Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “hinge loss”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 19 records

Levenberg–Marquardt multi-classification using hinge loss function

Incorporating higher-order optimization functions, such as Levenberg-Marquardt (LM) have revealed better generalizable solutions for deep learning problems. However, these higher-order optimization functions suffer from very large processing time and training complexity especially as training datasets become large, such as in multi-view classification problems, where finding global optima is a very costly problem. To solve this issue, we develop a solution for LM-enabled classification with, to the best of knowledge first-time implementation of hinge loss, for multiview classification. Hinge loss allows the neural network to converge faster and perform better than other loss functions such as logistic or square loss rates. Here we prove our method by experimenting with various multiclass classification challenges of varying complexity and training data size. The empirical results show the training time and accuracy rates achieved, highlighting how our method outperforms in all cases, especially when training time is limited. Our paper presents important results in the relationship between optimization and loss functions and how these can impact deep learning problems.

97 MATHEMATICS AND COMPUTING↗

Using support vector machines for anomalous change detonation

We cast anomalous change detection as a binary classification problem, and use a support vector machine (SVM) to build a detector that does not depend on assumptions about the underlying data distribution. To speed up the computation, our SVM is implemented, in part, on a graphical processing unit. Results on real and simulated anomalous changes are used to compare performance to algorithms which effectively assume a Gaussian distribution. In this paper, we investigate the use of support vector machines (SVMs) with radial basis kernels for finding anomalous changes. Compared to typical applications of SVMs, we are operating in a regime of very low false alarm rate. This means that even for relatively large training sets, the data are quite meager in the regime of operational interest. This drives us to use larger training sets, which in turn places more of a computational burden on the SVM. We initially considered three different approaches to to address the need to work in the very low false alarm rate regime. The first is a standard SVM which is trained at one threshold (where more reliable estimates of false alarm rates are possible) and then re-thresholded for the low false alarm rate regime. The second uses the same thresholding approach, but employs a so-called least squares SVM; here a quadratic (instead of a hinge-based) loss function is employed, and for this model, there are good theoretical arguments in favor of adjusting the threshold in a straightforward manner. The third approach employs a weighted support vector machine, where the weights for the two types of errors (false alarm and missed detection) are automatically adjusted to achieve the desired false alarm rate. We have found in previous experiments (not shown here) that the first two types can in some cases work well, while in other cases they do not. This renders both approaches unreliable for automated change detection. By contrast, the third approach reliably produces good results, but at the cost of larger computational requirements caused by the need to estimate very small false alarm rates. To address these computational requirements, we employ a recently developed in-house solver for SVMs that is significantly faster than freely available standard solvers. But these computational issues are secondary to the larger question: do kernelized solutions provide better performance, in terms of detection rates and false alarm rates, than more traditional methods for change detection that effectively assume Gaussian data distributions? To this end, we will compare ROC curves obtained from the SVM with those from chronochrome, covariance equalization, and hyperbolic anomalous change detection.

97 MATHEMATICS AND COMPUTING↗

Gyrokinetic analysis and simulation of pedestals to identify the culprits for energy losses using ‘fingerprints’

Fusion performance in tokamaks hinges critically on the efficacy of the edge transport barrier (ETB) in suppressing energy losses. The new concept of 'fingerprints' is introduced to identify the instabilities that cause transport losses in the ETBs of many of today's experiments, from among widely posited candidates. Analysis of the gyrokinetic-Maxwell equations and gyrokinetic simulations of experiments reveals that each mode type produces characteristic ratios of transport in the various channels: density, heat, and impurities. This, together with experimental observations of transport in some channel or of the relative size of the driving sources of channels, can identify or determine the dominant modes causing energy transport. In multiple H-mode cases with edge-localized modes that are examined, these fingerprints indicate that magnetohydrodynamic (MHD)-like modes are apparently not the dominant agent of energy transport; rather, this role is played by micro-tearing modes (MTMs) and electron temperature gradient (ETG) modes, and furthermore, possibly by ion temperature gradient/trapped electron modes (ITG/TEM) on JET (Joint European Torus). MHD-like modes may dominate the electron particle losses. Fluctuation frequency can also be a vital means of identification, and is often closely related to the transport fingerprint. The analytical arguments unify and explain previously disparate experimental observations on multiple devices, including DIII-D, JET, and ASDEX-U. Detailed simulations of two DIII-D ETBs also demonstrate and corroborate this.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗

Three-dimensional folding of pre-strained polymer sheets via absorption of laser light

Patterned light from a laser can induce rapid self-folding of pre-strained polymer sheets. Black ink coated on the sheet absorbs the light, which converts the photon energy into thermal energy that heats the sheet locally; the temperature of the sheet is highest at the surface where the light impinges on the sheet and decreases through the sheet thickness. The gradient of temperature induces a gradient of strain relaxation through the depth of the sheet, which causes folding within seconds of irradiation. The pattern of laser light that irradiates the compositionally homogeneous two-dimensional (2D) substrate dictates the resulting three-dimensional (3D) shape. Unlike most approaches to self-folding, the methodology described here requires no patterning of pre-defined hinges. It opens up the possibility of using a patterning technique that is inherently 2D to form 3D shapes. The use of lasers also enables systematic control of key process parameters such as power, intensity, and the pattern of light (i.e., beam width and shape). The rate of folding and folding angle measured with respect to these parameters provide an indirect quantification of heat loss in the sample and thereby identify the threshold power and power intensity that must be delivered to the hinge for folding to occur.

75 CONDENSED MATTER PHYSICS, SUPERCONDUCTIVITY AND↗

Novel Technique for Developing Linearized Convex System Models from Experimentally Derived Data

This paper presents a novel technique for generating a convex system model from an experimentally derived data set which features variance among repeated measurements. The convex system model developed as a test case characterizes the dynamic system losses of a vanadium redox flow battery as a function of the active power output and the battery state of charge. The technique hinges on a pre-cleaning via clustering procedure which precedes the formation of a planar convex hull comprised of triangular simplices. The clustering procedure efficiently reduces the experimental data set while mitigating variance among repeated measurements and removes outliers. Ultimately, the lower evelope of the planar convex hull serves as the desired convex system model. The proposed technique reduces systematic model error which is otherwise present when directly developing a planar convex hull model based on an unreduced data set.

24 POWER TRANSMISSION AND DISTRIBUTION↗

Accessing the exceptional points of parity-time symmetric acoustics

Parity-time (PT) symmetric systems experience phase transition between PT exact and broken phases at exceptional point. These PT phase transitions contribute significantly to the design of single mode lasers, coherent perfect absorbers, isolators, and diodes. However, such exceptional points are extremely difficult to access in practice because of the dispersive behaviour of most loss and gain materials required in PT symmetric systems. Here we introduce a method to systematically tame these exceptional points and control PT phases. Our experimental demonstration hinges on an active acoustic element that realizes a complex-valued potential and simultaneously controls the multiple interference in the structure. The manipulation of exceptional points offers new routes to broaden applications for PT symmetric physics in acoustics, optics, microwaves and electronics, which are essential for sensing, communication and imaging.

42 ENGINEERING↗

Poster - 18: New features in EGSnrc for photon cross sections

Purpose: To implement two new features in the EGSnrc Monte Carlo system. The first is an option to account for photonuclear attenuation, which can contribute a few percent to the total cross section at the higher end of the energy range of interest to medical physics. The second is an option to use exact NIST XCOM photon cross sections. Methods: For the first feature, the photonuclear total cross sections are generated from the IAEA evaluated data. In the current, first-order implementation, after a photonuclear event, there is no energy deposition or secondary particle generation. The implementation is validated against deterministic calculations and experimental measurements of transmission signals. For the second feature, before this work, if the user explicitly requested XCOM photon cross sections, EGSnrc still used its own internal incoherent scattering cross sections. These differ by up to 2% from XCOM data between 30 keV and 40 MeV. After this work, exact XCOM incoherent scattering cross sections are an available option. Minor interpolation artifacts in pair and triplet XCOM cross sections are also addressed. The default for photon cross section in EGSnrc is XCOM except for the new incoherent scattering cross sections, which have to be explicitly requested. The photonuclear, incoherent, pair and triplet data from this work are available for elements and compounds for photon energies from 1 keV to 100 GeV. Results: Both features are implemented and validated in EGSnrc.Conclusions: The two features are part of the standard EGSnrc distribution as of version 4.2.3.2.

