Moving Multiscale Modelling to the Edge: Benchmarking and Load Optimization for Cellular Automata on Low Power Microcomputers

Numerical computations are usually associated with the High Performance Computing. Nevertheless, both industry and science tend to involve devices with lower power in computations. This is especially true when the data collecting devices are able to partially process them at place, thus increasing the system reliability. This paradigm is known as Edge Computing. In this paper, we propose the use of devices at the edge, with lower computing power, for multi-scale modelling calculations. A system was created, consisting of a high-power device—a two-processor workstation, 8 RaspberryPi 4B microcomputers and 8 NVidia Jetson Nano units, equipped with GPU processor. As a part of this research, benchmarking was performed, on the basis of which the computational capabilities of the devices were classified. Two parameters were considered: the number and performance of computing units (CPUs and GPUs) and the energy consumption of the loaded machines. Then, using the calculated weak scalability and energy consumption, a min–max-based load optimization algorithm was proposed. The system was tested in laboratory conditions, giving similar computation time with same power consumption for 24 physical workstation cores vs. 8x RaspberryPi 4B and 8x Jetson Nano. The work ends with a proposal to use this solution in industrial processes on example of hot rolling of flat products.

EAR-Net: Efficient Atrous Residual Network for Semantic Segmentation of Street Scenes Based on Deep Learning

Segmentation of street scenes is a key technology in the field of autonomous vehicles. However, conventional segmentation methods achieve low accuracy because of the complexity of street landscapes. Therefore, we propose an efficient atrous residual network (EAR-Net) to improve accuracy while maintaining computation costs. First, we performed feature extraction and restoration, utilizing depthwise separable convolution (DSConv) and interpolation. Compared with conventional methods, DSConv and interpolation significantly reduce computation costs while minimizing performance degradation. Second, we utilized residual learning and atrous spatial pyramid pooling (ASPP) to achieve high accuracy. Residual learning increases the ability to extract context information by preventing the problem of feature and gradient losses. In addition, ASPP extracts additional context information while maintaining the resolution of the feature map. Finally, to alleviate the class imbalance between the image background and objects and to improve learning efficiency, we utilized focal loss. We evaluated EAR-Net on the Cityscapes dataset, which is commonly used for street scene segmentation studies. Experimental results showed that the EAR-Net had better segmentation results and similar computation costs as the conventional methods. We also conducted an ablation study to analyze the contributions of the ASPP and DSConv in the EAR-Net.

Spectrum of end of the world branes in holographic BCFTs

We study overlaps between two regularized boundary states in conformal field theories. Regularized boundary states are dual to end of the world branes in an AdS black hole via the AdS/BCFT. Thus they can be regarded as microstates of a single sided black hole. Owing to the open-closed duality, such an overlap between two different regularized boundary states is exponentially suppressed as $$ \left\langle \left.{\psi}_a\right|{\psi}_b\right\rangle \sim {e}^{-O\left({h}_{ab}^{\left(\min \right)}\right)} $$ ψ a ψ b ∼ e − O h ab min , where $$ {h}_{ab}^{\left(\min \right)} $$ h ab min is the lowest energy of open strings which connect two different boundaries a and b. Our gravity dual analysis leads to $$ {h}_{ab}^{\left(\min \right)} $$ h ab min = c/24 for a pure AdS3 gravity. This shows that a holographic boundary state is a random vector among all left-right symmetric states, whose number is given by a square root of the number of all black hole microstates. We also perform a similar computation in higher dimensions, and find that $$ {h}_{ab}^{\left(\min \right)} $$ h ab min depends on the tensions of the branes. In our analysis of holographic boundary states, the off diagonal elements of the inner products can be computed directly from on-shell gravity actions, as opposed to earlier calculations of inner products of microstates in two dimensional gravity.

Assessing the Evolution of Transit User Behavior from Smart Card Data

There is a huge potential for exploiting information centered on individual transit users’ behavior through longitudinal smart card data. This is particularly true for cities like Gatineau, Canada, where the bus system serves passengers with different travel patterns. Understanding the evolution of these patterns marks an important point in improving transit demand forecasting models. Indeed, better models can help transit planners to create optimized networks. This paper proposes a comparison of a traditional and an experimental methodology aiming to identify the evolution of travel structure among transit users. These methodologies are based on the clustering of multi-week travel patterns derived from a large sample of smart card transactions (35.4 million). Representing users’ behavior, these patterns are constructed using the number of trips made by every card on each day of a week. Behavior vectors are defined by seven components (one for each day) and are clustered using a K-means algorithm. The experimental week-to-week method consists in clustering the population on each week, while using the clustering results from the previous week as seed. This latter approach makes it possible to observe the evolution of users’ behaviors and also has a better clustering quality in a similar computation time than the traditional method.

Simplified Method for Elastic Distortional Buckling Stress of Cold-Formed Thin-Wall Steel Members

For the distortional buckling of cold-formed thin-wall steel members, the Direct Strength Method (DSM) is a recently adopted design approach by foreign standard, and Chinese professional standard Technical specification for low-rise cold-formed thin-wall steel buildings present the calculation formulas based on the Effective Area Method (EAM) .The key of both methods is the elastic distortional buckling stress .At present, the calculation methods about it are over conservative .Compare the calculation results of method in AISI S100with the one calculated by software CUFSM ,for C and Z sections in bending and axial compression loading respectively in the North American Specifications for Cold-Formed Steel Structural Members (AISI S100). This paper presents revised simplified methods for calculating the elastic distortional buckling stress. The new methods yield more accurate results but similar computation cost compared to the existing methods.

An ‘ ab initio ’ Gaussian orbital calculation of the (100) surface of crystalline lithium hydride

An ab initio computation has been performed for the (100) face of a lithium hydride ionic crystal. The computation follows a similar computation performed earlier for an infinite crystal of lithium hydride. A simple wavefunction of a type proposed by A. A. Frost was used in which pairs of electrons are assigned to orbitals described by simple spherical Gaussian functions, the positions and sizes of which are allowed to float to minimize the energy. In the surface computation the nuclei and the Gaussian orbitals in the ultimate and penultimate layers were allowed to float freely, those in the remaining layers being disposed as in the infinite crystal. It was found that, in the surface layer, the hydrogen nuclei were farther from the fixed (third) layer than the lithium nuclei. Also, while the orbitals of the lithium ions remained centred at the same place as the nuclei, corresponding to zero polarization of the lithium ions, the hydride ions showed considerable polarization, the centres of the orbitals being displaced considerably from the nuclei. The changes in the penultimate layer are also discussed. An estimate was made of the surface energy for this face of lithium hydride and the value found was reasonable when compared with semi-empirical values for the (100) surfaces of alkali halide crystals.