196746 research outputs found
Sort by
Dynamic robustness evaluation for automated model selection in operation
No full textContext:The increasing use of artificial neural network (ANN) classifiers in systems, especially safety-critical systems (SCSs), requires ensuring their robustness against out-of-distribution (OOD) shifts in operation, which are changes in the underlying data distribution from the data training the classifier. However, measuring the robustness of classifiers in operation with only unlabeled data is challenging. Additionally, machine learning engineers may need to compare different models or versions of the same model and switch to an optimal version based on their robustness.Objective:This paper explores the problem of dynamic robustness evaluation for automated model selection. We aim to find efficient and effective metrics for evaluating and comparing the robustness of multiple ANN classifiers using unlabeled operational data.Methods:To quantitatively measure the differences between the model outputs and assess robustness under OOD shifts using unlabeled data, we choose distance-based metrics. An empirical comparison of five such metrics, suitable for higher-dimensional data like images, is performed. The selected metrics include Wasserstein distance (WD), maximum mean discrepancy (MMD), Hellinger distance (HL), Kolmogorov–Smirnov statistic (KS), and Kullback–Leibler divergence (KL), known for their efficacy in quantifying distribution differences. We evaluate these metrics on 20 state-of-the-art models (ten CIFAR10-based models, five CIFAR100-based models, and five ImageNet-based models) from a widely used robustness benchmark (RobustBench) using data perturbed with various types and magnitudes of corruptions to mimic real-world OOD shifts.Results:Our findings reveal that the WD metric outperforms others when ranking multiple ANN models for CIFAR10- and CIFAR100-based models, while the KS metric demonstrates superior performance for ImageNet-based models. MMD can be used as a reliable second option for both datasets.Conclusion:This study highlights the effectiveness of distance-based metrics in ranking models’ robustness for automated model selection. It also emphasizes the significance of advancing research in dynamic robustness evaluation
Multimodal 3D quantification of particle stimulated nucleation in industrially manufactured aluminium AA5182 sheet
No full textParticle stimulated nucleation is a dominant recrystallisation mechanism observed in many industrially relevant aluminium alloys during thermomechanical processing. In this work, we quantify particle stimulated nucleation in 3D in an aluminium AA5182 alloy sheet cold-rolled to 75% thickness reduction. Second phase particles and nuclei are mapped in the same sample volume by conventional laboratory absorption X-ray tomography and synchrotron X-ray Laue micro-diffraction. The large second phase particles are classified as Fe- and Mg-rich phases. It is found that 84% of the nuclei are particle stimulated and 40% of the particles stimulate nucleation. The critical particle diameter is found to be 4 μm . Deviatoric elastic strains are derived from micro-diffraction data and it is found that elastic strains are present in the recrystallised nuclei. The effects of the different particle types, particle clustering, particle size and aspect ratio as well as strain inheritance are discussed. This work provides a full 3D quantification of particle stimulated nucleation behaviour in AA5182 alloy sheet deformed to high strain
Application of CRISPR/Cas9 Genome Editing to Improve Recombinant Protein Production in CHO Cells
No full textGenome editing has become an important aspect of Chinese hamster ovary (CHO) cell line engineering for improving the production of recombinant protein therapeutics. Currently, the engineering focus is directed toward expanding product diversity while controlling and improving product quality and yields. In this chapter, we present our protocol for using the genome editing tool Clustered Regularly Interspaced Short Palindromic Repeat (CRISPR)/CRISPR-associated protein 9 (Cas9) to knock out engineering target genes in CHO cells. As an example, we describe how to knock out the glutamine synthetase (GS) gene, which increases the selection efficiency of the GS-mediated gene amplification system
Effect of grain size and orientation on magnetron sputtering yield of tantalum
No full textThe electron beam melting (EBM) technique was employed to prepare ultra-highly pure (99.999 wt%) Tantalum (Ta) cast ingot for application in chips. Subsequently, the Ta cast ingot were forged, rolled, and annealed with different durations to gain three different grain sizes (centimeter scale, 99.8 μm, and 36.7 μm). Sputtering experiments conducted under identical conditions revealed that the rolled target (36.7 μm) film deposition rate was increased by 60.6 % compared to the cast ingot target with a centimeter-scale grain size (columnar crystal). Ta targets with a fine grain size and homogeneous distribution demonstrate superior film deposition performance. The sputtering rate is directly related to the atomic packing density of grains. The (111)-oriented grains of BCC targets (Ta target) exhibit sputtering resistance, and the order of sputtering rate of Ta atoms was S(101) > S(001) > S(111)
Democratizing uncertainty quantification
No full textUncertainty Quantification (UQ) is vital to safety-critical model-based analyses, but the widespread adoption of sophisticated UQ methods is limited by technical complexity. In this paper, we introduce UM-Bridge (the UQ and Modeling Bridge), a high-level abstraction and software protocol that facilitates universal interoperability of UQ software with simulation codes. It breaks down the technical complexity of advanced UQ applications and enables separation of concerns between experts. UM-Bridge democratizes UQ by allowing effective interdisciplinary collaboration, accelerating the development of advanced UQ methods, and making it easy to perform UQ analyses from prototype to High Performance Computing (HPC) scale. In addition, we present a library of ready-to-run UQ benchmark problems, all easily accessible through UM-Bridge. These benchmarks support UQ methodology research, enabling reproducible performance comparisons. We demonstrate UM-Bridge with several scientific applications, harnessing HPC resources even using UQ codes not designed with HPC support.</p
UltraCommander: Ultrasonic Side Channel Attack via Browser Extensions
