TopoAct: Visually Exploring the Shape of Activations in Deep Learning

Deep neural networks such as GoogLeNet, ResNet, and BERT have achieved impressive performance in tasks such as image and text classification. To understand how such performance is achieved, we probe a trained deep neural network by studying neuron activations, i.e.combinations of neuron firings, at various layers of the network in response to a particular input. With a large number of inputs, we aim to obtain a global view of what neurons detect by studying their activations. In particular, we develop visualizations that show the shape of the activation space, the organizational principle behind neuron activations, and the relationships of these activations within a layer. Applying tools from topological data analysis, we present TopoAct , a visual exploration system to study topological summaries of activation vectors. We present exploration scenarios using TopoAct that provide valuable insights into learned representations of neural networks. We expect TopoAct to give a topological perspective that enriches the current toolbox of neural network analysis, and to provide a basis for network architecture diagnosis and data anomaly detection.     

Surface modification technique to enhance metallurgical and mechanical properties of alloy C-276 weldment by laser shock peening without coating

Microsegregation of alloying elements is prone to hot cracking in the weldment of alloy C-276. The formation of topologically close packed phases P and µ is largely responsible for the hot cracking. The present study articulates the effect of laser shock peening (LSP) to improve the metallurgical and mechanical properties of the weld joint. The weld joint was fabricated by pulsed current gas tungsten arc welding (PCGTAW) using an ERNiCrMo-3 filler wire. LSP without coating was carried out on the cap surface of the weldment. Microstructural studies were carried out to compare the as-welded and laser-peened microstructure on the fusion zone. The results show that a fine equiaxed dendritic structure was observed in both conditions. EDS analysis was carried out to evaluate the microsegregation of alloying elements. EDS analysis indicates that there are no secondary intermetallic phases. X-ray diffraction analysis was carried out to evaluate the phase change and crystallite size in the as-welded and laser shock peened fusion zone. The result shows 48.99% reduction in crystallite size after LSP. Hardness and tensile strength results indicate there is a consequential increase in laser shock peened specimen compared with as-welded specimen.     

Hot Corrosion Behaviour of Refractory and Rare Earth Oxide Reinforced CoCrAlY APS Coatings at 700 °C

This paper investigates cyclic hot corrosion of plasma sprayed CoCrAlY?+?Al₂O₃?+?YSZ (C1) and CoCrAlY?+?CeO₂ (C2) composite coatings on MDN 321 and Superni 76 substrates in molten salt (Na₂SO₄-60%V₂O₅) environment exposed to 700 °C. Weight change technique is used to evaluate the corrosion performance. Both C1 and C2 coatings showed better corrosion resistance than uncoated alloy. Both the coatings showed linear weight gain during the initial cycles and parabolic weight gain nature with subsequent hot corrosion cycles. The parabolic rate constant (K_p) of C1 and C2 coating was observed to be in the range 0.29–0.32?×?10^?10 g² cm^?4 s^?1 and 1.0–1.13?×?10^?10 g² cm^?4 s^?1 respectively. In C1 coating, the globular and continuously packed structure on the corroded surface having CoO, Cr₂O₃, CoCr₂O₄ and CoAl₂O₄ spinel oxides provided superior hot corrosion resistance. While in case of C2 coating, the outward growth of CeVO₄ irregular crystals as a corrosion product of CeO₂ and V₂O₅ salt deteriorated the oxide scales resulting in higher corrosion rate.     

Molecular dynamics modeling of thermal conductivity enhancement in metal nanoparticle suspensions

A theoretical approach based on molecular dynamics modeling, for the estimation of the enhancement of the thermal conductivity of liquids by the introduction of suspended metallic nanoparticles is proposed. Algorithms are developed for simulating the nanofluid abiding the procedural steps of the molecular dynamics method. The method is presented as a solution to the generic problem of thermal conductivity enhancement of liquids in the presence of nanoparticles, and illustrated using a specific simulation procedure with properties representing water and platinum nanoparticles. The thermal conductivity enhancement estimated using the simulations are compared with existing experimental results and those predicted by conventional effective medium theories. Parametric studies are conducted to obtain the variation of thermal conductivity enhancement with the temperature and the volume fraction of the nanoparticles in the suspension.     

