Search PubMed⌕ Search

SEARCH · Search PubMed

Results for “Machine 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 1,747 records · Page 97Linked to original sources
PubMed2000

Automating parallel implementation of neural learning algorithms.

Neural learning algorithms generally involve a number of identical processing units, which are fully or partially connected, and involve an update function, such as a ramp, a sigmoid or a Gaussian function for instance. Some variations also exist, where units can be heterogeneous, or where an alternative update technique is employed, such as a pulse stream generator. Associated with connections are numerical values that must be adjusted using a learning rule, and and dictated by parameters that are learning rule specific, such as momentum, a learning rate, a temperature, amongst others. Usually, neural learning algorithms involve local updates, and a global interaction between units is often discouraged, except in instances where units are fully connected, or involve synchronous updates. In all of these instances, concurrency within a neural algorithm cannot be fully exploited without a suitable implementation strategy. A design scheme is described for translating a neural learning algorithm from inception to implementation on a parallel machine using PVM or MPI libraries, or onto programmable logic such as FPGAs. A designer must first describe the algorithm using a specialised Neural Language, from which a Petri net (PN) model is constructed automatically for verification, and building a performance model. The PN model can be used to study issues such as synchronisation points, resource sharing and concurrency within a learning rule. Specialised constructs are provided to enable a designer to express various aspects of a learning rule, such as the number and connectivity of neural nodes, the interconnection strategies, and information flows required by the learning algorithm. A scheduling and mapping strategy is then used to translate this PN model onto a multiprocessor template. We demonstrate our technique using a Kohonen and backpropagation learning rules, implemented on a loosely coupled workstation cluster, and a dedicated parallel machine, with PVM libraries.

Algorithms↗
PubMed2004

Dynamics of the adaptive natural gradient descent method for soft committee machines.

Adaptive natural gradient descent (ANGD) method realizes natural gradient descent (NGD) without needing to know the input distribution of learning data and reduces the calculation cost from a cubic order to a square order. However, no performance analysis of ANGD has been done. We have developed a statistical-mechanical theory of the simplified version of ANGD dynamics for soft committee machines in on-line learning; this method provides deterministic learning dynamics expressed through a few order parameters, even though ANGD intrinsically holds a large approximated Fisher information matrix. Numerical results obtained using this theory were consistent with those of a simulation, with respect not only to the learning curve but also to the learning failure. Utilizing this method, we numerically evaluated ANGD efficiency and found that ANGD generally performs as well as NGD. We also revealed the key condition affecting the learning plateau in ANGD.

Journal Article↗
PubMedDate not supplied

[Reading machines for the blind].

Direct conversion reading machines for the blind with tactile and acoustic information output were analyzed. It is shown that direct conversion machines are simple enough, inexpensive and may be recommended for individual use. But an effective use of them requires a protracted period of learning. The present state and future prospects for constructing reading machines with vocal information output (machines with letter-wise recognition and the ones capable of synthetizing the speech) are enquired into. These machines are shown to help reach a greater reading speed, but are of a more complex nature.

Acoustics↗
PubMed2006

The brain as a system of nested but partially overlapping networks. Heuristic relevance of the model for brain physiology and pathology.

A new model of the brain organization is proposed. The model is based on the assumption that a global molecular network enmeshes the entire central nervous system. Thus, brain extra-cellular and intra-cellular molecular networks are proposed to communicate at the level of special plasma membrane regions (e.g., the lipid rafts) where horizontal molecular networks can represent input/output regions allowing the cell to have informational exchanges with the extracellular environment. Furthermore, some "pervasive signals" such as field potentials, pressure waves and thermal gradients that affect large parts of the brain cellular and molecular networks are discussed. Finally, at least two learning paradigms are analyzed taking into account the possible role of Volume Transmission: the so-called model of "temporal difference learning" and the "Turing B-unorganised machine". The relevance of this new view of brain organization for a deeper understanding of some neurophysiological and neuropathological aspects of its function is briefly discussed.

Animals↗
PubMed1986

Computers and videodiscs in pathology education: ECLIPS as an example of one approach.

