Search PubMed⌕ Search

SEARCH · Search PubMed

Results for “Benchmark”

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 1,603 records · Page 89Linked to original sources
PubMed2026

Network methods for diagonal integration of unpaired single-cell multiomics data: a review.

MOTIVATION: Advances in single-cell sequencing have enabled multiomics profiling at unprecedented resolution; however, mass spectrometry-based single-cell proteomics (scMS) remains inherently destructive, precluding simultaneous transcriptomic capture. Unlike antibody-based methods such as CITE-seq, which permit paired profiling but are restricted to targeted protein panels, scMS provides unbiased, genome-scale coverage of the intracellular proteome yet necessitates post hoc integration of unpaired datasets. This diagonal integration challenge, where transcriptomes and proteomes are measured in separate cells lacking shared anchors, remains underserved by existing reviews, which focus predominantly on vertical integration strategies enabled by non-destructive assays. RESULTS: We survey the complete computational pipeline for constructing mechanistic proteogenomic networks from unpaired single-cell data, covering: (i) unimodal network inference such as knowledge-based approaches, probabilistic graphical models, temporal directionality inference, and generative and foundation model strategies that establish the transcriptomic scaffold; (ii) cross-modal integration architectures such as network propagation, graph neural networks (scMRDR, scmFormer, scCotag), and consensus frameworks designed explicitly for the unpaired proteomics setting; and (iii) benchmarking paradigms spanning network reconstruction (BEELINE, GRETA, CausalBench) and multi-task integration evaluation (scMultiBench, SCMMIB), with guidance on metric selection under network sparsity and class imbalance. We identify three principal axes of future development: generative proteomic translation from transcriptomic precursors, inductive prior embedding in next-generation architectures, and perturbation-based causal benchmarking. AVAILABILITY AND IMPLEMENTATION: This is a review article; no novel software is distributed. A curated benchmark resource table, methods starter guide, and per-method bottleneck annotations are provided in the Supplementary Material.

Multiomics↗
PubMed2026

GUANinE v1.1 reveals complementarity of supervised and genomic language models.

There has been much debate about the benefits of supervised versus unsupervised learning on genomes. Determining which is better in what contexts requires developing comprehensive benchmarks spanning functional and evolutionary tasks. Importantly, such benchmarks need large sample sizes to enable well-powered ranking of models. Having developed and applied such a benchmark here (GUANinE v1.1), we conclusively demonstrate each paradigm offers key advantages and outperforms on certain tasks. In accordance with training, supervised sequence-to-function models exhibit strong performance when annotating functional states characterized by chromatin accessibility or histone marks, while self-supervised language models outperform on evolutionary conservation. Our hundreds of new evaluations in this v1.1 expansion provide evidence for a tradeoff between input context size and model parameter count for a fixed compute budget, which we depict with new metrics such as kiloparameters/base pair. We also construct two new large-scale variant interpretation tasks in v1.1: cadd-snv measuring deleteriousness, and clinvar-snv measuring clinical pathogenicity. We find that conservation scores, and by extension, genomic language models, predict deleteriousness well, but successfully translating deleteriousness predictions to pathogenicity remains challenging. GUANinE v1.1 newly evaluates dozens of pretrained genomic models, and we conclude that moderate-context hybrid or post-trained language models may define the next era of machine learning in genomics.

Genomics↗
PubMed2003

PDP: protein domain parser.

UNLABELLED: We have developed a program for automatic identification of domains in protein three-dimensional structures. Performance of the program was assessed by three different benchmarks: (i) by comparison with the expert-curated SCOP database of structural domains; (ii) by comparison with a collection of manual domain assignments; and (iii) by comparison with a set of 55 proteins, frequently used as a benchmark for automatic domain assignment. In all these benchmarks PDP identified domains correctly in more than 80% of proteins. AVAILABILITY: http://123d.ncifcrf.gov/.

Algorithms↗
PubMed2006

Probalign: multiple sequence alignment using partition function posterior probabilities.

