Search NASA⌕ Search

SEARCH · Search NASA

Results for “small sample size”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

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

At least 1,009 records · Page 56

Deep active learning for classifying cancer pathology reports

Abstract Background Automated text classification has many important applications in the clinical setting; however, obtaining labelled data for training machine learning and deep learning models is often difficult and expensive. Active learning techniques may mitigate this challenge by reducing the amount of labelled data required to effectively train a model. In this study, we analyze the effectiveness of 11 active learning algorithms on classifying subsite and histology from cancer pathology reports using a Convolutional Neural Network as the text classification model. Results We compare the performance of each active learning strategy using two differently sized datasets and two different classification tasks. Our results show that on all tasks and dataset sizes, all active learning strategies except diversity-sampling strategies outperformed random sampling, i.e., no active learning. On our large dataset (15K initial labelled samples, adding 15K additional labelled samples each iteration of active learning), there was no clear winner between the different active learning strategies. On our small dataset (1K initial labelled samples, adding 1K additional labelled samples each iteration of active learning), marginal and ratio uncertainty sampling performed better than all other active learning techniques. We found that compared to random sampling, active learning strongly helps performance on rare classes by focusing on underrepresented classes. Conclusions Active learning can save annotation cost by helping human annotators efficiently and intelligently select which samples to label. Our results show that a dataset constructed using effective active learning techniques requires less than half the amount of labelled data to achieve the same performance as a dataset constructed using random sampling.

59 BASIC BIOLOGICAL SCIENCES↗

Pu Anion Exchange Process Intensification

This research is focused on improving the efficiency of the anion exchange process for purifying plutonium. While initially focused on plutonium, the technology could also be applied to other ion-exchange processes. Work in FY17 focused on the improvement and optimization of porous foam columns that were initially developed in FY16. These foam columns were surface functionalized with poly(4-vinylpyridine) (PVP) to provide the Pu specific anion-exchange sites. Two different polymerization methods were explored for maximizing the surface functionalization with the PVP. The open-celled polymeric foams have large open pores and large surface areas available for sorption. The fluid passes through the large open pores of this material, allowing convection to be the dominant mechanism by which mass transport takes place. These materials generally have very low densities, open-celled structures with high cell interconnectivity, small cell sizes, uniform cell size distributions, and high structural integrity. These porous foam columns provide advantages over the typical porous resin beads by eliminating the slow diffusion through resin beads, making the anion-exchange sites easily accessible on the foam surfaces. The best performing samples exceeded the Pu capacity of the commercially available resin, and also offered the advantage of sharper elution profiles, resulting in a more concentrated product, with less loss of material to the dilute heads and tails cuts. An alternate approach to improving the efficiency of this process was also explored through the development of a microchannel array system for performing the anion exchange.

38 RADIATION CHEMISTRY, RADIOCHEMISTRY, AND NUCLEA↗

Redshift differences of galaxies in nearby groups

It is reported that galaxies in nearby groups exhibit anomalous nonvelocity redshifts. In this discussion, (1) four classes of nearby groups of galacies are analyzed, and no significant nonvelocity redshift effect is found; and (2) it is pointed out that transverse velocities (i.e., velocities transverse to the line of sight of the main galaxy, or center of mass) contribute components to the redshift measurements of companion galaxies. The redshifts of galaxies in nearby groups of appreciable angular size are considerably affected by these velocity projection effects. The transverse velocity contributions average out in rich, isotropic groups, and also in large samples of irregular groups of low membership, as in the four classes referred to in (1), but can introduce apparent discrepancies in small samples (as studied by Arp) of nearby groups of low membership.

