Search for: All records

Using the newly introduced ``occupation kernels,'' the present manuscript develops an approach to dynamic mode decomposition (DMD) that treats continuous time dynamics, without discretization, through the Liouville operator. The technical and theoretical differences between Koopman based DMD for discrete time systems and Liouville based DMD for continuous time systems are highlighted, which includes an examination of these operators over several reproducing kernel Hilbert spaces.

This paper introduces robustness verification for semantic segmentation neural networks (in short, semantic segmentation networks [SSNs]), building on and extending recent approaches for robustness verification of image classification neural networks. Despite recent progress in developing verification methods for specifications such as local adversarial robustness in deep neural networks (DNNs) in terms of scalability, precision, and applicability to different network architectures, layers, and activation functions, robustness verification of semantic segmentation has not yet been considered. We address this limitation by developing and applying new robustness analysis methods for several segmentation neural network architectures, specifically by addressing reachability analysis of up-sampling layers, such as transposed convolution and dilated convolution. We consider several definitions of robustness for segmentation, such as the percentage of pixels in the output that can be proven robust under different adversarial perturbations, and a robust variant of intersection-over-union (IoU), the typical performance evaluation measure for segmentation tasks. Our approach is based on a new relaxed reachability method, allowing users to select the percentage of a number of linear programming problems (LPs) to solve when constructing the reachable set, through a relaxation factor percentage. The approach is implemented within NNV, then applied and evaluated on segmentation datasets, such as amore »multi-digit variant of MNIST known as M2NIST. Thorough experiments show that by using transposed convolution for up-sampling and average-pooling for down-sampling, combined with minimizing the number of ReLU layers in the SSNs, we can obtain SSNs with not only high accuracy (IoU), but also that are more robust to adversarial attacks and amenable to verification. Additionally, using our new relaxed reachability method, we can significantly reduce the verification time for neural networks whose ReLU layers dominate the total analysis time, even in classification tasks.« less

Tellurium (Te) stable isotope measurements have the potential to serve as tracers of Te mobility and redox conditions in modern and ancient environments. Here, we present a method to measure Te isotope ratios by MC-ICP-MS utilizing a hydride generation system to efficiently deliver Te to the plasma, in combination with a 120 Te– 124 Te double spike. This approach allows for precise δ 130 Te/ 126 Te (2 σ : 0.09‰) measurements while using less than 8.75 ng of natural Te. Although hydride generation methods usually produce higher sensitivity than more conventional methods, for Te, the sensitivity is similar, on our instrument, to that achieved using a desolvating nebulizer. Nonetheless, hydride generation has an advantageous ability to exclude interfering elements such as Ba and allow analysis of samples without chemical separation of Te in some cases. We also demonstrate successfully a modified ion exchange procedure to separate various matrix components and isobaric interferences from Te in natural sediments. Analyses of multiple digestions of USGS standard reference materials, mine tailings, ancient sediments, and soils utilizing this approach show the largest spread in terrestrial Te isotopic composition to date (δ 130 Te/ 126 Te ∼ 1.21‰) and a lack of detectable mass-independentmore »fractionation.« less

Abstract We present the analysis and results of the first datasetcollected with the MARS neutron detectordeployed at the Oak Ridge NationalLaboratory Spallation Neutron Source (SNS) for the purpose ofmonitoring and characterizing the beam-related neutron (BRN) backgroundfor the COHERENT collaboration. MARS was positionednext to the COH-CsI coherent elastic neutrino-nucleus scattering detectorin the SNS basement corridor. This is the basement location ofclosest proximity to the SNS target and thus, of highest neutrino flux,but it is also well shielded from the BRN flux by infill concreteand gravel. These data show the detector registered roughly one BRN per day.Using MARS' measured detection efficiency, the incomingBRN flux is estimated to be 1.20 ± 0.56 neutrons/m^2/MWhfor neutron energies above ∼3.5 MeV and up to a few tens of MeV.We compare our results with previous BRN measurements in the SNS basement corridorreported by other neutron detectors.

In this letter, two different switching modes are identified for RF synchronous rectifier circuits which lead to a change in the polarity of the rectified output voltage. The operation of the two modes is defined in terms of the dynamic current-voltage (I-V) characteristics of the switching device and the polarity of the output load voltage. In one mode, the dc output voltage is positive and associated with a switching condition where the device is ON in quadrant-III and OFF in quadrant-I. This mode of operation has been reported widely in the literature. In the second mode, the quadrants associated with the on- and off-state are swapped leading to a polarity reversal across the load. Theoretical justification for the two modes is given and the experimental results are shown to confirm the two different switching modes in the synchronous rectifier. Experiments with a GaN inverse class-F rectifier were made to verify the theory, and both modes have similar measured peak rectification efficiencies of 77% at a frequency of 1.8 GHz.