73 NUCLEAR PHYSICS AND RADIATION PHYSICS↗

From broadband biphotons to frequency combs via spectral compression with time-varying cavities

Biphoton frequency combs are promising resources for quantum networking due in large part to their compatibility with the telecommunication infrastructure. Here, in this work, we propose a method to periodically compress broadband frequency-entangled photons into biphoton frequency combs by utilizing time-varying linear cavities. Our approach hinges on rapid modulation of the input cavity coupling, yielding high spectral purity in each output comb line similar to that achieved with narrowband filters, but without the associated loss in flux. We examine the dependence of spectral purity and compression on coupling strength, cavity loss, and switching speed, finding realistic regimes supporting purities in excess of 0.999 and peak enhancement factors of 100× and beyond.

Myilswamy, Karthik V. [Purdue Univ., West Lafayett↗

An IRE1-proteasome system signalling cohort controls cell fate determination in unresolved proteotoxic stress of the plant endoplasmic reticulum

Excessive accumulation of misfolded proteins in the endoplasmic reticulum (ER) causes ER stress, which is an underlying cause of major crop losses and devastating human conditions. Here, ER proteostasis surveillance is mediated by the conserved master regulator of the unfolded protein response (UPR), Inositol Requiring Enzyme 1 (IRE1), which determines cell fate by controlling pro-life and pro-death outcomes through as yet largely unknown mechanisms. Here we report that Arabidopsis IRE1 determines cell fate in ER stress by balancing the ubiquitin–proteasome system (UPS) and UPR through the plant-unique E3 ligase, PHOSPHATASE TYPE 2CA (PP2CA)-INTERACTING RING FINGER PROTEIN 1 (PIR1). Indeed, PIR1 loss leads to suppression of pro-death UPS and the lethal phenotype of an IRE1 loss-of-function mutant in unresolved ER stress in addition to activating pro-survival UPR. Specifically, in ER stress, PIR1 loss stabilizes ABI5, a basic leucine zipper (bZIP) transcription factor, that directly activates expression of the critical UPR regulator gene, bZIP60, triggering transcriptional cascades enhancing pro-survival UPR. Collectively, our results identify new cell fate effectors in plant ER stress by showing that IRE1’s coordination of cell death and survival hinges on PIR1, a key pro-death component of the UPS, which controls ABI5, a pro-survival transcriptional activator of bZIP60.

59 BASIC BIOLOGICAL SCIENCES↗

Design of a Loss of Coolant Blowdown Capsule for Remote Assembly with High Burnup Fuel

The Transient Water Irradiation System is an enhanced capability capsule type irradiation vehicle designed to support fuel safety research for light water reactor specimens in the Transient Reactor Test Facility and is designed to simulate loss of coolant and reactivity-initiated accidents. The capsule was designed, deployed, and commissioned with fresh fuel specimens to validate instrumentation and prepare for previously irradiated high burnup specimens. The irradiation system features an extensive in-situ instrumentation package to detect phenomena typical to light water reactor fuels. To accommodate the assembly with high burnup specimens inside the Hot Fuel Examination Facility, and to ensure instrumentation integrity is maintained throughout assembly, the design was updated to support remote handling. The updated design features a hinge mechanism which allows for remote pre-irradiated specimen loading and assembly while protecting sensitive instrumentation by relocating during loading. Fixtures and equipment have been developed to handle the experiment capsule and components in cell, and to support remote assembly. The experiment module also includes radiation shielding and contamination control to support operations after removal from the hot cell.

11 NUCLEAR FUEL CYCLE AND FUEL MATERIALS↗

Efficient charge generation from triplet excitons in metal-organic heterojunctions

The success of many emerging molecular electronics concepts hinges on an atomistic understanding of the underlying electronic dynamics. We employ picosecond time-resolved x-ray photoemission spectroscopy (tr-XPS) to elucidate the roles of singlet and triplet excitons for photoinduced charge generation at a copper-phthalocyanine-C 60 heterojunction. Contrary to common belief, fast intersystem crossing to triplet excitons after photoexcitation is not a loss channel but contributes to a significantly larger extent to the time-integrated interfacial charge generation than the initially excited singlet excitons. The tr-XPS data provide direct access to the diffusivity of the triplet excitons D CuPc =(1.8±1.2)×10 -5 cm 2 /s (where CuPc is copper-phthalocyanine) and their diffusion length L diff =(8±3)nm.