Contact-Barrier Free,High Mobility,Dual-Gated Junctionless Transistor Using Tellurium Nanowire

The gate-all-around nanowire transistor, due to its extremely tight electrostatic control and vertical integration capability, is a highly promising candidate for sub-5 nm technology nodes. In particular, the junctionless nanowire transistors are highly scalable with reduced variability due to avoidance of steep source/drain junction formation by ion implantation. Here a dual-gated junctionless nanowire p-type field effect transistor is demonstrated using tellurium nanowire as the channel. The dangling-bond-free surface due to the unique helical crystal structure of the nanowire, coupled with an integration of dangling-bond-free, high quality hBN gate dielectric, allows for a phonon-limited field effect hole mobility of 570 cm² V⁻¹ s⁻¹ at 270 K, which is well above state-of-the-art strained Si hole mobility. By lowering the temperature, the mobility increases to 1390 cm² V⁻¹ s⁻¹ and becomes primarily limited by Coulomb scattering. The combination of an electron affinity of ≈ 4 eV and a small bandgap of tellurium provides zero Schottky barrier height for hole injection at the metal-contact interface, which is remarkable for reduction of contact resistance in a highly scaled transistor. Exploiting these properties, coupled with the dual-gated operation, we achieve a high drive current of 216 μA μm⁻¹ while maintaining an on-off ratio in excess of 2 × 10⁴. The findings have intriguing prospects for alternate channel material based next-generation electronics.     

Molecular Insights on the Possible Role of Annexin A2 in COVID-19 Pathogenesis and Post-Infection Complications

Severe acute respiratory syndrome-coronavirus 2 (SARS-CoV-2) has infected >235 million people and killed over 4.8 million individuals worldwide. Although vaccines have been developed for prophylactic management, there are no clinically proven antivirals to treat the viral infection. Continuous efforts are being made all over the world to develop effective drugs but these are being delayed by periodic outbreak of mutated SARS-CoV-2 and a lack of knowledge of molecular mechanisms underlying viral pathogenesis and post-infection complications. In this regard, the involvement of Annexin A2 (AnxA2), a lipid-raft related phospholipid-binding protein, in SARS-CoV-2 attachment, internalization, and replication has been discussed. In addition to the evidence from published literature, we have performed in silico docking of viral spike glycoprotein and RNA-dependent RNA polymerase with human AnxA2 to find the molecular interactions. Overall, this review provides the molecular insights into a potential role of AnxA2 in the SARS-CoV-2 pathogenesis and post-infection complications, especially thrombosis, cytokine storm, and insulin resistance.     

Characterization of the Hot Deformation Behavior and Microstructure Evolution of a New γ-γ′ Strengthened Cobalt-Based Superalloy

Recently, cobalt-based γ-γ′ microstructured superalloys have attracted attention. However, studies on their processing behavior [i.e., processing maps (the variation of strain rate sensitivity (m) with temperature)] are limited. Thus, the high-temperature flow behavior of a γ-γ′ Co-30Ni-10Al-5Mo-2Ta-2Ti-5Cr (at. pct) superalloy was investigated using isothermal compression tests between 1348 and 1498 K at strain rates from 0.001 to 10 s⁻¹. The m contour map was generated using the experimental flow stress values, which were used to locate the optimum hot workability and desired microstructural processing range. A strong dependence of m on the deformation parameters (temperature, strain rate, and strain) was observed. A maximum m value of around 0.3 at 1460 K to 1498 K and strain rates of 0.01 to 0.5 s⁻¹ was found. The deformed samples show a fully recrystallized microstructure at high m. Unstable domains showed the formation of cavities at the grain boundary triple points and cracks along the grain boundaries at high strain rates (1 to 10 s⁻¹), corresponding to m < 0.10. A constitutive model was developed using an Arrhenius hyperbolic sine function, yielding an apparent activation energy of 540 ± 30 kJ mol⁻¹ for hot deformation. This study indicates reasonable formability under certain conditions below the solvus, thus opening possibilities for further thermomechanical treatment.

     

微流道交流电水力泵的数值模拟及优化

描述了微通道内交流电水力泵(EHD)的工作原理,并用IntelliSuite(R)的Microfluidic模块对电水力泵进行数值模拟.并分析了影响电水力泵速度场分布的因素,包括流体电导率、驱动交流电频率和相位、电极维度以及电极间隙和数量等.通过数值模拟不同参数下的电水力泵,比较驱动压力大小,最后给出优化的设计结果,为给电水力泵设计参数的选择提供依据.本文对流道长度为5mm,宽0.4mm,厚度0.4mm的电水力泵做了优化设计的数值模拟,最后得到驱动电极宽度为0.01mm,电极间距为0.02mm.     

Computational Pipeline for Reference-Free Comparative Analysis of RNA 3D Structures Applied to SARS-CoV-2 UTR Models

RNA is a unique biomolecule that is involved in a variety of fundamental biological functions, all of which depend solely on its structure and dynamics. Since the experimental determination of crystal RNA structures is laborious, computational 3D structure prediction methods are experiencing an ongoing and thriving development. Such methods can lead to many models; thus, it is necessary to build comparisons and extract common structural motifs for further medical or biological studies. Here, we introduce a computational pipeline dedicated to reference-free high-throughput comparative analysis of 3D RNA structures. We show its application in the RNA-Puzzles challenge, in which five participating groups attempted to predict the three-dimensional structures of 5′- and 3′-untranslated regions (UTRs) of the SARS-CoV-2 genome. We report the results of this puzzle and discuss the structural motifs obtained from the analysis. All simulated models and tools incorporated into the pipeline are open to scientific and academic use.