MOTIVATION: The maximum expected accuracy optimization criterion for multiple sequence alignment uses pairwise posterior probabilities of residues to align sequences. The partition function methodology is one way of estimating these probabilities. Here, we combine these two ideas for the first time to construct maximal expected accuracy sequence alignments. RESULTS: We bridge the two techniques within the program Probalign. Our results indicate that Probalign alignments are generally more accurate than other leading multiple sequence alignment methods (i.e. Probcons, MAFFT and MUSCLE) on the BAliBASE 3.0 protein alignment benchmark. Similarly, Probalign also outperforms these methods on the HOMSTRAD and OXBENCH benchmarks. Probalign ranks statistically highest (P-value < 0.005) on all three benchmarks. Deeper scrutiny of the technique indicates that the improvements are largest on datasets containing N/C-terminal extensions and on datasets containing long and heterogeneous length proteins. These points are demonstrated on both real and simulated data. Finally, our method also produces accurate alignments on long and heterogeneous length datasets containing protein repeats. Here, alignment accuracy scores are at least 10% and 15% higher than the other three methods when standard deviation of length is >300 and 400, respectively. AVAILABILITY: Open source code implementing Probalign as well as for producing the simulated data, and all real and simulated data are freely available from http://www.cs.njit.edu/usman/probalign

Algorithms↗
PubMed2005

MEDock: a web server for efficient prediction of ligand binding sites based on a novel optimization algorithm.

The prediction of ligand binding sites is an essential part of the drug discovery process. Knowing the location of binding sites greatly facilitates the search for hits, the lead optimization process, the design of site-directed mutagenesis experiments and the hunt for structural features that influence the selectivity of binding in order to minimize the drug's adverse effects. However, docking is still the rate-limiting step for such predictions; consequently, much more efficient algorithms are required. In this article, the design of the MEDock web server is described. The goal of this sever is to provide an efficient utility for predicting ligand binding sites. The MEDock web server incorporates a global search strategy that exploits the maximum entropy property of the Gaussian probability distribution in the context of information theory. As a result of the global search strategy, the optimization algorithm incorporated in MEDock is significantly superior when dealing with very rugged energy landscapes, which usually have insurmountable barriers. This article describes four different benchmark cases that span a diverse set of different types of ligand binding interactions. These benchmarks were compared with the use of the Lamarckian genetic algorithm (LGA), which is the major workhorse of the well-known AutoDock program. These results demonstrate that MEDock consistently converged to the correct binding modes with significantly smaller numbers of energy evaluations than the LGA required. When judged by a threshold of the number of energy evaluations consumed in the docking simulation, MEDock also greatly elevates the rate of accurate predictions for all benchmark cases. MEDock is available at http://medock.csie.ntu.edu.tw/ and http://bioinfo.mc.ntu.edu.tw/medock/.

Algorithms↗
PubMed2001

Mitochondrial versus nuclear gene sequences in deep-level mammalian phylogeny reconstruction.

Both mitochondrial and nuclear gene sequences have been employed in efforts to reconstruct deep-level phylogenetic relationships. A fundamental question in molecular systematics concerns the efficacy of different types of sequences in recovering clades at different taxonomic levels. We compared the performance of four mitochondrial data sets (cytochrome b, cytochrome oxidase II, NADH dehydrogenase subunit I, 12S rRNA-tRNA-16S rRNA) and eight nuclear data sets (exonic regions of alpha-2B adrenergic receptor, aquaporin, ss-casein, gamma-fibrinogen, interphotoreceptor retinoid binding protein, kappa-casein, protamine, von Willebrand Factor) in recovering deep-level mammalian clades. We employed parsimony and minimum-evolution with a variety of distance corrections for superimposed substitutions. In 32 different pairwise comparisons between these mitochondrial and nuclear data sets, we used the maximum set of overlapping taxa. In each case, the variable-length bootstrap was used to resample at the size of the smaller data set. The nuclear exons consistently performed better than mitochondrial protein and rRNA-tRNA coding genes on a per-residue basis in recovering benchmark clades. We also concatenated nuclear genes for overlapping taxa and made comparisons with concatenated mitochondrial protein-coding genes from complete mitochondrial genomes. The variable-length bootstrap was used to score the recovery of benchmark clades as a function of the number of resampled base pairs. In every case, the nuclear concatenations were more efficient than the mitochondrial concatenations in recovering benchmark clades. Among genes included in our study, the nuclear genes were much less affected by superimposed substitutions. Nuclear genes having appropriate rates of substitution should receive strong consideration in efforts to reconstruct deep-level phylogenetic relationships.

Animals↗
PubMed2005

Radiation shielding and dosimetry experiments updates in the SINBAD database.

