RNAGCN: RNA tertiary structure assessment with a graph convolutional network

RNAs play crucial and versatile roles in cellular biochemical reactions. Since experimental approaches of determining their three-dimensional (3D) structures are costly and less efficient, it is greatly advantageous to develop computational methods to predict RNA 3D structures. For these methods, designing a model or scoring function for structure quality assessment is an essential step but this step poses challenges. In this study, we designed and trained a deep learning model to tackle this problem. The model was based on a graph convolutional network (GCN) and named RNAGCN. The model provided a natural way of representing RNA structures, avoided complex algorithms to preserve atomic rotational equivalence, and was capable of extracting features automatically out of structural patterns. Testing results on two datasets convincingly demonstrated that RNAGCN performs similarly to or better than four leading scoring functions. Our approach provides an alternative way of RNA tertiary structure assessment and may facilitate RNA structure predictions. RNAGCN can be downloaded from https://gitee.com/dcw-RNAGCN/rnagcn.     

A general stability result for the semilinear viscoelastic wave model under localized effects

Our main goal in the present work is to address an integro-differential model under localized viscoelastic and frictional effects arising in the Boltzmann theory of viscoelasticity. More precisely, we consider a general version in the history context of the pioneer localized viscoelastic problem approached by Cavalcanti and Oquendo (2003) in the null history scenario, and more recently by Cavalcanti et al. (2018) in the history framework. By means of a new observability inequality, we prove a general stability result to the model under a weaker assumption on the localized frictional damping and a slower condition on the decreasing memory kernel (of polynomial type) than the previously mentioned works. To achieve such stability results, we still work in a general setting by removing the assumption on complementary damping mechanisms and show, in some reasonable situations concerning the density coefficient, that the localized viscoelastic effect is enough to ensure the general stability (of polynomial type) to the problem.     

Uniform stability of semilinear wave equations with arbitrary local memory effects versus frictional dampings

This paper is concerned with the mixed initial–boundary value problem for semilinear wave equations with complementary frictional dampings and memory effects. We successfully establish uniform exponential and polynomial decay rates for the solutions to this initial–boundary value problem under much weak conditions concerning memory effects. More specifically, we obtain the exponential and polynomial decay rates after removing the fundamental condition that the memory-effect region includes a part of the system boundary, while the condition is a necessity in the previous literature; moreover, for the polynomial decay rates we only assume minimal conditions on the memory kernel function g, without the usual assumption of $g^{\prime }$ controlled by g.     

The observation of inherent spin Seebeck effect in Rh/YIG hybrid structure

Rhodium (Rh) is a 4d metal possessing a large spin orbit coupling strength and spin-Hall conductivity with a very small magnetic susceptibility, implying an insignificant magnetic proximity effect (MPE). We report here the observation of longitudinal spin Seebeck effect (LSSE) using Rh as a normal metal. A Rh film was sputtered on nanometer thick YIG films of highly crystalline nature and extremely low magnetic damping to obtain Rh/YIG hybrid structure. A clear thermal voltage V_th (SSE voltage) was obtained when a temperature gradient was applied on the Rh/YIG hybrid. The Rh film showed a very weak anomalous Hall resistance and the magneto-resistive testing clearly ruled out the magnetization of the Rh films via MPE. The anisotropic magnetoresistance (AMR) revealed a clear spin hall magnetoresistance (SMR) signal in Rh film implying a purely intrinsic spin current generation, free from any parasitic magnetic effects. The work can open a new window in the study of pure and uncontaminated spin current, generated in ferromagnetic insulators, using Rh as spin current detector.     

Semiotic Aggregation in Deep Learning

Convolutional neural networks utilize a hierarchy of neural network layers. The statistical aspects of information concentration in successive layers can bring an insight into the feature abstraction process. We analyze the saliency maps of these layers from the perspective of semiotics, also known as the study of signs and sign-using behavior. In computational semiotics, this aggregation operation (known as superization) is accompanied by a decrease of spatial entropy: signs are aggregated into supersign. Using spatial entropy, we compute the information content of the saliency maps and study the superization processes which take place between successive layers of the network. In our experiments, we visualize the superization process and show how the obtained knowledge can be used to explain the neural decision model. In addition, we attempt to optimize the architecture of the neural model employing a semiotic greedy technique. To the extent of our knowledge, this is the first application of computational semiotics in the analysis and interpretation of deep neural networks.     

Semilinear wave models with friction and viscoelastic damping

In this paper, we study the global (in time) existence of small data solutions to the Cauchy problem for the semilinear wave equation with friction, viscoelastic damping, and a power nonlinearity. We are interested in the connection between regularity assumptions for the data and the admissible range of exponents p in the power nonlinearity.     

核电结构土-结相互作用分析分区混合计算方法

土-结构相互作用分析是核电结构进行抗震设计和安全评估的重要环节.在核电结构的土-结相互作用分析中,阻尼和非线性是影响结构反应的重要因素. 若采用频域分析,可以方便考虑阻尼,但需通过等效线性化来考虑非线性,不适合于强震作用下的土体非线性.若采用时域分析的逐步积分方法,适合于考虑非线性,但材料阻尼一般采用瑞利阻尼模型,除了紧靠指定阻尼比的少数几个振型外,其他振型的反应将受到瑞利阻尼模型所确定的大阻尼所抑制,造成地震反应与真实情形有较大差异.若采用时域分析的模态叠加法,可合理计入阻尼效应,但模态叠加法不能考虑非线性.因此,如何合理考虑阻尼和非线性是核电结构土-结相互作用分析需要关注的问题.基于此,本文提出一种模态叠加和时步积分结合的土-结相互作用分区算法.其中,出于安全性考虑,地震作用下核电主体结构一般不允许进入非线性,因此结构可采用模态叠加方法,以便合理考虑结构阻尼;土体和基础采用显式时步积分法,可考虑土体非线性;通过人工边界条件考虑无限域的影响 (辐射阻尼).通过简单算例对该方法进行了验证,并用于CAP1400核电结构的土-结相互作用分析中,对比分析了采用模态阻尼和瑞利阻尼时核电结构和场地反应的差异,结果表明结构阻尼模型对场地的反应影响不大,但对结构反应影响明显,在实际工程中应合理选取阻尼模型.      

Nonexistence of global solutions of wave equations with weak time-dependent damping and combined nonlinearity

In our previous two works, we studied the blow-up and lifespan estimates for damped wave equations with a power nonlinearity of the solution or its derivative, with scattering damping independently. In this work, we are devoted to establishing a similar result for a combined nonlinearity. Comparing to the result of wave equation without damping, one can say that the scattering damping has no influence.     

基于级联卷积神经网络的前列腺磁共振图像分类

针对深度学习训练成本高，以及基于磁共振图像的前列腺癌临床诊断需要大量医学常识且极为耗时的问题，本文提出了一种基于级联卷积神经网络（Convolutional Neural Network，CNN）和磁共振图像的前列腺癌（Prostate Cancer，PCa）自动分类诊断方法，该网络以Faster-RCNN作为前网络，对前列腺区域进行提取分割，用于排除前列腺附近组织器官的干扰；以基于ResNet改进的网络结构CNN40_bottleneck作为后网络，用于对前列腺区域病变进行分类.后网络由瓶颈结构串联组成，其中使用批量标准化（Batch Normalization，BN）、全局平均池化（Global Average Pooling，GAP）进行优化.实验结果证明，本文方法对前列腺癌诊断结果较好，而且缩减了训练时间和参数量，有效降低了训练成本.