Translational symmetry in convolutions with localized kernels causes an implicit bias toward high frequency adversarial examples.

Adversarial attacks are still a significant challenge for neural networks. Recent efforts have shown that adversarial perturbations typically contain high-frequency features, but the root cause of this phenomenon remains unknown. Inspired by theoretical work on linear convolutional models, we hypothesize that together with , and that this is one of the main causes of .

Advancing Carbon Fiber Composite Inspection: Deep Learning-Enabled Defect Localization and Sizing via 3-D U-Net Segmentation of Ultrasonic Data.

In nondestructive evaluation (NDE), accurately characterizing defects within components relies on accurate sizing and localization to evaluate the severity or criticality of defects. This study presents for the first time a deep learning (DL) methodology using 3-D U-Net to localize and size defects in carbon fiber reinforced polymer (CFRP) composites through volumetric segmentation of ultrasonic testing (UT) data. Using a previously developed approach, synthetic training data, closely representative of experimental data, was used for the automatic generation of ground truth segmentation masks.

Unsupervised machine learning for flaw detection in automated ultrasonic testing of carbon fibre reinforced plastic composites.

The use of Carbon Fibre Reinforced Plastic (CFRP) composite materials for critical components has significantly surged within the energy and aerospace industry. With this rapid increase in deployment, reliable post-manufacturing Non-Destructive Evaluation (NDE) is critical for verifying the mechanical integrity of manufactured components. To this end, an automated Ultrasonic Testing (UT) NDE process delivered by an industrial manipulator was developed, greatly increasing the measurement speed, repeatability, and locational precision, while increasing the throughput of data generated by the selected NDE modality.

Widespread support for a global species list with a formal governance system.

Taxonomic data are a scientific common. Unlike nomenclature, which has strong governance institutions, there are currently no generally accepted governance institutions for the compilation of taxonomic data into an accepted global list. This gap results in challenges for conservation, ecological research, policymaking, international trade, and other areas of scientific and societal importance.

Tightening the requirements for species diagnoses would help integrate DNA-based descriptions in taxonomic practice.

Modern advances in DNA sequencing hold the promise of facilitating descriptions of new organisms at ever finer precision but have come with challenges as the major Codes of bionomenclature contain poorly defined requirements for species and subspecies diagnoses (henceforth, species diagnoses), which is particularly problematic for DNA-based taxonomy. We, the commissioners of the International Commission on Zoological Nomenclature, advocate a tightening of the definition of "species diagnosis" in future editions of Codes of bionomenclature, for example, through the introduction of requirements for specific information on the character states of differentiating traits in comparison with similar species. Such new provisions would enhance taxonomic standards and ensure that all diagnoses, including DNA-based ones, contain adequate taxonomic context.

Shallow Univariate ReLU Networks as Splines: Initialization, Loss Surface, Hessian, and Gradient Flow Dynamics.

Understanding the learning dynamics and inductive bias of neural networks (NNs) is hindered by the opacity of the relationship between NN parameters and the function represented. Partially, this is due to symmetries inherent within the NN parameterization, allowing multiple different parameter settings to result in an identical output function, resulting in both an unclear relationship and redundant degrees of freedom. The NN parameterization is invariant under two symmetries: permutation of the neurons and a continuous family of transformations of the scale of weight and bias parameters.

Uncertainty Quantification for Deep Learning in Ultrasonic Crack Characterization.

Deep learning for nondestructive evaluation (NDE) has received a lot of attention in recent years for its potential ability to provide human level data analysis. However, little research into quantifying the uncertainty of its predictions has been done. Uncertainty quantification (UQ) is essential for qualifying NDE inspections and building trust in their predictions.

A new species of Trimma (Pisces: Gobiidae) from the deep reefs of Palau, western Pacific Ocean.

A new species of Trimma is described from three specimens from deep reefs (91.4 m) at Uchelbeluu Reef, Palau, western Pacific Ocean. Trimma panemorfum n.

The Herbarium 2021 Half-Earth Challenge Dataset and Machine Learning Competition.

Herbarium sheets present a unique view of the world's botanical history, evolution, and biodiversity. This makes them an all-important data source for botanical research. With the increased digitization of herbaria worldwide and advances in the domain of fine-grained visual classification which can facilitate automatic identification of herbarium specimen images, there are many opportunities for supporting and expanding research in this field.

Domain Adapted Deep-Learning for Improved Ultrasonic Crack Characterization Using Limited Experimental Data.

Deep learning is an effective method for ultrasonic crack characterization due to its high level of automation and accuracy. Simulating the training set has been shown to be an effective method of circumventing the lack of experimental data common to nondestructive evaluation (NDE) applications. However, a simulation can neither be completely accurate nor capture all variability present in the real inspection.

Identification of 22 novel BTK gene variants in B cell deficiency with hypogammaglobulinemia.

X-linked agammaglobulinemia (XLA) is an inborn error of immunity caused by pathogenic variants in the BTK gene, resulting in impaired B cell differentiation and maturation. Over 900 variants have already been described in this gene, however, new pathogenic variants continue to be identified. In this report, we describe 22 novel variants in BTK, associated with B cell deficiency with hypo- or agammaglobulinemia in male patients or in asymptomatic female carriers.

Pneumococcal serotype-specific cut-offs based on antibody responses to pneumococcal polysaccharide vaccination in healthy adults.

Antibody responses to pneumococcal polysaccharide vaccination are frequently used as a diagnostic tool for humoral immunodeficiencies, part of the larger collection of inborn errors of immunity. Currently, arbitrary criteria, such as a serotype specific titer of >/= 1.3 µg/mL is most often used as a cut-off for interpretation of pneumococcal antibody responses.

