Hyperparameter on-line learning of stochastic resonance based threshold networks

Aiming at training the feed-forward threshold neural network consisting of nondifferentiable activation functions, the approach of noise injection forms a stochastic resonance based threshold network that can be optimized by various gradient-based optimizers. The introduction of injected noise extends the noise level into the parameter space of the designed threshold network, but leads to a highly non-convex optimization landscape of the loss function. Thus, the hyperparameter on-line learning procedure with respective to network weights and noise levels becomes of challenge. It is shown that the Adam optimizer, as an adaptive variant of stochastic gradient descent, manifests its superior learning ability in training the stochastic resonance based threshold network effectively. Experimental results demonstrate the significant improvement of performance of the designed threshold network trained by the Adam optimizer for function approximation and image classification.     

Synthesis of Small‐Sized,Porous, and Low‐Toxic Magnetite Nanoparticles by Thin POSS Silica Coating

In this communication, we report the synthesis of small‐sized (<10 nm), water‐soluble, magnetic nanoparticles (MNPs) coated with polyhedral oligomeric silsesquioxanes (POSS), which contain either polyethylene glycol (PEG) or octa(tetramethylammonium) (OctaTMA) as functional groups. The POSS‐coated MNPs exhibit superparamagnetic behavior with saturation magnetic moments (51–53 emu g^?1) comparable to silica‐coated MNPs. They also provide good colloidal stability at different pH and salt concentrations, and low cytotoxicity to MCF‐7 human breast epithelial cells. The relaxivity data and magnetic resonance (MR) phantom images demonstrate the potential application of these MNPs in bioimaging.     

Close‐Packed Two‐Dimensional Silver Nanoparticle Arrays: Quadrupolar and Dipolar Surface Plasmon Resonance Coupling

Silver nanoparticles (NPs) ranging in size from 40 to 100 nm were prepared in high yield by using an improved seed‐mediated method. The homogeneous Ag NPs were used as building blocks for 2D assembled Ag NP arrays by using an oil/water interface. A close‐packed 2D array of Ag NPs was fabricated by using packing molecules (3‐mercaptopropyltrimethoxysilane) to control the interparticle spacing. The homogeneous 2D Ag NP array exhibited a strong quadrupolar cooperative plasmon mode resonance and a dipolar red‐shift relative to individual Ag NPs suspended in solution. A well‐arranged 2D Ag NP array was embedded in polydimethylsiloxane film and, with biaxial stretching to control the interparticle distance, concomitant variations of the quadrupolar and dipolar couplings were observed. As the interparticle distance increased, the intensity of the quadrupolar cooperative plasmon mode resonance decreased and dipolar coupling completely disappeared. The local electric field of the 2D Ag NP array was calculated by using finite difference time domain simulation and qualitatively showed agreement with the experimental measurements.     

Local linear convergence analysis of Primal–Dual splitting methods

In this paper, we study the local linear convergence properties of a versatile class of Primal–Dual splitting methods for minimizing composite non-smooth convex optimization problems. Under the assumption that the non-smooth components of the problem are partly smooth relative to smooth manifolds, we present a unified local convergence analysis framework for these methods. More precisely, in our framework, we first show that (i) the sequences generated by Primal–Dual splitting methods identify a pair of primal and dual smooth manifolds in a finite number of iterations, and then (ii) enter a local linear convergence regime, which is characterized based on the structure of the underlying active smooth manifolds. We also show how our results for Primal–Dual splitting can be specialized to cover existing ones on Forward–Backward splitting and Douglas–Rachford splitting/ADMM (alternating direction methods of multipliers). Moreover, based on these obtained local convergence analysis result, several practical acceleration techniques are discussed. To exemplify the usefulness of the obtained result, we consider several concrete numerical experiments arising from fields including signal/image processing, inverse problems and machine learning. The demonstration not only verifies the local linear convergence behaviour of Primal–Dual splitting methods, but also the insights on how to accelerate them in practice.     