We have enumerated ways in which the evolving computer and videodisc technologies are being used in pathology education and discussed in some detail the particular use with which we are most familiar, text management. While it is probably premature to speculate as to how these technologies will ultimately affect pathology education, one recent trend--the convergence that seems to be developing between those working on expert consulting systems and those working primarily on educational applications--will probably influence this impact substantially. We believe that we are moving, from opposite directions, toward the same end result, namely, the use of machine intelligence to facilitate and augment human learning. We expect that, as the two groups come closer together, very powerful, interesting, and eminently useful educational tools will emerge. While this is occurring, we think that most would agree that one of the very urgent needs is to develop forums in which the academic and practice communities can interact with researchers and developers. With apologies to Clemenceau, computers are rapidly becoming too important to be left exclusively to computer scientists. Such forums would serve to give these communities a chance to learn what the new technologies have to offer and give developers a better idea of where these technologies can make the greatest contributions.

Artificial Intelligence↗
PubMed2003

Urinary nucleosides as potential tumor markers evaluated by learning vector quantization.

Modified nucleosides were recently presented as potential tumor markers for breast cancer. The patterns of the levels of urinary nucleosides are different for tumor bearing individuals and for healthy individuals. Thus, a powerful pattern recognition method is needed. Although backpropagation (BP) neural networks are becoming increasingly common in medical literature for pattern recognition, it has been shown that often-superior methods exist like learning vector quantization (LVQ) and support vector machines (SVM). The aim of this feasibility study is to get an indication of the performance of urinary nucleoside levels evaluated by LVQ in contrast to the evaluation the popular BP and SVM networks. Urine samples were collected from female breast cancer patients and from healthy females. Twelve different ribonucleosides were isolated and quantified by a high performance liquid chromatography (HPLC) procedure. LVQ, SVM and BP networks were trained and the performance was evaluated by the classification of the test sets into the categories "cancer" and "healthy". All methods showed a good classification with a sensitivity ranging from 58.8 to 70.6% at a specificity of 88.4-94.2% for the test patterns. Although the classification performance of all methods is comparable, the LVQ implementations are superior in terms of more qualitative features: the results of LVQ networks are more reproducible, as the initialization is deterministic. The LVQ networks can be trained by unbalanced sizes of the different classes. LVQ networks are fast during training, need only few parameters adjusted for training and can be retrained by patterns of "local individuals". As at least some of these features play an important role in an implementation into a medical decision support system, it is recommended to use LVQ for an extended study.

Adult↗
PubMed2002

Robots that imitate humans.

The study of social learning in robotics has been motivated by both scientific interest in the learning process and practical desires to produce machines that are useful, flexible, and easy to use. In this review, we introduce the social and task-oriented aspects of robot imitation. We focus on methodologies for addressing two fundamental problems. First, how does the robot know what to imitate? And second, how does the robot map that perception onto its own action repertoire to replicate it? In the future, programming humanoid robots to perform new tasks might be as simple as showing them.

Journal Article↗
PubMed2001

Multi-class protein fold recognition using support vector machines and neural networks.

MOTIVATION: Protein fold recognition is an important approach to structure discovery without relying on sequence similarity. We study this approach with new multi-class classification methods and examined many issues important for a practical recognition system. RESULTS: Most current discriminative methods for protein fold prediction use the one-against-others method, which has the well-known 'False Positives' problem. We investigated two new methods: the unique one-against-others and the all-against-all methods. Both improve prediction accuracy by 14-110% on a dataset containing 27 SCOP folds. We used the Support Vector Machine (SVM) and the Neural Network (NN) learning methods as base classifiers. SVMs converges fast and leads to high accuracy. When scores of multiple parameter datasets are combined, majority voting reduces noise and increases recognition accuracy. We examined many issues involved with large number of classes, including dependencies of prediction accuracy on the number of folds and on the number of representatives in a fold. Overall, recognition systems achieve 56% fold prediction accuracy on a protein test dataset, where most of the proteins have below 25% sequence identity with the proteins used in training.

Discriminant Analysis↗
PubMed1974

Nonprint media as information resources: software and hardware.

Nonprint materials are rapidly becoming important information and learning resource materials for the health science library. Because of their long experience in organizing and utilizing informational materials, libraries represent highly appropriate repositories and sites for utilization of these new materials. Nonprint materials differ from printed materials in several ways, and this may account for the resistance of some librarians to dealing with them. One of the most important differences is that a machine must serve as mediator between the information and the user of nonprint materials. Also, the great variety of formats and machines can confuse the novice. The librarian must learn to deal with these differences in a creative way through a process of cooperation and collaboration with media and educational technology specialists.

Attitude↗
PubMed1999

Recurrent sampling models for the Helmholtz machine.