Harrison, E. R.↗

Viral coefficient and hidden mass in the galaxy groups

The purpose is the verification of the virial mass estimations for small galaxy groups. The dynamical evolution of triple and quintuple galaxies was studied by the numerical simulations. The dependence of the virial coefficient k(t) versus time was derived. Initial k(O) = O. The function k(t) has some strong oscillations from 0.02 to 0.99. Generally, these oscillations are quasiperiodical ones. Such a behavior of k(t) is caused by formation in a system of close isolated temporary double subsystems. A strong correlation between the virial coefficient and the least mutual distance in the system is observed. Such wide oscillations may add into the estimation of virial mass of the galaxy groups an uncertainty of more than one order. An additional uncertainty is introduced by the projection effect. This uncertainty for the individual estimations of the masses approach three orders. Thus any individual estimation of the virial mass is impossible for small galaxy groups. Some possibility of statistical estimation (median or average) of the total mass, including a hidden mass, is shown for the homogeneous samples. The authors propose a method for these estimations based on a comparison of the medians of dynamical parameters (a mean size in projection and a dispersion of relative radial velocities) for the simulated and observed ensembles of the galaxy groups. This method has been applied to a sample of 46 probably physical triplets of galaxies. The probable median of the hidden mass in a volume of the triplet is about 4 M, where M is the total mass of visible matter.

Anosova, Joanna P.↗

Calculation of Heavy Ion Inactivation and Mutation Rates in Radial Dose Model of Track Structure

In the track structure model, the inactivation cross section is found by summing an inactivation probability over all impact parameters from the ion to the sensitive sites within the cell nucleus. The inactivation probability is evaluated by using the dose response of the system to gamma rays and the radial dose of the ions and may be equal to unity at small impact parameters. We apply the track structure model to recent data with heavy ion beams irradiating biological samples of E. Coli, B. Subtilis spores, and Chinese hamster (V79) cells. Heavy ions have observed cross sections for inactivation that approach and sometimes exceed the geometric size of the cell nucleus. We show how the effects of inactivation may be taken into account in the evaluation of the mutation cross sections in the track structure model through correlation of sites for gene mutation and cell inactivation. The model is fit to available data for HPRT (hypoxanthine guanine phosphoribosyl transferase) mutations in V79 cells, and good agreement is found. Calculations show the high probability for mutation by relativistic ions due to the radial extension of ions track from delta rays. The effects of inactivation on mutation rates make it very unlikely that a single parameter such as LET (linear energy transfer) can be used to specify radiation quality for heavy ion bombardment.

Cucinotta, Francis A.↗

Large Scale Experiments on Spacecraft Fire Safety

Full scale fire testing complemented by computer modelling has provided significant know how about the risk, prevention and suppression of fire in terrestrial systems (cars, ships, planes, buildings, mines, and tunnels). In comparison, no such testing has been carried out for manned spacecraft due to the complexity, cost and risk associated with operating a long duration fire safety experiment of a relevant size in microgravity. Therefore, there is currently a gap in knowledge of fire behaviour in spacecraft. The entire body of low-gravity fire research has either been conducted in short duration ground-based microgravity facilities or has been limited to very small fuel samples. Still, the work conducted to date has shown that fire behaviour in low-gravity is very different from that in normal-gravity, with differences observed for flammability limits, ignition delay, flame spread behaviour, flame colour and flame structure. As a result, the prediction of the behaviour of fires in reduced gravity is at present not validated. To address this gap in knowledge, a collaborative international project, Spacecraft Fire Safety, has been established with its cornerstone being the development of an experiment (Fire Safety 1) to be conducted on an ISS resupply vehicle, such as the Automated Transfer Vehicle (ATV) or Orbital Cygnus after it leaves the ISS and before it enters the atmosphere. A computer modelling effort will complement the experimental effort. Although the experiment will need to meet rigorous safety requirements to ensure the carrier vehicle does not sustain damage, the absence of a crew removes the need for strict containment of combustion products. This will facilitate the possibility of examining fire behaviour on a scale that is relevant to spacecraft fire safety and will provide unique data for fire model validation. This unprecedented opportunity will expand the understanding of the fundamentals of fire behaviour in spacecraft. The experiment is being developed by an international topical team that is collaboratively defining the experiment requirements and performing supporting analysis, experimentation and technology development. This paper presents the objectives, status and concept of this project.