The Shielding Integral Benchmark Archive Database (SINBAD) is an internationally established set of radiation shielding and dosimetry data related to experiments relevant in reactor shielding, fusion blanket neutronics and accelerator shielding. In addition to the characterisation of the radiation source, it describes shielding materials and instrumentation and the relevant detectors. The experimental results, be it dose or reaction rates, or unfolded spectra, are presented in tabular ASCII form that can easily be exported to different computer environments for further use. Most sets in SINBAD also contain the computer model used for the interpretation of the experiment and, where available, results from uncertainty analysis. This is an international effort between the Organization for Economic Cooperation and Development, Nuclear Energy Agency Data Bank (http://www.nea.fr/html/databank/) (OECD/NEA Data Bank) and Oak Ridge National Laboratory, Radiation Safety Information Computational Center (http://www-rsicc.ornl.gov/rsic.html) (ORNL/RSICC). Cooperation from many organisations, authors and benchmark analysts have helped SINBAD become a 'living database'--one which involves continuous information updates, preservation and additions of nuclear benchmarks in the areas of fusion, fission and accelerator science and engineering. This paper focuses on the increased comprehensiveness of experiments that have been carried out in recent years and the validation of computer code and cross section library using these experiments.

Database Management Systems↗
PubMed2000

Predicting exposure conditions that facilitate the potentiation of noise-induced hearing loss by carbon monoxide.

Hearing loss is the most common occupational disease in the United States, with noise serving as the presumed causative agent in most instances. This investigation characterizes the exposure conditions that facilitate the potentiation of noise-induced hearing loss (NIHL) by carbon monoxide (CO). Auditory function was compared in rats exposed 4 weeks earlier to noise alone, CO alone, combined exposure, and air in the exposure chamber. This interval between exposure and auditory threshold assessment was selected to permit recovery of temporary threshold shifts. The compound action potential (CAP) threshold evoked by pure tone stimuli was used as a measure of auditory sensitivity. The no adverse effect level (NOAEL) with respect to potentiation of NIHL was found to be 300 ppm CO. Potentiation of NIHL by CO increases linearly as CO concentration increases between 500 -1500 ppm. Benchmark dose software (version 1. 1B) published by the U.S. EPA National Center for Environmental Assessment was employed to determine a benchmark concentration of CO that produced either a 5-dB potentiation of NIHL or an increase in auditory threshold equivalent to 10% of the effect of noise alone. The lower bound for these benchmark concentrations were 320 and 194 ppm CO, respectively. Unlike CO dose, the relationship between noise severity and potentiation of NIHL by CO shows a nonlinear relationship. The greatest potentiation was observed at moderate noise exposures (100 dB, 2-h, octave band-limited noise, or OBN) that produce limited permanent threshold shifts. Repeated exposures to 95-dB noise for 2-h periods in combination with 1200 ppm CO also yielded potentiation of NIHL, though such effects were not observed following a single combined exposure. These results underscore the potential risk of hearing loss from combined exposure to noise and CO, and the risks associated with repeated exposure.

Action Potentials↗
PubMed2000

Wall Street's assessment of plastic surgery--related technology: a clinical and financial analysis.

Many plastic surgeons develop technologies that are manufactured by Wall Street-financed companies. Others participate in the stock market as investors. This study examines the bioengineered skin industry to determine whether it integrates clinical and financial information as Wall Street tenets would predict, and to see whether the financial performance of these companies provides any lessons for practicing plastic surgeons. In efficient markets, the assumptions on which independent financial analysts base their company sales and earnings projections are clinically reasonable, the volatility of a company's stock price does not irrationally differ from that of its industry sector, and the buy/sell recommendations of analysts are roughly congruent. For the companies in this study, these key financial parameters were compared with a benchmark index of 69 biotech companies of similar age and annual revenues (Student's t test). Five bioengineered skin companies were included in the study. Analysts estimated that each company would sell its product to between 24 and 45 percent of its target clinical population. The average stock price volatility was significantly higher for study companies than for those in the benchmark index (p < 0.05). Similarly, buy/sell recommendations of analysts for the study companies were significantly less congruent than those for the benchmark companies (p < 0.05). These results indicate clinically unrealistic projections for market penetration, significantly high price volatility, and significantly high discordance among professional analysts. In all cases, the market is inefficient-an unusual finding on Wall Street. A likely explanation for this market failure is a cycle of poor clinical correlation when assigning sales projections, which in turn leads to price volatility and discordance of buy/sell recommendations. This study's findings have implications for plastic surgeons who develop new technology or who participate in the equities markets as investors. Plastic surgeons who develop new medical devices or technology cannot universally depend on the market to drive clinically reasonable financial performance. Although inflated sales estimates have benefits in the short term, failure to meet projections exacts severe financial penalties. Plastic surgeons who invest in the stock market, because of their unique clinical experience, may sometimes be in the position to evaluate new technologies and companies better than Wall Street experts. Well-timed trades that use this expertise can result in opportunities for profit.