Many recent analysis-by-synthesis density estimation models of cortical learning and processing have made the crucial simplifying assumption that units within a single layer are mutually independent given the states of units in the layer below or the layer above. In this article, we suggest using either a Markov random field or an alternative stochastic sampling architecture to capture explicitly particular forms of dependence within each layer. We develop the architectures in the context of real and binary Helmholtz machines. Recurrent sampling can be used to capture correlations within layers in the generative or the recognition models, and we also show how these can be combined.

Animals↗
PubMed1997

Phacoemulsification. Technology and fundamentals.

The number one rule of phacoemulsification and aspiration cataract surgery is to know your machine. This chapter is designed to help the surgeon who is currently using phacoemulsification, or those who wish to understand more about technique, learn the basics and technology of the various types of phacoemulsification machines. Fluidics, pump design, handpiece mechanics, phacoemulsification needles, and fundamentals of phacoemulsification of cataracts will be reviewed.

Animals↗
PubMed2006

Prediction of compounds with specific pharmacodynamic, pharmacokinetic or toxicological property by statistical learning methods.

Computational methods for predicting compounds of specific pharmacodynamic, pharmacokinetic, or toxicological property are useful for facilitating drug discovery and drug safety evaluation. The quantitative structure-activity relationship (QSAR) and quantitative structure-property relationship (QSPR) methods are the most successfully used statistical learning methods for predicting compounds of specific property. More recently, other statistical learning methods such as neural networks and support vector machines have been explored for predicting compounds of higher structural diversity than those covered by QSAR and QSPR. These methods have shown promising potential in a number of studies. This article is intended to review the strategies, current progresses and underlying difficulties in using statistical learning methods for predicting compounds of specific property. It also evaluates algorithms commonly used for representing structural and physicochemical properties of compounds.

Pharmacokinetics↗
PubMedDate not supplied

Alumina-zirconia machinable abutments for implant-supported single-tooth anterior crowns.

Innovative materials and application techniques are constantly being developed in the ongoing search for improved restorations. This article describes a new material and the fabrication process of aesthetic machinable ceramic anterior implant abutments. The ceramic material utilized is a mixture of alumina (aluminum oxide) and ceria (cerium oxide) with partially stabilized zirconia (zirconium oxide). The initial core material is a cylinder with a 9-mm diameter and a 15-mm height, obtained by ceramic injection and presintering processes. The resultant alumina-zirconia core is porous and readily machinable. It is secured to the analog, and its design is customized by machining the abutment to suit the particular clinical circumstances. The machining is followed by glass infiltration, and the crown is finalized. The learning objective of this article is to gain a basic knowledge of the fabrication and clinical application of the custom machinable abutments.

Aluminum Oxide↗
PubMed2005

Regularized Least Squares Cancer classifiers from DNA microarray data.

BACKGROUND: The advent of the technology of DNA microarrays constitutes an epochal change in the classification and discovery of different types of cancer because the information provided by DNA microarrays allows an approach to the problem of cancer analysis from a quantitative rather than qualitative point of view. Cancer classification requires well founded mathematical methods which are able to predict the status of new specimens with high significance levels starting from a limited number of data. In this paper we assess the performances of Regularized Least Squares (RLS) classifiers, originally proposed in regularization theory, by comparing them with Support Vector Machines (SVM), the state-of-the-art supervised learning technique for cancer classification by DNA microarray data. The performances of both approaches have been also investigated with respect to the number of selected genes and different gene selection strategies. RESULTS: We show that RLS classifiers have performances comparable to those of SVM classifiers as the Leave-One-Out (LOO) error evaluated on three different data sets shows. The main advantage of RLS machines is that for solving a classification problem they use a linear system of order equal to either the number of features or the number of training examples. Moreover, RLS machines allow to get an exact measure of the LOO error with just one training. CONCLUSION: RLS classifiers are a valuable alternative to SVM classifiers for the problem of cancer classification by gene expression data, due to their simplicity and low computational complexity. Moreover, RLS classifiers show generalization ability comparable to the ones of SVM classifiers also in the case the classification of new specimens involves very few gene expression levels.

Algorithms↗
PubMed2004

Performance of the Alex AVX-2 MIMD architecture in learning the NetTalk database.

