REVISE: A Tool for Measuring and Mitigating Bias in Visual Datasets

International Journal of Computer Vision - Machine learning models are known to perpetuate and even amplify the biases present in the data. However, these data biases frequently do not become...     

Analytical‐Based Methodologies for Examining the In Vitro Absorption,Distribution, Metabolism,and Elimination (ADME) of Silver Nanoparticles

The clinical applications of silver nanoparticles (AgNPs) remain limited due to the lack of well‐established methodologies for studying their nanokinetics. Hereby, the primary goal is to adapt a suite of analytical‐based methodologies for examining the in vitro absorption, distribution, metabolism, and elimination of AgNPs. Vero 76 and HEK 293 cells are exposed to ≈10‐nm spherical AgNPs⁺ and AgNPs^? at relevant concentrations (0–300 µg mL^?1) and times (4–48 h). Absorption: Inductively coupled plasma optical emission spectroscopy (ICP‐OES) demonstrates that the two AgNP formulations are not bioequivalent. For example, different bioavailabilities (C _maximum < 20.7 ± 4% and 6.82 ± 0.4%), absorption times (T _maximum > 48 and ≈24 h), and absorption rate laws (first‐ and zeroth‐order at 300 µg mL^?1) are determined in Vero 76 for AgNPs⁺ and AgNPs^?, respectively. Distribution: Raman and CytoViva hyperspectral imaging show different cellular localizations for AgNPs⁺ and AgNPs^?. Metabolism: Cloud point extraction (CPE)‐tangential flow filtration (TFF) reveal that ≤ 11% ± 4% of the administered, sublethal AgNPs release Ag⁺ and contribute to the observed cytotoxicity. Elimination: ICP‐OES‐CPE suggests that AgNPs are cleared via exocytosis.     

Feedstock powder processing research needs for additive manufacturing development

Additive manufacturing (AM) promises to redesign traditional manufacturing by enabling the ultimate in agility for rapid component design changes in commercial products and for fabricating complex integrated parts. By significantly increasing quality and yield of metallic alloy powders, the pace for design, development, and deployment of the most promising AM approaches can be greatly accelerated, resulting in rapid commercialization of these advanced manufacturing methods. By successful completion of a critical suite of processing research tasks that are intended to greatly enhance gas atomized powder quality and the precision and efficiency of powder production, researchers can help promote continued rapid growth of AM. Other powder-based or spray-based advanced manufacturing methods could also benefit from these research outcomes, promoting the next wave of sustainable manufacturing technologies for conventional and advanced materials.     

Whole body vibration exposures in forklift operators: comparison of a mechanical and air suspension seat

Using a repeated measures design, this study compared differences in whole body vibration (WBV) exposures when 12 forklift operators drove the same forklift with a mechanical suspension and an air suspension seat. A portable PDA-based WBV data acquisition system collected and analysed time-weighted and raw WBV data per ISO 2631-1 and 2631-5 WBV measurement standards. Tri-axial measurements of weighted vibration (A(w)), crest factor, vibration dose values, time-weighted average-peak, raw (+) peak, raw (-) peak and static compression dose (S(ed)) were compared between seats. There were significant differences in z-axis WBV exposures with the air suspension seat, yielding lower WBV exposures. In addition, there were differences between seats in how they attenuated WBV exposures based on the driver's weight. In the mechanical suspension seat, WBV exposures were weight-dependent, with lighter drivers having higher WBV exposures, whereas with the air suspension seat, the same trends were not as prevalent. STATEMENT OF RELEVANCE: This study contributes to the understanding of how different seat suspensions can influence WBV transmission and how some components of vibration transmission are dependent on the weight of the driver. Additional systematic studies are needed to quantify how various factors can influence WBV exposures.     

Persistent ocean monitoring with underwater gliders: Adapting sampling resolution

Ocean processes are dynamic and complex and occur on multiple spatial and temporal scales. To obtain a synoptic view of such processes, ocean scientists collect data over long time periods. Historically, measurements were continually provided by fixed sensors, e.g., moorings, or gathered from ships. Recently, an increase in the utilization of autonomous underwater vehicles has enabled a more dynamic data acquisition approach. However, we still do not utilize the full capabilities of these vehicles. Here we present algorithms that produce persistent monitoring missions for underwater vehicles by balancing path following accuracy and sampling resolution for a given region of interest, which addresses a pressing need among ocean scientists to efficiently and effectively collect high‐value data. More specifically, this paper proposes a path planning algorithm and a speed control algorithm for underwater gliders, which together give informative trajectories for the glider to persistently monitor a patch of ocean. We optimize a cost function that blends two competing factors: maximize the information value along the path while minimizing deviation from the planned path due to ocean currents. Speed is controlled along the planned path by adjusting the pitch angle of the underwater glider, so that higher resolution samples are collected in areas of higher information value. The resulting paths are closed circuits that can be repeatedly traversed to collect long‐term ocean data in dynamic environments. The algorithms were tested during sea trials on an underwater glider operating off the coast of southern California, as well as in Monterey Bay, California. The experimental results show improvements in both data resolution and path reliability compared to previously executed sampling paths used in the respective regions. © 2011 Wiley Periodicals, Inc.     