Urban, David L.↗

Large Scale Experiments on Spacecraft Fire Safety

Full scale fire testing complemented by computer modelling has provided significant knowhow about the risk, prevention and suppression of fire in terrestrial systems (cars, ships, planes, buildings, mines, and tunnels). In comparison, no such testing has been carried out for manned spacecraft due to the complexity, cost and risk associated with operating a long duration fire safety experiment of a relevant size in microgravity. Therefore, there is currently a gap in knowledge of fire behaviour in spacecraft. The entire body of low-gravity fire research has either been conducted in short duration ground-based microgravity facilities or has been limited to very small fuel samples. Still, the work conducted to date has shown that fire behaviour in low-gravity is very different from that in normal gravity, with differences observed for flammability limits, ignition delay, flame spread behaviour, flame colour and flame structure. As a result, the prediction of the behaviour of fires in reduced gravity is at present not validated. To address this gap in knowledge, a collaborative international project, Spacecraft Fire Safety, has been established with its cornerstone being the development of an experiment (Fire Safety 1) to be conducted on an ISS resupply vehicle, such as the Automated Transfer Vehicle (ATV) or Orbital Cygnus after it leaves the ISS and before it enters the atmosphere. A computer modelling effort will complement the experimental effort. Although the experiment will need to meet rigorous safety requirements to ensure the carrier vehicle does not sustain damage, the absence of a crew removes the need for strict containment of combustion products. This will facilitate the possibility of examining fire behaviour on a scale that is relevant to spacecraft fire safety and will provide unique data for fire model validation. This unprecedented opportunity will expand the understanding of the fundamentals of fire behaviour in spacecraft. The experiment is being developed by an international topical team that is collaboratively defining the experiment requirements and performing supporting analysis, experimentation and technology development. This paper presents the objectives, status and concept of this project.

Urban, David↗

An estimate of field size distributions for selected sites in the major grain producing countries

The field size distributions for the major grain producing countries of the World were estimated. LANDSAT-1 and 2 images were evaluated for two areas each in the United States, People's Republic of China, and the USSR. One scene each was evaluated for France, Canada, and India. Grid sampling was done for representative sub-samples of each image, measuring the long and short axes of each field; area was then calculated. Each of the resulting data sets was computer analyzed for their frequency distributions. Nearly all frequency distributions were highly peaked and skewed (shifted) towards small values, approaching that of either a Poisson or log-normal distribution. The data were normalized by a log transformation, creating a Gaussian distribution which has moments readily interpretable and useful for estimating the total population of fields. Resultant predictors of the field size estimates are discussed.

Podwysocki, M. H.↗

SUPERDENSE GALAXIES AND THE MASS-SIZE RELATION AT LOW REDSHIFT

We search for massive and compact galaxies (superdense galaxies, hereafter SDGs) at z = 0.03-0.11 in the Padova-Millennium Galaxy and Group Catalogue, a spectroscopically complete sample representative of the general field population of the local universe. We find that compact galaxies with radii and mass densities comparable to high-z massive and passive galaxies represent 4.4% of all galaxies with stellar masses above 3 Multiplication-Sign 10{sup 10} M {sub Sun }, yielding a number density of 4.3 Multiplication-Sign 10{sup -4} h {sup 3} Mpc{sup -3}. Most of them are S0s (70%) or ellipticals (23%), are red, and have intermediate-to-old stellar populations, with a median luminosity-weighted age of 5.4 Gyr and a median mass-weighted age of 9.2 Gyr. Their velocity dispersions and dynamical masses are consistent with the small radii and high stellar mass estimates. Comparing with the WINGS sample of cluster galaxies at similar redshifts, the fraction of SDGs is three times smaller in the field than in clusters, and cluster SDGs are on average 4 Gyr older than field SDGs. We confirm the existence of a universal trend of smaller radii for older luminosity-weighted ages at fixed galaxy mass. As a consequence, the median mass-size relation shifts toward smaller radii for galaxies with older stars, but the effect is much more pronounced in clusters than in the field. Our results show that, on top of the well-known dependence of stellar age on galaxy mass, the luminosity-weighted age of galaxies depends on galaxy compactness at fixed mass and, for a fixed mass and radius, on environment. This effect needs to be taken into account in order not to overestimate the evolution of galaxy sizes from high to low z. Our results and hierarchical simulations suggest that a significant fraction of the massive compact galaxies at high z have evolved into compact galaxies in galaxy clusters today. When stellar age and environmental effects are taken into account, the average amount of size evolution of individual galaxies between high and low z is mild, a factor {approx}1.6.