The process of training neural networks on parallel architectures has been used to assess the performance of so many parallel machines. In this paper, we are investigating the implementation of backpropagation (BP) on the Alex AVX-2 coarse-grained MIMD machine. A host-worker parallel implementation is carried out in order to train different networks to learn the NetTalk dictionary. First, a computational model is constructed using a single processor to complete the learning process. Also, a communication model for the host-worker topology is developed in order to compute the communication overhead in the broadcasting/gathering process. Both models are then used to predict the machine performance when p processors are used and a comparison with the actual measured performance of the parallel architecture implementation is carried out. Simulation results show that both models can be used effectively to predict the machine performance for the NetTalk problem. Finally, a comparison between the AVX-2 NetTalk implementation and the performance of other parallel platforms is presented.

Artificial Intelligence↗
PubMed2003

Depth cue reliance in surgeons and medical students.

BACKGROUND: Depth perception is reduced in endoscopic surgery, although little is known about the effect this has on surgical performance. METHODS: To assess the role of depth cues, 45 subjects completed tests of depth cue reliance. Surgical skill was assessed using the Minimally Invasive Surgical Trainer-Virtual Reality, a previously validated laparoscopic simulator. RESULTS: We could demonstrate no difference in cue reliance for three depth cues--namely stereo, texture, and outline--between surgeons and medical students. Greater dominance on stereo for medical students was a positive finding and a negative finding for the surgeons when correlated with surgical performance. CONCLUSIONS: We suggest that surgeons learn to adapt to the nonstereo environment in MIS, and this is the first study to show evidence of this phenomenon. This difference in stereo reliance is a reflection of the experience that surgeons have with laparoscopy compared with medical students, who have none.

Adult↗
PubMed2006

Construct and face validity and task workload for laparoscopic camera navigation: virtual reality versus videotrainer systems at the SAGES Learning Center.

BACKGROUND: Laparoscopic camera navigation (LCN) training on simulators has demonstrated transferability to actual operations, but no comparative data exist. The objective of this study was to compare the construct and face validity, as well as workload, of two previously validated virtual reality (VR) and videotrainer (VT) systems. METHODS: Attendees (n = 90) of the SAGES 2005 Learning Center performed two repetitions on both VR (EndoTower) and VT (Tulane Trainer) LCN systems using 30 degrees laparoscopes and completed a questionnaire regarding demographics, simulator characteristics, and task workload. Construct validity was determined by comparing the performance scores of subjects with various levels of experience according to five parameters and face validity according to eight. The validated NASA-TLX questionnaire that rates the mental, physical, and temporal demand of a task as well as the performance, effort, and frustration of the subject was used for workload measurement. RESULTS: Construct validity was demonstrated for both simulators according to the number of basic laparoscopic cases (p = 0.005), number of advanced cases (p < 0.001), and frequency of angled scope use (p < 0.001), and only for VT according to training level (p < 0.001) and fellowship training (p = 0.008). Face validity ratings on a 1-20 scale averaged 15.4 +/- 3 for VR vs. 16 +/- 2.6 for VT (p = 0.04). Ninety-six percent of participants rated both simulators as valid educational tools. The NASA-TLX overall workload score was 69.5 +/- 24 for VR vs. 68.8 +/- 20.5 for VT (p = 0.31). CONCLUSIONS: This is the largest study to date that compares two validated LCN simulators. While subtle differences exist, both VR and VT simulators demonstrated excellent construct validity, good face validity, and acceptable workload parameters. These systems thus represent useful training devices and should be widely used to improve surgical performance.

Adult↗
PubMed2004

Benign/malignant classifier of soft tissue tumors using MR imaging.

UNLABELLED: This article presents a pattern-recognition approach to the soft tissue tumors (STT) benign/malignant character diagnosis using magnetic resonance (MR) imaging applied to a large multicenter database. OBJECTIVE: To develop and test an automatic classifier of STT into benign or malignant by using classical MR imaging findings and epidemiological information. MATERIALS AND METHODS: A database of 430 patients (62% benign and 38% malignant) from several European multicenter registers. There were 61 different histologies (36 with benign and 25 with malignant nature). Three pattern-recognition methods (artificial neural networks, support vector machine, k-nearest neighbor) were applied to learn the discrimination between benignity and malignancy based on a defined MR imaging findings protocol. After the systems had learned by using training samples (with 302 cases), the clinical decision support system was tested in the diagnosis of 128 new STT cases. RESULTS: An 88-92% efficacy was obtained in a not-viewed set of tumors using the pattern-recognition techniques. The best results were obtained with a back-propagation artificial neural network. CONCLUSION: Benign vs. malignant STT discrimination is accurate by using pattern-recognition methods based on classical MR image findings. This objective tool will assist radiologists in STT grading.

Algorithms↗