The train driver recovery problem—A set partitioning based model and solution method

The need to recover a train driver schedule occurs during major disruptions in the daily railway operations. Based on data from the Danish passenger railway operator DSB S-tog A/S, a solution method to the train driver recovery problem (TDRP) is developed. The TDRP is formulated as a set partitioning problem. We define a disruption neighbourhood by identifying a small set of drivers and train tasks directly affected by the disruption. Based on the disruption neighbourhood, the TDRP model is formed and solved. If the TDRP solution provides a feasible recovery for the drivers within the disruption neighbourhood, we consider that the problem is solved. However, if a feasible solution is not found, the disruption neighbourhood is expanded by adding further drivers or increasing the recovery time period. Fractional solutions to the LP relaxation of the TDRP are resolved with a constraint branching strategy using the depth-first search of the Branch & Bound tree. The LP relaxation of the TDRP possesses strong integer properties. We present test scenarios generated from the historical real-life operations data of DSB S-tog A/S. The numerical results show that all but one tested instances produce integer solutions to the LP relaxation of the TDRP and solutions are found within a few seconds.     

Polynomial regression under arbitrary product distributions

In recent work, Kalai, Klivans, Mansour, and Servedio (2005) studied a variant of the “Low-Degree (Fourier) Algorithm” for learning under the uniform probability distribution on {0,1} ⁿ . They showed that the L ₁ polynomial regression algorithm yields agnostic (tolerant to arbitrary noise) learning algorithms with respect to the class of threshold functions—under certain restricted instance distributions, including uniform on {0,1} ⁿ and Gaussian on ? ⁿ . In this work we show how all learning results based on the Low-Degree Algorithm can be generalized to give almost identical agnostic guarantees under arbitrary product distributions on instance spaces X ₁×???×X _n . We also extend these results to learning under mixtures of product distributions. The main technical innovation is the use of (Hoeffding) orthogonal decomposition and the extension of the “noise sensitivity method” to arbitrary product spaces. In particular, we give a very simple proof that threshold functions over arbitrary product spaces have δ-noise sensitivity $O(\sqrt{\delta})$ , resolving an open problem suggested by Peres (2004).     

Using LiDAR and quickbird data to model plant production and quantify uncertainties associated with wetland detection and land cover generalizations

Spatiotemporal data from satellite remote sensing and surface meteorology networks have made it possible to continuously monitor global plant production, and to identify global trends associated with land cover/use and climate change. Gross primary production (GPP) and net primary production (NPP) are routinely derived from the Moderate Resolution Imaging Spectroradiometer (MODIS) onboard satellites Terra and Aqua, and estimates generally agree with independent measurements at validation sites across the globe. However, the accuracy of GPP and NPP estimates in some regions may be limited by the quality of model input variables and heterogeneity at fine spatial scales. We developed new methods for deriving model inputs (i.e., land cover, leaf area, and photosynthetically active radiation absorbed by plant canopies) from airborne laser altimetry (LiDAR) and Quickbird multispectral data at resolutions ranging from about 30 m to 1 km. In addition, LiDAR-derived biomass was used as a means for computing carbon-use efficiency. Spatial variables were used with temporal data from ground-based monitoring stations to compute a six-year GPP and NPP time series for a 3600 ha study site in the Great Lakes region of North America. Model results compared favorably with independent observations from a 400 m flux tower and a process-based ecosystem model (BIOME-BGC), but only after removing vapor pressure deficit as a constraint on photosynthesis from the MODIS global algorithm. Fine-resolution inputs captured more of the spatial variability, but estimates were similar to coarse-resolution data when integrated across the entire landscape. Failure to account for wetlands had little impact on landscape-scale estimates, because vegetation structure, composition, and conversion efficiencies were similar to upland plant communities. Plant productivity estimates were noticeably improved using LiDAR-derived variables, while uncertainties associated with land cover generalizations and wetlands in this largely forested landscape were considered less important.