75 CONDENSED MATTER PHYSICS, SUPERCONDUCTIVITY AND↗

Minimizing the impacts of the ammonia economy on the nitrogen cycle and climate

Ammonia (NH 3 ) is an attractive low-carbon fuel and hydrogen carrier. However, losses and inefficiencies across the value chain could result in reactive nitrogen emissions (NH 3 , NO x , and N 2 O), negatively impacting air quality, the environment, human health, and climate. A relatively robust ammonia economy (30 EJ/y) could perturb the global nitrogen cycle by up to 65 Mt/y with a 5% nitrogen loss rate, equivalent to 50% of the current global perturbation caused by fertilizers. Moreover, the emission rate of nitrous oxide (N 2 O), a potent greenhouse gas and ozone-depleting molecule, determines whether ammonia combustion has a greenhouse footprint comparable to renewable energy sources or higher than coal (100 to 1,400 gCO 2 e/kWh). The success of the ammonia economy hence hinges on adopting optimal practices and technologies that minimize reactive nitrogen emissions. We discuss how this constraint should be included in the ongoing broad engineering research to reduce environmental concerns and prevent the lock-in of high-leakage practices.

54 ENVIRONMENTAL SCIENCES↗

Structures of trehalose-6-phosphate phosphatase from pathogenic fungi reveal the mechanisms of substrate recognition and catalysis

Trehalose is a disaccharide essential for the survival and virulence of pathogenic fungi. The biosynthesis of trehalose requires trehalose-6-phosphate synthase, Tps1, and trehalose-6-phosphate phosphatase, Tps2. Here, we report the structures of the N-terminal domain of Tps2 (Tps2NTD) from Candida albicans , a transition-state complex of the Tps2 C-terminal trehalose-6-phosphate phosphatase domain (Tps2PD) bound to BeF 3 and trehalose, and catalytically dead Tps2PD(D24N) from Cryptococcus neoformans bound to trehalose-6-phosphate (T6P). The Tps2NTD closely resembles the structure of Tps1 but lacks any catalytic activity. The Tps2PD–BeF 3 –trehalose and Tps2PD(D24N)–T6P complex structures reveal a “closed” conformation that is effected by extensive interactions between each trehalose hydroxyl group and residues of the cap and core domains of the protein, thereby providing exquisite substrate specificity. Disruption of any of the direct substrate–protein residue interactions leads to significant or complete loss of phosphatase activity. Notably, the Tps2PD–BeF 3 –trehalose complex structure captures an aspartyl-BeF 3 covalent adduct, which closely mimics the proposed aspartyl-phosphate intermediate of the phosphatase catalytic cycle. Structures of substrate-free Tps2PD reveal an “open” conformation whereby the cap and core domains separate and visualize the striking conformational changes effected by substrate binding and product release and the role of two hinge regions centered at approximately residues 102–103 and 184–188. Significantly, tps2 Δ, tps2NTD Δ, and tps2D705N strains are unable to grow at elevated temperatures. Combined, these studies provide a deeper understanding of the substrate recognition and catalytic mechanism of Tps2 and provide a structural basis for the future design of novel antifungal compounds against a target found in three major fungal pathogens.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

Structural insights into lipid-dependent reversible dimerization of human GLTP

It is shown that dimerization is promoted by glycolipid binding to human GLTP. The importance of dimer flexibility in wild-type protein is manifested by point mutation that ‘locks’ the dimer while diversifying ligand/protein adaptations. Human glycolipid transfer protein (hsGLTP) forms the prototypical GLTP fold and is characterized by a broad transfer selectivity for glycosphingolipids (GSLs). The GLTP mutation D48V near the ‘portal entrance’ of the glycolipid binding site has recently been shown to enhance selectivity for sulfatides (SFs) containing a long acyl chain. Here, nine novel crystal structures of hsGLTP and the SF-selective mutant complexed with short-acyl-chain monoSF and diSF in different crystal forms are reported in order to elucidate the potential functional roles of lipid-mediated homodimerization. In all crystal forms, the hsGLTP–SF complexes displayed homodimeric structures supported by similarly organized intermolecular interactions. The dimerization interface always involved the lipid sphingosine chain, the protein C-terminus (C-end) and α-helices 6 and 2, but the D48V mutant displayed a ‘locked’ dimer conformation compared with the hinge-like flexibility of wild-type dimers. Differences in contact angles, areas and residues at the dimer interfaces in the ‘flexible’ and ‘locked’ dimers revealed a potentially important role of the dimeric structure in the C-end conformation of hsGLTP and in the precise positioning of the key residue of the glycolipid recognition centre, His140. ΔY207 and ΔC-end deletion mutants, in which the C-end is shifted or truncated, showed an almost complete loss of transfer activity. The new structural insights suggest that ligand-dependent reversible dimerization plays a role in the function of human GLTP.