A simple three‐wave approximate Riemann solver for the Saint‐Venant‐Exner equations

Erosion and sediments transport processes have a great impact on industrial structures and on water quality. Despite its limitations, the Saint‐Venant‐Exner system is still (and for sure for some years) widely used in industrial codes to model the bedload sediment transport. In practice, its numerical resolution is mostly handled by a splitting technique that allows a weak coupling between hydraulic and morphodynamic distinct softwares but may suffer from important stability issues. In recent works, many authors proposed alternative methods based on a strong coupling that cure this problem but are not so trivial to implement in an industrial context. In this work, we then pursue 2 objectives. First, we propose a very simple scheme based on an approximate Riemann solver, respecting the strong coupling framework, and we demonstrate its stability and accuracy through a number of numerical test cases. However, second, we reinterpret our scheme as a splitting technique and we extend the purpose to propose what should be the minimal coupling that ensures the stability of the global numerical process in industrial codes, at least, when dealing with collocated finite volume method. The resulting splitting method is, up to our knowledge, the only one for which stability properties are fully demonstrated.     

An investigation on the defect structures and spin Hamiltonian parameters for the two orthorhombic Ti3+ centers in ZnWO4

By employing the perturbation formulae of the spin Hamiltonian parameters (SHPs) (g factors g_xx, g_yy, g_zz, hyperfine structure constants A_xx, A_yy, A_zz and superhyperfine parameters A_xx׳, A_yy׳, A_zz׳) for a 3d¹ ion in orthorhombically elongated octahedra and tetrahedra, the defect structures and the experimental EPR spectra are theoretically and systematically investigated for the two orthorhombic Ti³⁺ centers C₁ and C₂ in ZnWO₄. Center C₁ is ascribed to the impurity Ti³⁺ at host W⁶⁺ site associated with two nearest neighbor oxygen vacancies due to charge compensation. The resultant tetrahedral [TiO₄]^5– cluster is determined to undergo the local orthorhombic elongation distortion, characterized by the axial distortion angle Δθ (=θ–θ₀≈–6.84°) of the local impurity-ligand bond angle θ related to θ₀ (≈54.74°) and the perpendicular distortion angle Δε (=ε–ε₀≈2.5°) related to ε₀ (≈45°) of an ideal tetrahedron because of the Jahn–Teller effect. Center C₂ is attributed to Ti³⁺ on Zn²⁺ site, and this octahedral [TiO₆]^9– cluster may experience the local axial elongation ΔZ (≈0.001 Ǻ) and the planar bond angle variation Δφ (≈9.1°) due to the Jahn–Teller effect, resulting in a more regular oxygen octahedron. All the calculated SHPs (i.e., g factors for both centers, the hyperfine structure constants for center C₂ and superhyperfine parameters of next nearest neighbor ligand W for center C₁) show good agreement with the observed values. However, the theoretical results based on the previous assignment of center C₁ as Ti³⁺ on W⁶⁺ site with only one nearest planar oxygen vacancy (i.e., five-fold coordinated octahedral [TiO₅]^7– cluster) show much worse agreement with the experimental data. The defect structures and the SHPs (especially the g anisotropies) are discussed for both centers. The present studies on the superhyperfine parameters of ligand W⁶⁺ for center C₁ would be helpful to further investigations on the superhyperfine interactions of cation ligands which were rather scarcely treated before.     

紫色球杆菌视紫红质光谱特性的机器学习研究

近年来，机器学习等人工智能技术被应用于蛋白质工程，其在蛋白质结构、功能预测、催化活性等研究中具有独特优势。在未知蛋白质结构的情况下，将蛋白质序列和功能特性与机器学习相结合，基于序列-活性关系（innovative sequence-activity relationship，ISAR）算法，将蛋白质氨基酸序列数字化，用快速傅里叶变换（fast four transform，FFT）进行预处理，再进行偏最小二乘回归建模，可在数据集较少情况下拟合得到最佳模型。通过机器学习对紫色球杆菌视紫红质（gloeobacter violaceus rhodopsin，GR）的突变体蛋白质氨基酸序列与光谱最大吸收波长进行建模，获得了最佳模型。用最佳索引LEVM760106建模得到的确定系数R² 为0.944，均方误差E为11.64。用小波变换进行的预处理，其R² 虽也约为0.944，但E大于11.64，不及FFT进行的预处理。方法较好地解决了蛋白质序列与功能特性之间的数学建模问题，在蛋白质工程中可为预测更优的突变体提供支持。     

A paper-based resonance energy transfer nucleic acid hybridization assay using upconversion nanoparticles as donors and quantum dots as acceptors

