Search for: All records

Metric magnitude is a measure of the “size” of point clouds with many desirable geometric properties. It has been adapted to various mathematical contexts and recent work suggests that it can enhance machine learning and optimization algorithms. But its usability is limited due to the computational cost when the dataset is large or when the computation must be carried out repeatedly (e.g. in model training). In this paper, we study the magnitude computation problem, and show efficient ways of approximating it. We show that it can be cast as a convex optimization problem, but not as a submodular optimization. The paper describes two new algorithms – an iterative approximation algorithm that converges fast and is accurate, and a subset selection method that makes the computation even faster. It has been previously proposed that magnitude of model sequences generated during stochastic gradient descent is correlated to generalization gap. Extension of this result using our more scalable algorithms shows that longer sequences in fact bear higher correlations. We also describe new applications of magnitude in machine learning – as an effective regularizer for neural network training, and as a novel clustering criterion. 

Abstract Ingestion of microplastics (MP) by suspension‐feeding bivalves has been well‐documented. However, it is unclear whether exposure to MP could damage the stomach and digestive gland (gut) of these animals, causing ramifications for organism and ecosystem health. Here, we show no apparent effects of nylon microfiber (MF) ingestion on the gut microbiome or digestive tissues of the blue mussel,Mytilus edulis. We exposed mussels to two low concentrations (50 and 100 particles/L) of either nylon MF orSpartinaspp. particles (dried, ground marsh grass), ca. 250–500 μm in length, or a no particle control laboratory treatment for 21 days. Results showed that nylon MF, when aged in coarsely filtered seawater, developed a different microbial community thanSpartinaspp. particles and seawater, however, even after exposure to this different community, mussel gut microbial communities resisted disturbance from nylon MF. The microbial communities of experimental mussels clustered together in ordination and were similar in taxonomic composition and measures of alpha diversity. Additionally, there was no evidence of damage to gut tissues after ingestion of nylon MF orSpartinaspp. Post‐ingestive particle processing likely mediated a short gut retention time of these relatively large particles, contributing to the negligible treatment effects. 

Abstract Continent‐scale observations of seismic phenomena have provided multi‐scale constraints of the Earth's interior. Of those analyzed, array‐based observations of slowness vector properties (backazimuth and horizontal slowness) and multipathing have yet to be made on a continental scale. Slowness vector measurements give inferences on mantle heterogeneity properties such as velocity perturbation and velocity gradient strength and quantify their effect on the wavefield. Multipathing is a consequence of waves interacting with strong velocity gradients resulting in two arrivals with different slowness vector properties and times. The mantle structure beneath the contiguous Unites States has been thoroughly analyzed by previous seismic studies and is data‐rich, making it an excellent testing ground to both analyze mantle structure with our approach and compare with other imaging techniques. We apply an automated array‐analysis technique to an SKS data set to create the first continent‐scale data set of multipathing and slowness vector measurements. We analyze the divergence of the slowness vector deviation field to highlight seismically slow and fast regions. Our results resolve several slow mantle anomalies beneath Yellowstone, the Appalachian mountains and fast anomalies throughout the mantle. Many of the anomalies cause multipathing in frequency bands 0.15–0.30 and 0.20–0.40 Hz which suggests velocity transitions over at most 500 km exist. Comparing our observations to synthetics created from tomography models, we find model NA13 (Bedle et al., 2021,https://doi.org/10.1029/2021GC009674) fits our data best but differences still remain. We therefore suggest slowness vector measurements should be used as an additional constraint in tomographic inversions and will lead to better resolved models of the mantle. 

