capacity, although at a cost of more elaborate biological machinery and slower learning.

Animals↗
PubMed2001

Blind source separation by sparse decomposition in a signal dictionary.

The blind source separation problem is to extract the underlying source signals from a set of linear mixtures, where the mixing matrix is unknown. This situation is common in acoustics, radio, medical signal and image processing, hyperspectral imaging, and other areas. We suggest a two-stage separation process: a priori selection of a possibly overcomplete signal dictionary (for instance, a wavelet frame or a learned dictionary) in which the sources are assumed to be sparsely representable, followed by unmixing the sources by exploiting the their sparse representability. We consider the general case of more sources than mixtures, but also derive a more efficient algorithm in the case of a nonovercomplete dictionary and an equal numbers of sources and mixtures. Experiments with artificial signals and musical sounds demonstrate significantly better separation than other known techniques.

Journal Article↗
PubMed2025

SHICEDO: single-cell Hi-C data enhancement with reduced over-smoothing.

MOTIVATION: Single-cell Hi-C (scHi-C) technologies have significantly advanced our understanding of the 3D genome organization. However, scHi-C data are often sparse and noisy, leading to substantial computational challenges in downstream analyses. RESULTS: In this study, we introduce SHICEDO, a novel deep-learning model specifically designed to enhance scHi-C contact matrices by imputing missing or sparsely captured chromatin contacts through a generative adversarial framework. SHICEDO leverages the unique structural characteristics of scHi-C matrices to derive customized features that enable effective data enhancement. Additionally, the model incorporates a channel-wise attention mechanism to mitigate the over-smoothing issue commonly associated with scHi-C enhancement methods. Through simulations and real-data applications, we demonstrate that SHICEDO outperforms the state-of-the-art methods, achieving superior quantitative and qualitative results. Moreover, SHICEDO enhances key structural features in scHi-C data, thus enabling more precise delineation of chromatin structures such as A/B compartments, TAD-like domains, and chromatin loops. AVAILABILITY AND IMPLEMENTATION: SHICEDO is publicly available at https://github.com/wmalab/SHICEDO.

Single-Cell Analysis↗
PubMed1996

Problem-based medical education in general practice and health care quality assurance.

States that health systems globally are in transition, for which also a correspondingly reformed education is required, involving new--and renewed--basic sciences, principles, skills and methods. To that end, demonstrates how problem-based learning in general practice is almost ideal, although still sparsely implemented, and has much in common with health care quality assurance. Contends that primary care in many ways comprises the most important platform and vehicle of modern health services and their interaction with the population. Uses the example of the Faculty of Health Sciences in Linköping, where all curricula are integrated, community-oriented, problem-based and have a focus on primary care and general practice. Represents a useful and hitherto insufficiently recognized widening of the overall field of health care quality assurance, and summarizes some of the experiences therein. Also provides a brief overview of the literature.

Competency-Based Education↗
PubMed2004

The generalized LASSO.

In the last few years, the support vector machine (SVM) method has motivated new interest in kernel regression techniques. Although the SVM has been shown to exhibit excellent generalization properties in many experiments, it suffers from several drawbacks, both of a theoretical and a technical nature: the absence of probabilistic outputs, the restriction to Mercer kernels, and the steep growth of the number of support vectors with increasing size of the training set. In this paper, we present a different class of kernel regressors that effectively overcome the above problems. We call this approach generalized LASSO regression. It has a clear probabilistic interpretation, can handle learning sets that are corrupted by outliers, produces extremely sparse solutions, and is capable of dealing with large-scale problems. For regression functionals which can be modeled as iteratively reweighted least-squares (IRLS) problems, we present a highly efficient algorithm with guaranteed global convergence. This defies a unique framework for sparse regression models in the very rich class of IRLS models, including various types of robust regression models and logistic regression. Performance studies for many standard benchmark datasets effectively demonstrate the advantages of this model over related approaches.

Algorithms↗