75 CONDENSED MATTER PHYSICS, SUPERCONDUCTIVITY AND↗

Structure of FcγRI in complex with Fc reveals the importance of glycan recognition for high-affinity IgG binding

Fc gamma receptor I (FcγRI) contributes to protective immunity against bacterial infections, but exacerbates certain autoimmune diseases. The sole high-affinity IgG receptor, FcγRI plays a significant role in immunotherapy. To elucidate the molecular mechanism of its high-affinity IgG binding, we determined the crystal structure of the extracellular domains of human FcγRI in complex with the Fc domain of human IgG 1 . FcγRI binds to the Fc in a similar mode as the low-affinity FcγRII and FcγRIII receptors. In addition to many conserved contacts, FcγRI forms additional hydrogen bonds and salt bridges with the lower hinge region of Fc. Unique to the high-affinity receptor-Fc complex, however, is the conformation of the receptor D2 domain FG loop, which enables a charged KHR motif to interact with proximal carbohydrate units of the Fc glycans. Both the length and the charge of the FcγRI FG loop are well conserved among mammalian species. Ala and Glu mutations of the FG loop KHR residues showed significant contributions of His-174 and Arg-175 to antibody binding, and the loss of the FG loop–glycan interaction resulted in an ~20- to 30-fold decrease in FcγRI affinity to all three subclasses of IgGs. Furthermore, deglycosylation of IgG 1 resulted in a 40-fold loss in FcγRI binding, demonstrating involvement of the receptor FG loop in glycan recognition. Here, these results highlight a unique glycan recognition in FcγRI function and open potential therapeutic avenues based on antibody glycan engineering or small molecular glycan mimics to target FcγRI for certain autoimmune diseases.

59 BASIC BIOLOGICAL SCIENCES↗

Nonadiabatic Force Matching for Alchemical Free-Energy Estimation

We propose a method to compute free-energy differences from nonadiabatic alchemical transformations by using flow-based generative models. The method, nonadiabatic force matching, hinges on estimating the dissipation along an alchemical switching process in terms of a nonadiabatic force field that can be learned through stochastic flow matching. The learned field can be used in conjunction with short-time trajectory data to evaluate upper and lower bounds on the alchemical free energy that variationally converge to the exact value if the field is optimal. Applying the method to evaluate the alchemical free energy of atomistic models shows that it can substantially reduce the simulation cost of a free-energy estimate at a negligible loss of accuracy when compared with thermodynamic integration.

Computational chemistry↗

Influence of grain boundary characteristics on thermal stability in nanotwinned copper

High density grain boundaries provide high strength, but may introduce undesirable features, such as high Fermi levels and instability. We investigated the kinetics of recovery and recrystallization of Cu that was manufactured to include both nanotwins (NT) and high-angle columnar boundaries. We used the isothermal Johnson-Mehl-Avrami-Kolmogorov (JMAK) model to estimate activation energy values for recovery and recrystallization and compared those to values derived using the non-isothermal Kissinger equation. The JMAK model hinges on an exponent that expresses the growth mechanism of a material. The exponent for this Cu was close to 0.5, indicating low-dimensional microstructure evolution, which is associated with anisotropic twin coarsening, heterogeneous recrystallization, and high stability. Since this Cu was of high purity, there was a negligible impurity-drag-effect on boundaries. The twin coarsening and heterogeneous recrystallization resulted from migration of high-angle columnar boundaries with their triple junctions in one direction, assisted by the presence of high concentration vacancies at boundaries. Analyses performed by electron energy loss spectroscopy of atomic columns at twin boundaries (TBs) and in the interior showed similar plasma peak shapes and L3 edge positions. As a result, this implies that values for conductivity and Fermi level are equal for atoms at TBs and in the interior.

36 MATERIALS SCIENCE↗