79 ASTRONOMY AND ASTROPHYSICS↗

Effect of spacer layer thickness on magnetic interactions in self-assembled single domain iron nanoparticles

The magnetic characteristics of iron nanoparticles embedded in an alumina thin film matrix have been studied as a function of spacer layer thickness. Alumina as well as iron nanoparticles were deposited in a multilayered geometry using sequential pulsed laser deposition. The role of spacer layer thickness was investigated by making layered thin film composites with three different spacer layer thicknesses 6, 12, and 18 nm with fixed iron particle size of 13 nm. Intralayer magnetic interactions being the same in each sample, the variation in coercivity and saturation magnetization is attributed to thickness dependent interlayer magnetic interactions of three types: exchange, strong dipolar, and weak dipolar. A thin film composite multilayer structure offers a continuously tunable strength of interparticle dipole-dipole interaction and is thus well suited for studies of the influence of interaction on the magnetic properties of small magnetic particle systems.

36 MATERIALS SCIENCE↗

Hydrogen accumulation in and at the perimeter of U–C–N–O inclusions in uranium – A SIMS analysis

We use imaging secondary ion mass spectrometry (SIMS) with a Cameca NanoSIMS 50 to determine the distribution and relative concentration of H in two uranium metal samples: sample1(LANL), as cast with less than 1 wppm H, sample2(Y12), outgassed 6 h in vacuum at 630 °C to remove hydrogen. H ion counts appear almost exclusively associated with ‘carbide’ inclusions, based on H – , C – and O – ion images for uranium surfaces sputter-cleaned in situ with a 16 keV Cs + ion beam. Here, two classes of inclusions are identified: small, micrometer to sub-micrometer inclusions and larger, clearly angular inclusion (≥3 μm). In sample1(LANL) the large inclusions (≥250/mm 2 ) show a low H – /C – ratio inside, and have H – /C – ratios at their perimeters comparable in magnitude to that seen in μm-size inclusions. Small inclusions (~2500/mm 2 ) contain H more uniformly throughout and, averaged over the inclusion, the small inclusions have approximately 50 times higher relative H concentration than the large inclusions. Sample2(Y12) was found to have comparable H – /C – ratios in the large carbides, but no small inclusions were observed. Because of the matrix-sensitivity of SIMS, H/C ratios representing the actual composition of the inclusions cannot be derived from the H – /C – ratios without calibration UC samples with known H content, which are not currently available.

36 MATERIALS SCIENCE↗

Colloidal Three-Dimensional Covalent Organic Frameworks and Their Application as Porous Liquids

Developing solid porous materials into free-flowing liquids with permanent porosity is a promising strategy for overcoming certain limitations of conventional sorbent materials employed in gas storage and separation applications. The ability to control the pore size and chemical functionalities of organic frameworks gives these particular nanoporous materials distinct advantages over other small cage-like molecules or hollow particles that have been developed into porous liquids. Here, we describe the synthesis of a 3D imine-linked colloidal covalent organic framework (COF)-based porous liquid, designed for efficient size-exclusion of solvent, as well as for long-term stability. By tethering ionic liquids to the colloid surface, the colloids can be dried, purified, and resuspended in a variety of solvents without irreversible aggregation typically observed of COF colloids. Colloid size could be controlled between 50 and 400 nm, with surface areas as high as 800 m2/g. The 3D intertwining morphology of the colloids had pore sizes ranging from 5 to 14 Å, allowing them to efficiently sizeexclude bulky ionic liquids. The COF colloids were stable towards flocculation in an ionic liquid for > 1 year. Permanent porosity was confirmed with a combination of 19F NMR measurements and gas sorption techniques. CO2 and CH4 uptake in these porous liquids increased more than 10 and 20-fold, respectively, over non-porous, neat ionic liquid control samples. The work not only advances the state of COF-based colloid science but also represents a practical advance towards developing more robust, tunable framework-based porous liquid materials for a host of gas storage and separation applications.