Biomedical Engineering↗
PubMed2005

An international otology database.

There are many reports in the literature on the outcome of ear surgery. However, it is difficult to make comparisons from these reports because of a lack of uniformity in the way in which outcomes are reported. At the same time, the general public expects doctors to provide them with information regarding the benefits and risks of surgery. Many ear surgeons will not be able to provide this information because few have an up-to-date database. A group of otologists in Europe has attempted to establish a common otology database. The aim is to create a framework whereby surgeons in Europe can contribute to a common ear database using a web-based system. The proposed methodology involves two levels of data entry. Level 1 is a minimal database where the main outcomes are included. Level 2 is a comprehensive database where detailed information on pathologies, risk factors, and surgical procedures is also recorded. As both databases share the same core data, clinicians using database 1 can still compare their outcomes with those using database 2. Clinicians will be able to input and retrieve ear data onto or from the web-based database. There has already been an international consensus on the content of the common ear database. The website address is www.ear-audit.net. The international project has now been in operation for more than 6 months, and Ear UK has already given its endorsement. The founder members include 27 otologists from 12 European countries. The cumulative outcome from this group of surgeons will be used as the benchmark for the purpose of comparative audit. These benchmark sites will be subjected to regular external validation. The web-based system is interactive and gives instant feedback to individual surgeons who wish to compare their results to the benchmark. Therefore, it can be a good learning tool for trainees or less experienced surgeons. The common data entry system also provides an opportunity for clinicians to collaborate in clinical research.

Databases, Factual↗
PubMed2002

Indicators of AEI applied to the Delaware Estuary.

We evaluated the impacts of entrainment and impingement at the Salem Generating Station on fish populations and communities in the Delaware Estuary. In the absence of an agreed-upon regulatory definition of "adverse environmental impact" (AEI), we developed three independent benchmarks of AEI based on observed or predicted changes that could threaten the sustainability of a population or the integrity of a community. Our benchmarks of AEI included: (1) disruption of the balanced indigenous community of fish in the vicinity of Salem (the "BIC" analysis); (2) a continued downward trend in the abundance of one or more susceptible fish species (the "Trends" analysis); and (3) occurrence of entrainment/impingement mortality sufficient, in combination with fishing mortality, to jeopardize the future sustainability of one or more populations (the "Stock Jeopardy" analysis). The BIC analysis utilized nearly 30 years of species presence/absence data collected in the immediate vicinity of Salem. The Trends analysis examined three independent data sets that document trends in the abundance of juvenile fish throughout the estuary over the past 20 years. The Stock Jeopardy analysis used two different assessment models to quantify potential long-term impacts of entrainment and impingement on susceptible fish populations. For one of these models, the compensatory capacities of the modeled species were quantified through meta-analysis of spawner-recruit data available for several hundred fish stocks. All three analyses indicated that the fish populations and communities of the Delaware Estuary are healthy and show no evidence of an adverse impact due to Salem. Although the specific models and analyses used at Salem are not applicable to every facility, we believe that a weight of evidence approach that evaluates multiple benchmarks of AEI using both retrospective and predictive methods is the best approach for assessing entrainment and impingement impacts at existing facilities.

Animals↗
PubMed2005

A study into the plaque-inhibitory activity of experimental toothpaste formulations containing antimicrobial agents.

BACKGROUND/AIMS: The use of specific antimicrobial agents in toothpastes may help reduce plaque and gingivitis. There would also appear to be some value in formulating products that contain combinations of such agents that may potentiate any activity present. The aims of this exploratory and pragmatic study were twofold: (1) Exploratory: to compare the effects on plaque re-growth of two zinc citrate/triclosan formulations, one of which contained bromochlorophene and hence demonstrate any additional beneficial effects produced by the addition of the phenol. (2) Pragmatic: to assess whether both pastes were significantly better than a benchmark control, proprietary fluoride toothpaste at inhibiting plaque formation. METHODS: Following an initial prophylaxis to remove all plaque and calculus, toothpaste slurry rinses were used over a 96 h period by 24 volunteers, while omitting all other oral hygiene procedures. After 24, 48 and 96 h, plaque was measured by plaque area and by plaque index. For comparative purposes, a conventional commercial fluoride toothpaste rinse was also used as a benchmark control in this triple cross-over double-blind study. RESULTS: With one exception, comparisons between the three pastes failed to show any significant differences in plaque accumulation at 96 h whether assessed by plaque index or area. At this time period, significantly more plaque was seen with the zinc citrate paste without bromochlorophene, compared with that of the control paste. CONCLUSIONS: The findings from this study failed to demonstrate a plaque-inhibitory action from the two novel formulations beyond that of a conventional benchmark toothpaste, although overall levels of plaque formed by the volunteers, especially on the control paste were generally lower than in previous studies. Nevertheless, it remains to be determined whether the test formulations could exert a direct anti-inflammatory action against gingivitis by way of the triclosan delivery system. Neither test formulation was subsequently marketed.