Domain-driven models yield better predictions at lower cost than reservoir computers in Lorenz systems.

Recent advances in computing algorithms and hardware have rekindled interest in developing high-accuracy, low-cost models for simulating physical systems. The idea is to replace expensive numerical integration of complex coupled partial differential equations at fine time scales performed on supercomputers, with machine-learned surrogates that efficiently and accurately forecast future system states using data sampled from the underlying system. One particularly popular technique being explored within the weather and climate modelling community is the (ESN), an attractive alternative to other well-known deep learning architectures.

Principles for creating a single authoritative list of the world's species.

Lists of species underpin many fields of human endeavour, but there are currently no universally accepted principles for deciding which biological species should be accepted when there are alternative taxonomic treatments (and, by extension, which scientific names should be applied to those species). As improvements in information technology make it easier to communicate, access, and aggregate biodiversity information, there is a need for a framework that helps taxonomists and the users of taxonomy decide which taxa and names should be used by society whilst continuing to encourage taxonomic research that leads to new species discoveries, new knowledge of species relationships, and the refinement of existing species concepts. Here, we present 10 principles that can underpin such a governance framework, namely (i) the species list must be based on science and free from nontaxonomic considerations and interference, (ii) governance of the species list must aim for community support and use, (iii) all decisions about list composition must be transparent, (iv) the governance of validated lists of species is separate from the governance of the names of taxa, (v) governance of lists of accepted species must not constrain academic freedom, (vi) the set of criteria considered sufficient to recognise species boundaries may appropriately vary between different taxonomic groups but should be consistent when possible, (vii) a global list must balance conflicting needs for currency and stability by having archived versions, (viii) contributors need appropriate recognition, (ix) list content should be traceable, and (x) a global listing process needs both to encompass global diversity and to accommodate local knowledge of that diversity.

Mesophotic.org: a repository for scientific information on mesophotic ecosystems.

Mesophotic coral ecosystems (MCEs) and temperate mesophotic ecosystems (TMEs) occur at depths of roughly 30-150 m depth and are characterized by the presence of photosynthetic organisms despite reduced light availability. Exploration of these ecosystems dates back several decades, but our knowledge remained extremely limited until about a decade ago, when a renewed interest resulted in the establishment of a rapidly growing research community. Here, we present the 'mesophotic.

A Reservoir Computing Model of Reward-Modulated Motor Learning and Automaticity.

Reservoir computing is a biologically inspired class of learning algorithms in which the intrinsic dynamics of a recurrent neural network are mined to produce target time series. Most existing reservoir computing algorithms rely on fully supervised learning rules, which require access to an exact copy of the target response, greatly reducing the utility of the system. Reinforcement learning rules have been developed for reservoir computing, but we find that they fail to converge on complex motor tasks.

Circuit Models of Low-Dimensional Shared Variability in Cortical Networks.

Trial-to-trial variability is a reflection of the circuitry and cellular physiology that make up a neuronal network. A pervasive yet puzzling feature of cortical circuits is that despite their complex wiring, population-wide shared spiking variability is low dimensional. Previous model cortical networks cannot explain this global variability, and rather assume it is from external sources.

, a new basslet from deep coral reefs in Micronesia (Perciformes, Percoidei, Serranidae).

The new species is described from four specimens collected at depths of 115-148 m near Palau and Pohnpei in Micronesia. It differs from the other three species of this genus in life color and in certain morphological characters, such as body depth, snout length, anterior three dorsal-fin spine lengths, caudal-fin length, and other characters. There are also genetic differences from the other four species of (d ≈ 0.

Mesophotic coral ecosystems are threatened and ecologically distinct from shallow water reefs.

The rapid degradation of coral reefs is one of the most serious biodiversity problems facing our generation. Mesophotic coral reefs (at depths of 30 to 150 meters) have been widely hypothesized to provide refuge from natural and anthropogenic impacts, a promise for the survival of shallow reefs. The potential role of mesophotic reefs as universal refuges is often highlighted in reef conservation research.

, a new basslet (Perciformes, Percoidei, Serranidae) from deep coral reefs in the Northwestern Hawaiian Islands.

The new species is described from two specimens collected at a depth of 90-92 m off Kure Atoll and Pearl and Hermes Atoll, Northwestern Hawaiian Islands. It differs from the other two species of this genus in life color and in certain morphological characters, such as number of pored lateral-line scales, pectoral-fin rays, snout length, anterior three dorsal-fin spine lengths, dorsal-fin profile, and other characters. There are also substantial genetic differences from the other two species of (d ≈ 0.

Spatiotemporal Dynamics and Reliable Computations in Recurrent Spiking Neural Networks.

Randomly connected networks of excitatory and inhibitory spiking neurons provide a parsimonious model of neural variability, but are notoriously unreliable for performing computations. We show that this difficulty is overcome by incorporating the well-documented dependence of connection probability on distance. Spatially extended spiking networks exhibit symmetry-breaking bifurcations and generate spatiotemporal patterns that can be trained to perform dynamical computations under a reservoir computing framework.

A comprehensive investigation of mesophotic coral ecosystems in the Hawaiian Archipelago.

Although the existence of coral-reef habitats at depths to 165 m in tropical regions has been known for decades, the richness, diversity, and ecological importance of mesophotic coral ecosystems (MCEs) has only recently become widely acknowledged. During an interdisciplinary effort spanning more than two decades, we characterized the most expansive MCEs ever recorded, with vast macroalgal communities and areas of 100% coral cover between depths of 50-90 m extending for tens of km in the Hawaiian Archipelago. We used a variety of sensors and techniques to establish geophysical characteristics.