Monodisperse aqueous upconverting nanoparticles (UCNPs) were covalently immobilized on aldehyde modified cellulose paper via reduction amination to develop a luminescence resonance energy transfer (LRET)-based nucleic acid hybridization assay. This first account of covalent immobilization of UCNPs on paper for a bioassay reports an optically responsive method that is sensitive, reproducible and robust. The immobilized UCNPs were decorated with oligonucleotide probes to capture HPRT1 housekeeping gene fragments, which in turn brought reporter conjugated quantum dots (QDs) in close proximity to the UCNPs for LRET. This sandwich assay could detect unlabeled oligonucleotide target, and had a limit of detection of 13 fmol and a dynamic range spanning nearly 3 orders of magnitude. The use of QDs, which are excellent LRET acceptors, demonstrated improved sensitivity, limit of detection, dynamic range and selectivity compared to similar assays that have used molecular fluorophores as acceptors. The selectivity of the assay was attributed to the decoration of the QDs with polyethylene glycol to eliminate non-specific adsorption. The kinetics of hybridization were determined to be diffusion limited and full signal development occurred within 3 min.     

ParaHydrogen Polarized Ethyl-[1-13C]pyruvate in Water,a Key Substrate for Fostering the PHIP-SAH Approach to Metabolic Imaging

An efficient synthesis of vinyl-[1-¹³C]pyruvate has been reported, from which ¹³C hyperpolarized (HP) ethyl-[1-¹³C]pyruvate has been obtained by means of ParaHydrogen Induced Polarization (PHIP). Due to the intrinsic lability of pyruvate, which leads quickly to degradation of the reaction mixture even under mild reaction conditions, the vinyl-ester has been synthesized through the intermediacy of a more stable ketal derivative. ¹³C and ¹H hyperpolarizations of ethyl-[1-¹³C]pyruvate, hydrogenated using ParaHydrogen, have been compared to those observed on the more widely used allyl-derivative. It has been demonstrated that the spin order transfer from ParaHydrogen protons to ¹³C, is more efficient on the ethyl than on the allyl-esterdue to the larger J-couplings involved. The main requirements needed for the biological application of this HP product have been met, i. e. an aqueous solution of the product at high concentration (40 mM) with a good ¹³C polarization level (4.8 %) has been obtained. The in vitro metabolic transformation of the HP ethyl-[1-¹³C]pyruvate, catalyzed by an esterase, has been observed. This substrate appears to be a good candidate for in vivo metabolic investigations using PHIP hyperpolarized probes.     

A molecular fluorophore in citric acid/ethylenediamine carbon dots identified and quantified by multinuclear solid-state nuclear magnetic resonance

The composition of fluorescent polymer nanoparticles, commonly referred to as carbon dots, synthesized by microwave-assisted reaction of citric acid and ethylenediamine was investigated by ¹³C, ¹³C{¹H}, ¹H─¹³C, ¹³C{¹⁴N}, and ¹⁵N solid-state nuclear magnetic resonance (NMR) experiments. ¹³C NMR with spectral editing provided no evidence for significant condensed aromatic or diamondoid carbon phases. ¹⁵N NMR showed that the nanoparticle matrix has been polymerized by amide and some imide formation. Five small, resolved ¹³C NMR peaks, including an unusual ═CH signal at 84 ppm (¹H chemical shift of 5.8 ppm) and ═CN₂ at 155 ppm, and two distinctive ¹⁵N NMR resonances near 80 and 160 ppm proved the presence of 5-oxo-1,2,3,5-tetrahydroimidazo[1,2-a]pyridine-7-carboxylic acid (IPCA) or its derivatives. This molecular fluorophore with conjugated double bonds, formed by a double cyclization reaction of citric acid and ethylenediamine as first shown by Y. Song, B. Yang, and coworkers in 2015, accounts for the fluorescence of the carbon dots. Cross-peaks in a ¹H─¹³C HETCOR spectrum with brief ¹H spin diffusion proved that IPCA is finely dispersed in the polyamide matrix. From quantitative ¹³C and ¹⁵N NMR spectra, a high concentration (18 ± 2 wt%) of IPCA in the carbon dots was determined. A pronounced gradient in ¹³C chemical-shift perturbations and peak widths, with the broadest lines near the COO group of IPCA, indicated at least partial transformation of the carboxylic acid of IPCA by amide or ester formation.