Mow, Rachel E.↗

A Quantitative Molecular Atlas for Interactions Between Lignin and Cellulose

Lignin and cellulose are two of the most abundant polymers on Earth, and are found in close proximity within plant cell walls. Despite their ubiquity, relatively little is known quantitatively about their interactions within plants, and by extension how their interaction may affect industrial biomass utilization. Given the inherent heterogeneity of the lignin polymer and the structural complexity of cellulose, quantitative relationships between given cellulose faces and specific lignin chemistries have been challenging to ascertain experimentally. In this study, we use molecular simulation to examine individual combinations of cellulose binding face and lignin chemistry to build a quantitative lignin-cellulose interaction atlas, including contributions from both specific hydrogen bonds observed in simulation and nonspecific interactions between lignin and cellulose driven by solvent considerations. Over all monomeric and dimeric lignin chemistries tested, the hydrophobic 100 face is the preferred cellulose interaction site. Among the hydrophilic crystalline faces, binding of these molecules is strongest to the 110 cellulose face. The chemical composition of lignin monomers and dimers is also found to modulate the binding affinity, with additional methoxy groups increasing the contact area to the cellulose surface, which we hypothesize may protect native cellulose from degradation. Additional methoxylation and monomer linkages that facilitate planar aromatic ring orientations increase the binding affinity between small lignin-related molecules and cellulose, with the planar linkages between guaiacyl-type monomers leading to higher binding affinities. These trends can be extended to larger lignin polymers, where we quantify the relationships between lignin polymer size and binding affinity through replica exchange umbrella sampling simulations. Rather than molecular weight, the binding surface area between the lignin polymer and the cellulose surface dictates the interaction strength. As a result, based on the findings here, we expect there to be abundant and durable interaction between cellulose and lignin, particularly on the hydrophobic face of cellulose where cellulase enzymes typically bind.

09 BIOMASS FUELS↗

THE ORIGIN OF COLOR GRADIENTS IN EARLY-TYPE SYSTEMS AND THEIR COMPACTNESS AT HIGH-z

In this Letter, we present mean optical+NIR color gradient estimates for 5080 early-type galaxies (ETGs) in the grizY JHK wavebands of the Sloan Digital Sky Survey plus the UKIRT Infrared Deep Sky Survey. The color gradient is estimated as the logarithmic slope of the radial color profile in ETGs. With such a large sample size, we study the variation of the mean color gradient as a function of waveband with unprecedented accuracy. We find that (1) color gradients are mainly due, on average, to a metallicity variation of about -0.4 dex per decade in galaxy radius; and (2) a small, but significant, positive age gradient is present, on average, in ETGs, with the inner stellar population being slightly younger, by {approx}0.1 dex per radial decade, than the outer one. Also, we show that the presence of a positive mean age gradient in ETGs, as found in the present study, implies their effective radius to be smaller at high-z, consistent with observations.

99 GENERAL AND MISCELLANEOUS↗

Variable cascade dynamics and intermittency in the solar wind at 1 AU

In recent studies by ourselves and others of third-moment expressions for the rate of energy cascade in the solar wind, it has been shown that relatively large volumes of data are needed to produce convergent averages. These averages are in good agreement with independently obtained estimates for the average heating rate for a solar wind plasma under those conditions. The unanswered question has been whether the convergence issue is the result of intermittent dynamics or simple measurement uncertainties. In other words, is the difficulty in obtaining a single result that characterizes many similarly prepared samples due to in situ dynamics that create physically real variations or simple statistics? There have been publications showing evidence of intermittent dynamics in the solar wind. Here we show that the third-moment expressions and the computed energy cascade for relatively small samples of data comparable to the correlation length are generally well-formed estimates of the local dynamics. This leads us to conclude that intermittency and not simple measurement uncertainties are responsible for the slow convergence to expected heating rates. We partially characterize the scale size of the intermittency to be comparable to or smaller than the correlation length of the turbulence and we attempt to rephrase the discussion of third moments in terms of intermittent dynamics.

79 ASTRONOMY AND ASTROPHYSICS↗

Benchmarking Microscale Ductility Measurements (Final Report of the Project DE-NE0008799)

