Search for: All records

The three-dimensional characterization of internal features, via metrics such as orientation, porosity, and connectivity, is important to a wide variety of scientific questions. Many spatial and morphological metrics only can be measured accurately through direct in situ three-dimensional observations of large (i.e., big enough to be statistically representative) volumes. For samples that lack material contrast between phases, serial grinding and imaging—which relies solely on color and textural characteristics to differentiate features—is a viable option for extracting such information. Here, we present the Grinding, Imaging, Reconstruction Instrument (GIRI), which automatically serially grinds and photographs centimeter-scale samples at micron resolution. Although the technique is destructive, GIRI produces an archival digital image stack. This digital image stack is run through a supervised machine-learning-based image processing technique that quickly and accurately segments data into predefined classes. These classified data then can be loaded into three-dimensional visualization software for measurement. We share three case studies to illustrate how GIRI can address questions with a significant morphological component for which two-dimensional or small-volume three-dimensional measurements are inadequate. The analyzed metrics include: the morphologies of objects and pores in a granular material, the bulk mineralogy of polyminerallic solids, and measurements of the internal angles and symmetry of crystals.

 

Solving a bilevel optimization problem is at the core of several machine learning problems such as hyperparameter tuning, data denoising, meta- and few-shot learning, and training-data poisoning. Different from simultaneous or multi-objective optimization, the steepest descent direction for minimizing the upper-level cost in a bilevel problem requires the inverse of the Hessian of the lower-level cost. In this work, we propose a novel algorithm for solving bilevel optimization problems based on the classical penalty function approach. Our method avoids computing the Hessian inverse and can handle constrained bilevel problems easily. We prove the convergence of the method under mild conditions and show that the exact hypergradient is obtained asymptotically. Our method's simplicity and small space and time complexities enable us to effectively solve large-scale bilevel problems involving deep neural networks. We present results on data denoising, few-shot learning, and training-data poisoning problems in a large-scale setting. Our results show that our approach outperforms or is comparable to previously proposed methods based on automatic differentiation and approximate inversion in terms of accuracy, run-time, and convergence speed. 

Unsupervised domain adaptation (UDA) enables cross-domain learning without target domain labels by transferring knowledge from a labeled source domain whose distribution differs from the target. However, UDA is not always successful and several accounts of ‘negative transfer’ have been reported in the literature. In this work, we prove a simple lower bound on the target domain error that complements the existing upper bound. Our bound shows the insufficiency of minimizing source domain error and marginal distribution mismatch for a guaranteed reduction in the target domain error, due to the possible increase of induced labeling function mismatch. This insufficiency is further illustrated through simple distributions for which the same UDA approach succeeds, fails, and may succeed or fail with an equal chance. Motivated from this, we propose novel data poisoning attacks to fool UDA methods into learning representations that produce large target domain errors. We evaluate the effect of these attacks on popular UDA methods using benchmark datasets where they have been previously shown to be successful. Our results show that poisoning can significantly decrease the target domain accuracy, dropping it to almost 0% in some cases, with the addition of only 10% poisoned data in the source domain. The failure of UDA methods demonstrates the limitations of UDA at guaranteeing cross-domain generalization consistent with the lower bound. Thus, evaluation of UDA methods in adversarial settings such as data poisoning can provide a better sense of their robustness in scenarios unfavorable for UDA. 

The early‐middle Neoproterozoic is thought to have witnessed significant perturbations to marine P cycling, in turn facilitating the rise of eukaryote‐dominated primary production. However, with few robust constraints on aqueous P concentrations, current understanding of Neoproterozoic P cycling is generally model‐dependent. To provide new geochemical constraints, we combined microanalytical data sets with solid‐state Nuclear Magnetic Resonance, synchrotron‐based X‐ray Absorption Near Edge Structure spectroscopy, and micro‐X‐ray Fluorescence imaging to characterize the speciation and distribution of P in Tonian shallow‐water carbonate rocks. These data reflect shallow water phosphate concentrations 10–100× higher than modern systems, supporting the hypothesis that tectonically‐driven influxes in P periodically initiated kinetically‐controlled CaCO₃deposition, in turn destabilizing marine carbonate chemistry, climate, and nutrient inventories. Alongside these observations, a new compilation and statistical analysis of mudstone geochemistry data indicates that, in parallel, C_organd P burial increased across later Tonian continental margins until becoming decoupled at the close of the Tonian, implicating widespread N‐limitation triggered by increasing atmospheric O₂.

 

Strata from the Ediacaran Period (635 million to 538 million years ago [Ma]) contain several examples of enigmatic, putative shell-building metazoan fossils. These fossils may provide insight into the evolution and environmental impact of biomineralization on Earth, especially if their biological affinities and modern analogs can be identified. Recently, apparent morphological similarities with extant coralline demosponges have been used to assign a poriferan affinity toNamapoikia rietoogensis, a modular encrusting construction that is found growing between (and on) microbial buildups in Namibia. Here, we present three-dimensional reconstructions ofNamapoikiathat we use to assess the organism’s proposed affinity. Our morphological analyses, which comprise quantitative measurements of thickness, spacing, and connectivity, reveal thatNamapoikiaproduced approximately millimeter-thick meandering and branching/merging sheets. We evaluate this reconstructed morphology in the context of poriferan biology and determine thatNamapoikialikely is not a sponge-grade organism.