Adult↗
PubMed2005

Trauma bypass in south-eastern Sydney: an 8-year review.

OBJECTIVE: The New South Wales (NSW) Health Department and the Ambulance Service of NSW introduced a trauma bypass system in Sydney on 29 March 1992. This study aims to review the outcomes of trauma bypass patients brought to St George Hospital, a major trauma service in south-eastern Sydney, and to assess the performance of the current prehospital trauma triage protocol. METHODS: The St George Hospital Department of Trauma Services prospectively collected data on all trauma bypass patients for the 8-year period from 29 March 1992 to 29 March 2000. RESULTS: A total of 1990 patients were brought to hospital on trauma bypass. The average age was 32 years, 70% were men and 66% were from road traffic accidents. The positive predictive value of the prehospital triage tool for serious injury (Injury Severity Score [ISS] > 15) was 18.6% (95% CI 16.9-20.4). This is well below the benchmark previously established by the NSW Health Department Trauma System Advisory Committee. For all trauma bypass patients, 33.8% (95% CI 31.7-35.9) were discharged home from the ED. The overall death rate was 2.5% (95% CI 1.9-3.3). CONCLUSIONS: According to the proposed benchmark, current prehospital trauma triage guidelines are underperforming. This suggests that a review of the benchmarks of current local trauma systems and of the trauma triage tool is required.

Accidental Falls↗
PubMed2000

Validation of a precision radiochromic film dosimetry system for quantitative two-dimensional imaging of acute exposure dose distributions.

We present an evaluation of the precision and accuracy of image-based radiochromic film (RCF) dosimetry performed using a commercial RCF product (Gafchromic MD-55-2, Nuclear Associates, Inc.) and a commercial high-spatial resolution (100 microm pixel size) He-Ne scanning-laser film-digitizer (Personal Densitometer, Molecular Dynamics, Inc.) as an optical density (OD) imaging system. The precision and accuracy of this dosimetry system are evaluated by performing RCF imaging dosimetry in well characterized conformal external beam and brachytherapy high dose-rate (HDR) radiation fields. Benchmarking of image-based RCF dosimetry is necessary due to many potential errors inherent to RCF dosimetry including: a temperature-dependent time evolution of RCF dose response; nonuniform response of RCF; and optical-polarization artifacts. In addition, laser-densitometer imaging artifacts can produce systematic OD measurement errors as large as 35% in the presence of high OD gradients. We present a RCF exposure and readout protocol that was developed for the accurate dosimetry of high dose rate (HDR) radiation sources. This protocol follows and expands upon the guidelines set forth by the American Association of Physicists in Medicine (AAPM) Task Group 55 report. Particular attention is focused on the OD imaging system, a scanning-laser film digitizer, modified to eliminate OD artifacts that were not addressed in the AAPM Task Group 55 report. RCF precision using this technique was evaluated with films given uniform 6 MV x-ray doses between 1 and 200 Gy. RCF absolute dose accuracy using this technique was evaluated by comparing RCF measurements to small volume ionization chamber measurements for conformal external-beam sources and an experimentally validated Monte Carlo photon-transport simulation code for a 192Ir brachytherapy source. Pixel-to-pixel standard deviations of uniformly irradiated films were less than 1% for doses between 10 and 150 Gy; between 1% and 5% for lower doses down to 1 Gy and 1% and 1.5% for higher doses up to 200 Gy. Pixel averaging to form 200-800 microm pixels reduces these standard deviations by a factor of 2 to 5. Comparisons of absolute dose show agreement within 1.5%-4% of dose benchmarks, consistent with a highly accurate dosimeter limited by its observed precision and the precision of the dose standards to which it is compared. These results provide a comprehensive benchmarking of RCF, enabling its use in the commissioning of novel HDR therapy sources.

Brachytherapy↗
PubMed2002

Coupled-mode solutions in generalized ocean environments.