Conventional macroscale experimentation is generally considered to be straightforward with few limitations. Conversely, micro/nanoscale experimentation presents numerous challenges in loading device design, sample preparation and handling, as well as accurate understanding of grain size and local texture effects on recorded measurements. Despite these challenges, nanopillar compression, MEMs based micro-tension, and nanoindentation approaches have been able to provide fundamental contributions to the understanding of material behavior at small lengthscales. However, the overarching shortcoming of these micro/nanoscale experimentation approaches, is the inability to directly translate measurements evaluated at the nm and µm length scales (e.g., hardness) to macroscale tensile material behavior (i.e., elastic modulus, yield strength, and ductility). The objectives of the proposed study are, 1) to establish best practices for obtaining tensile microscale ductility measurements, and 2) to validate methodologies to for comparing microscale ductility measurements to macroscale ductility measurements. In order to achieve these objectives, a multi-lengthscale, multi-temperature testing protocol and simulation framework are executed first on copper as a model material to validate the following approach, and second on reactor grade Zircaloy-2. Experiments are conducted on specimens extracted from the same test piece to ensure nominally identical grain size and texture from specimens to specimen. Motivated by the need to isolate the contribution of size-effects on obtained mechanical property measurements, specimens are manufactured with thicknesses at the micro- (1-10 µm), meso- (10-100's µm), and macroscales (sub-sized ASTM E8). In-situ full-field deformation techniques (scanning electron microscopy (SEM) grid methods and optical DIC) are incorporated into testing at each specimen length-scale to capture plasticity localization and evolution. Experimental testing for all specimens is conducted at both room temperature and elevated temperatures to probe the role of thermal activation on plastic deformation accommodation processes. Simulation efforts focus on examining the mechanical behavior of microscale specimens using a finite element approach with explicitly resolved grain morphologies, and an embedded crystal plasticity model. The cost-efficient implementation method allows for the modeling of a statistically significant number of both real (i.e., digital twin) and generated microstructures to obtain an understanding of the interrelationships between specimen microstructure and geometric variables (grain size, texture, specimen geometry, etc.) on microscale mechanical behavior.

36 MATERIALS SCIENCE↗

Flocculation and aggregation in a microgravity environment (FAME)

An experiment to study flocculation phenomena in the constrained microgravity environment of a space shuttle or space station is described. The small size and light weight experiment easily fits in a Spacelab Glovebox. Using an integrated fiber optic dynamic light scattering (DLS) system we obtain high precision particle size measurements from dispersions of colloidal particles within seconds, needs no onboard optical alignment, no index matching fluid, and offers sample mixing and shear melting capabilities to study aggregation (flocculation and coagulation) phenomena under both quiescent and controlled agitation conditions. The experimental system can easily be adapted for other microgravity experiments requiring the use of DLS. Preliminary results of ground-based study are reported.

Ansari, Rafat R.↗

Cardiopulmonary Inflammatory Response to Meteorite Dust Exposures - Implications for Human Health on Earth and Beyond

This year marks the 50th anniversary of Apollo 11, the first time humans set foot on the Moon. The Apollo missions not only help answer questions related to our solar system, they also highlight many hazards associated with human space travel. One major concern is the effect of extraterrestrial dust on astronaut health. In an effort to expand upon previous work indicating lunar dust is respirable and reactive, the authors initiated an extensive study evaluating the role of a particulate’s innate geochemical features (e.g., bulk chemistry, internal composition, morphology, size, and reactivity) in generating adverse toxicological responses in vitro and in vivo. To allow for a broader planetary and geochemical assessment, seven samples were evaluated: six meteorites from either the Moon, Mars, or Asteroid 4 Vesta and a terrestrial basalt analogue. Even with the relatively small geochemical differences (all samples basaltic in nature), significant difference in cardiopulmonary inflammatory markers developed in both single exposure and multiple exposure studies. More specifically: 1) the single exposure studies reveal relationships between toxicity and a meteorite sample’s origin, its pre-ejected state (weathered versus un-weathered), and geochemical features (e.g. bulk iron content) and 2) multiple exposure studies reveal a correlation with particle derived reactive oxygen species (ROS) formation and neutrophil infiltration. Extended human exploration will further increase the probability of inadvertent and repeated exposures to extraterrestrial dusts. This comprehensive dataset allows for not only the toxicological evaluation of extraterrestrial materials but also clarifies important correlations between geochemistry and health. The utilization of an array of extraterrestrial samples from Moon, Mars, and asteroid 4Vesta will enable the development of a geochemical based toxicological hazard model that can be used for: 1) mission planning, 2) rapid risk assessment in cases of unexpected exposures, and 3) evaluation of the efficacy of various in situ techniques in gauging surface dust toxicity. Furthermore, by better understanding the importance of geochemical features on exposure related health outcomes in space, it is possible to better understand of the deleterious nature of dust exposure on Earth.

Harrington, A. D.↗