SUMMARY Horizontal slowness vector measurements using array techniques have been used to analyse many Earth phenomena from lower mantle heterogeneity to meteorological event location. While providing observations essential for studying much of the Earth, slowness vector analysis is limited by the necessary and subjective visual inspection of observations. Furthermore, it is challenging to determine the uncertainties caused by limitations of array processing such as array geometry, local structure, noise and their effect on slowness vector measurements. To address these issues, we present a method to automatically identify seismic arrivals and measure their slowness vector properties with uncertainty bounds. We do this by bootstrap sampling waveforms, therefore also creating random sub arrays, then use linear beamforming to measure the coherent power at a range of slowness vectors. For each bootstrap sample, we take the top N peaks from each power distribution as the slowness vectors of possible arrivals. The slowness vectors of all bootstrap samples are gathered and the clustering algorithm DBSCAN (Density-Based Spatial Clustering of Applications with Noise) is used to identify arrivals as clusters of slowness vectors. The mean of slowness vectors in each cluster gives the slowness vector measurement for that arrival and the distribution of slowness vectors in each cluster gives the uncertainty estimate. We tuned the parameters of DBSCAN using a data set of 2489 SKS and SKKS observations at a range of frequency bands from 0.1 to 1 Hz. We then present examples at higher frequencies (0.5–2.0 Hz) than the tuning data set, identifying PKP precursors, and lower frequency by identifying multipathing in surface waves (0.04–0.06 Hz). While we use a linear beamforming process, this method can be implemented with any beamforming process such as cross correlation beamforming or phase weighted stacking. This method allows for much larger data sets to be analysed without visual inspection of data. Phenomena such as multipathing, reflections or scattering can be identified automatically in body or surface waves and their properties analysed with uncertainties. 

Long-range transport of biogenic emissions from the coast of Antarctica, precipitation scavenging, and cloud processing are the main processes that influence the observed variability in Southern Ocean (SO) marine boundary layer (MBL) condensation nuclei (CN) and cloud condensation nuclei (CCN) concentrations during the austral summer. Airborne particle measurements on the HIAPER GV from north-south transects between Hobart, Tasmania and 62°S during the Southern Ocean Clouds, Radiation Aerosol Transport Experimental Study (SOCRATES) were separated into four regimes comprising combinations of high and low concentrations of CCN and CN. In 5-day HYSPLIT back trajectories, air parcels with elevated CCN concentrations were almost always shown to have crossed the Antarctic coast, a location with elevated phytoplankton emissions relative to the rest of the SO in the region south of Australia. The presence of high CCN concentrations was also consistent with high cloud fractions over their trajectory, suggesting there was substantial growth of biogenically formed particles through cloud processing. Cases with low cloud fraction, due to the presence of cumulus clouds, had high CN concentrations, consistent with previously reported new particle formation in cumulus outflow regions. Measurements associated with elevated precipitation during the previous 1.5-days of their trajectory had low CCN concentrations indicating CCN were effectively scavenged by precipitation. A coarse-mode fitting algorithm was used to determine the primary marine aerosol (PMA) contribution which accounted for < 20% of CCN (at 0.3% supersaturation) and cloud droplet number concentrations. Vertical profiles of CN and large particle concentrations (Dp > 0.07µm) indicated that particle formation occurs more frequently above the MBL; however, the growth of recently formed particles typically occurs in the MBL, consistent with cloud processing and the condensation of volatile compound oxidation products. 

This paper describes the creation of a virtual, interactive professional development course to build the capacity of community college faculty to recruit and retain women and underrepresented minorities in computing programs. The project was designed in response to community college faculty reporting need for practical methods to broaden participation in their programs and their feelings of isolation from like-minded faculty. The 12-session prototype has been piloted with eight community college faculty. The finalized PD will be available as free, standalone web-based modules. The course includes instruction on research-based practices for recruiting and retaining women and underrepresented minorities in computing. Evaluation mechanisms are developed to assess the impacts of the PD on faculty attitudes and teaching practices, and the effect of changed practices on introductory computing students’ engagement and persistence. Here we report preliminary findings from interviews. The project outputs will include polished online content modules, validated student survey instruments, a classroom observation protocol, and student and faculty interview instruments.