This paper presents the application of the differential equation approach to solving the second-order coupled-mode equations in inhomogeneous ocean environments. The model incorporates sound velocity profile points to construct depth-dependent, piecewise linear, ocean and bottom environments along a range grid. Modal solutions are evaluated in terms of Airy functions. The formalism to evaluate analytically the mode-coupling coefficients is presented. Comparisons to conventional expressions of the coefficients are made. The integro-differential form of the coupled equations is solved using an approach developed in nuclear theory that incorporates the Lanczos method [Knobles, J. Acoust. Soc. Am. 96, 1741-1747 (1994)]. Demonstration of the practicality of this approach is made by applying the results in actual calculations with realistic ocean environments. The formalism to evaluate analytically the mode-coupling coefficients is presented. Several benchmark examples were examined in order to validate the model and are discussed, including propagation over a hill, benchmark wedge problems, and a range-varying sound speed profile benchmark. The importance of this model is also demonstrated by the physical insight gained in having a coupled-mode approach to solving range-dependent problems.

Journal Article↗
PubMed2004

Survival after surgery or therapeutic catheterisation for congenital heart disease in children in the United Kingdom: analysis of the central cardiac audit database for 2000-1.

OBJECTIVES: To analyse simple national statistics and survival data collected in the central cardiac audit database after treatment for congenital heart disease and to provide long term comparative statistics for each contributing centre. DESIGN: Prospective, longitudinal, observational, national cohort survival study. SETTING: UK central cardiac audit database. MAIN OUTCOME MEASURES: Survival at 30 days and one year after treatment in the year April 2000-March 2001, assessed by using both volunteered life status and independently validated life status through the Office for National Statistics, using the patient's unique NHS number, or the general register offices of Scotland and Northern Ireland. Institutional results following a group of six benchmark operations and three benchmark catheterisation procedures. RESULTS: Since April 2000 data have been received from all 13 UK tertiary centres performing cardiac surgery or therapeutic cardiac catheterisation in children with congenital heart disease. Altogether 3666 surgical procedures and 1828 therapeutic catheterisations were performed. Central tracking of mortality identified 469 deaths, 194 occurring within 30 days and 275 later. Forty two of the 194 deaths within 30 days were detected by central tracking but not by volunteered data. For surgery overall, survival at 30 days was 94.9%, falling to 91.2% at one year; this effect was most marked for infants. For therapeutic catheterisation survival at 30 days was 99.1%, falling to 98.1% at one year. Survival of individual centres or individual operators did not differ from the national average after benchmark procedures. CONCLUSIONS: Independent data validation is essential for accurate survival analysis. One year survival gives a more realistic view of outcome than traditional perioperative mortality. Currently no detectable difference exists in survival between any of the 13 UK tertiary congenital heart disease centres, but confidence intervals for small centres are wide, limiting our power to detect underperformance from analysis of a single year's data. Appropriately resourced, focused national audit is capable of accurate data collection on which nationwide, long term quality control can be based.

Cardiac Catheterization↗
PubMed1995

Theoretical results for a class of neural networks.

The ability to derive minimal network architectures for neural networks has been at the center of attention for several years now. To this date numerous algorithms have been proposed to automatically construct networks. Unfortunately, these algorithms lack a fundamental theoretical analysis of their capabilities and only empirical evaluations on a few selected benchmark problems exist. Some theoretical results have been provided for small classes of well-known benchmark problems such as parity- and encoder-functions, but these are of little value due to their restrictiveness. In this work we describe a general class of 2-layer networks with 2 hidden units capable of representing a large set of problems. The cardinality of this class grows exponentially with regard to the inputs N. Furthermore, we outline a simple algorithm that allows us to determine, if any function (problem) is a member of this class. The class considered in this paper includes the benchmark problems parity and symmetry. Finally, we expand this class to include an even larger set of functions and point out several interesting properties it exhibits.

Models, Theoretical↗
PubMed2002

Detection of intimate-partner violence among members of a managed behavioral health organization.

This column describes a creative effort on the part of a national behavioral managed care plan to identify cases of domestic violence in its covered population. Although the screening efforts by intake counselors and clinicians yielded a lower rate of domestic violence than would have been expected, the effort established a baseline or benchmark from which the organization can now work. Clinical benchmarks are critical to developing an understanding of best practices. In this case, the benchmark finding initiated a best practices process through which this organization--and, hopefully, others--will continue to create better methods for identifying and helping people who experience domestic violence.

Adult↗