No full textUltrasound, imperceptible to the human ear, emerges as a potent carrier of information. Exploiting its covert nature, a growing number of both malicious entities and researchers delve into its potential privacy threats. Simultaneously, the surge in Browser Extensions, prized for their streamlined access to privileged browser resources, adds a layer of convenience while raising a big concern. This work introduces UltraCommander, a kind of cyber attack that leverages ultrasonic channel communication. In particular, UltraCommander covertly monitors user private data based on attacker-specified commands, facilitating surreptitious data transmission to the attacker. We then intricately detail the implementation of this attack and showcase its imperceptible nature. Our research underscores the successful exfiltration of private data through Chrome APIs, with future prospects extending to capturing additional information, such as SMS data, from browsers
Neural networks for reconstruction and uncertainty quantification of fast-ion phase-space distributions using FILD and INPA measurements
No full textThis study introduces the use of a deep convolutional neural network for reconstructing fast-ion velocity distributions from fast-ion loss detectors and imaging neutral particle analyzers (INPAs), automatically integrating uncertainty quantification through Monte Carlo dropout. The network-based reconstructions reveal pitch-angle splitting in high-energy features of lost fast-ion velocity distributions at ASDEX Upgrade during active neutral beam injection, a previously observed phenomenon now confirmed through neural networks. Moreover, contrary to common theories attributing these high-energy features to edge localized mode (ELM)-driven acceleration, we provide experimental evidence that they also occur in type-I ELM-quiescent phases. Additionally, we demonstrate improved reconstructions from INPA measurements, both synthetic and from an ASDEX Upgrade commissioning discharge, with the reconstructions closely matching TRANSP simulations. These findings suggest that neural networks can provide robust reconstructions with well-defined uncertainties, improving the reliability of interpretations of fast-ion behavior in magnetically confined plasmas
Testing of a passive foam fractionator prototype in a commercial recirculating trout farm
No full textFoam fractionation has emerged as a technical solution to reduce the build-up of microparticles and dissolved organic matter in recirculating aquaculture systems (RAS). However, commercial application in freshwater RAS is challenging and expensive. In the present study, a simple, low-cost passive foam fractionation (PFF) prototype was developed and tested under commercial conditions. The prototype was tested in a Model Trout Farm (MTF) in three different production raceways during winter and spring to assess the operation and removal potential. A number of different water quality parameters, including organic matter, particles, bacterial activity, and phosphorus were examined in the system water and in the removed foamate. Overall, the PFF prototype removed particles as well as particulate and dissolved organic matter, reduced the amount of bacteria and total phosphorus in the water, regardless of sampling time and place. By utilizing the existing airlifts in the MTF, the associated cost of construction and operation was kept low. Overall, the results demonstrate that the passive foam fractionation has the potential to help address some of aquaculture’s pressing issues in a cost effective manner
Locational Marginal Carbon Emission of Power Grids Approach: Optimal Scheduling of Recycling Electricity/Heat Rural Supply System Based on Waste Feedstock
No full textTo improve the power supply ability, heat supply ability, and waste recovery rate, a recycling electricity/heat rural supply system with waste feedstock is established. The energy supply system generates electricity/heat from biomass energy produced by wastes, which is also coupled to distributed renewable energy. The optimal scheduling of the established rural system will improve energy efficiency and cause emission reduction. Firstly, the waste recovery process is presented, and the architecture of the energy supply system is designed for the 100% absorption of renewable energy in rural areas. A carbon accounting model based on the locational marginal carbon emission factor is introduced, which considers the power exchange with the bulk power system and the carbon emission of biomass. Secondly, the optimal scheduling model for the recycling energy supply system is proposed to minimize both the total cost of energy supply and carbon emission, based on the constraints of energy balancing of electricity and heat, net carbon emissions, waste supply, etc. Finally, the IEEE 15-node system and PG&E 69-node system are employed for verification purposes. The proposed model contributes to 100% absorption of renewable energy and the efficient utilization of waste through the optimal cooperation of the waste supply, biomass power generation, and biomass heat, thereby supporting the achievement of zero carbon
Visualizing Metabolism in Biotechnologically Important Yeasts with dDNP NMR Reveals Evolutionary Strategies and Glycolytic Logic
No full textSaccharomyces cerevisiae has long been a pillar of biotechnological production and basic research. More recently, strides to exploit the functional repertoire of nonconventional yeasts for biotechnological production have been made. Genomes and genetic tools for these yeasts are not always available, and yeast genomics alone may be insufficient to determine the functional features in yeast metabolism. Hence, functional assays of metabolism, ideally in the living cell, are best suited to characterize the cellular biochemistry of such yeasts. Advanced in cell NMR methods can allow the direct observation of carbohydrate influx into central metabolism on a seconds time scale: dDNP NMR spectroscopy temporarily enhances the nuclear spin polarization of substrates by more than 4 orders of magnitude prior to functional assays probing central metabolism. We use various dDNP enhanced carbohydrates for in-cell NMR to compare the metabolism of S. cerevisiae and nonconventional yeasts, with an emphasis on the wine yeast Hanseniaspora uvarum. In-cell observations indicated more rapid exhaustion of free cytosolic NAD+ in H. uvarum and alternative routes for pyruvate conversion, in particular, rapid amination to alanine. In-cell observations indicated that S. cerevisiae outcompetes other biotechnologically relevant yeasts by rapid ethanol formation due to the efficient adaptation of cofactor pools and the removal of competing reactions from the cytosol. By contrast, other yeasts were better poised to use redox neutral processes that avoided CO2-emission. Beyond visualizing the different cellular strategies for arriving at redox neutral end points, in-cell dDNP NMR probing showed that glycolytic logic is more conserved: nontoxic precursors of cellular building blocks formed high-population intermediates in the influx of glucose into the central metabolism of eight different biotechnologically important yeasts. Unsupervised clustering validated that the observation of rapid intracellular chemistry is a viable means to functionally classify biotechnologically important organisms