Optical Multi-Imaging-Casting Accelerator for Fully Parallel Universal Convolution Computing

Over the last few years, optical computing has become a potential solution to computationally heavy convolution, aimed at accelerating various artificial intelligence applications. However, past schemes have never efficiently realized fully parallel optical convolution. Here, we propose a new paradigm for a universal convolution accelerator with truly massive parallelism and high precision based on optical multi-imaging-casting architecture. Specifically, a two-dimensional Dammann grating is adopted for the generation of multiple displaced images of the kernel, which is the core process for kernel sliding on the convolved matrix. Our experimental results indicate that the computing accuracy is typically close to 8-bit, and this accuracy can be improved further by using hybrid analog–digital coding method. In addition, a convolutional neural network for the standard MNIST dataset is demonstrated, and the recognition accuracy for inference is up to 97.3%. The paradigm reported here will open new opportunities for high-throughput universal convolution accelerators for real-time or quasi-real-time AI applications.

Assessment of OpenMP Master–Slave Implementations for Selected Irregular Parallel Applications

The paper investigates various implementations of a master–slave paradigm using the popular OpenMP API and relative performance of the former using modern multi-core workstation CPUs. It is assumed that a master partitions available input into a batch of predefined number of data chunks which are then processed in parallel by a set of slaves and the procedure is repeated until all input data has been processed. The paper experimentally assesses performance of six implementations using OpenMP locks, the tasking construct, dynamically partitioned for loop, without and with overlapping merging results and data generation, using the gcc compiler. Two distinct parallel applications are tested, each using the six aforementioned implementations, on two systems representing desktop and worstation environments: one with Intel i7-7700 3.60GHz Kaby Lake CPU and eight logical processors and the other with two Intel Xeon E5-2620 v4 2.10GHz Broadwell CPUs and 32 logical processors. From the application point of view, irregular adaptive quadrature numerical integration, as well as finding a region of interest within an irregular image is tested. Various compute intensities are investigated through setting various computing accuracy per subrange and number of image passes, respectively. Results allow programmers to assess which solution and configuration settings such as the numbers of threads and thread affinities shall be preferred.

Numerical methods for solving the equivalent inclusion equation in semi-analytical models

Equivalent inclusion method is the basis for semi-analytical models in tackling inhomogeneity problems. Equivalent eigenstrains are obtained by solving the consistency equation system of the equivalent inclusion method and then stress disturbances caused by inhomogeneities are determined. The equivalent inclusion method equation system can only be solved numerically, but the current fixed-point iteration method may not be able to achieve deep convergence when the Young's modulus of inhomogeneity is lower than that of the matrix material. The most significant innovation of this paper is to reveal the non-convergence mechanism of the current method. Considering the limitation, the Jacobian-free Newton Krylov algorithm is selected to solve the equivalent inclusion method equation. Results indicate that the new algorithm has significant advantages of computing accuracy and efficiency compared with the classic method.

Graphene memristive synapses for high precision neuromorphic computing

Memristive crossbar architectures are evolving as powerful in-memory computing engines for artificial neural networks. However, the limited number of non-volatile conductance states offered by state-of-the-art memristors is a concern for their hardware implementation since trained weights must be rounded to the nearest conductance states, introducing error which can significantly limit inference accuracy. Moreover, the incapability of precise weight updates can lead to convergence problems and slowdown of on-chip training. In this article, we circumvent these challenges by introducing graphene-based multi-level (>16) and non-volatile memristive synapses with arbitrarily programmable conductance states. We also show desirable retention and programming endurance. Finally, we demonstrate that graphene memristors enable weight assignment based on k-means clustering, which offers greater computing accuracy when compared with uniform weight quantization for vector matrix multiplication, an essential component for any artificial neural network.

Hybrid Invasive Weed Optimization Algorithm for Parameter Inversion Problems

A hybrid invasive weed optimization (HIWO) algorithm based on the Broyden–Fletcher–Goldfarb–Shanno (BFGS) algorithm is proposed for the problems on parameter inversion of the nonlinear models of sun shadow with integer variables in the study. Our presented algorithm can take full advantage of the local search ability of BFGS algorithm and the global search ability of invasive weed optimization (IWO) algorithm. The HIWO algorithm can not only reverse the date of sun shadow model successfully, but also conquer the weaknesses that the classic mathematical methods are hard to address integer nonlinear optimization problems by utilizing integers in some random variables from algorithms. The results of numerical experiments demonstrate that the HIWO algorithm has not only high computing accuracy, but also fast convergence speed. It can effectively improve the accuracy and efficiency of the techniques of sun shadow location, and afford an effective and efficient technique to handle the issues of integer parameter inversion in engineering applications.

Combined Hybrid Finite Element Method Applied in Elastic Thermal Stress Problem

In view of combinative stability, combinative variational principle based on domain decomposition for elastic thermal stress problem is constructed with the merits of avoiding Lax–Babuska–Brezzi (LBB) conditions. Compared with the principle of elasticity problem, new load items from thermal are involved. In addition, combined hybrid finite element is proposed to discretize the new principle and to formulate element stiffness matrix. Energy compatibility is introduced not only to simplify the variational principle and the corresponding element stiffness matrix but also to reduce the error of finite element solutions. On cuboid element, the energy compatible stress mode is given explicitly. The numerical results indicate that combined hybrid element with eight nodes can give almost the same computing accuracy of displacement and better computing accuracy of stress compared with cuboid element with 20 nodes, is not sensitive to mesh distortion and can circumvent Poisson-locking phenomenon.

A piecewise memory principle for fractional derivatives

In the numerical approximation of fractional order derivatives, the crucial point is to balance the computing complexity and the computing accuracy. We proposed a piecewise memory principle for fractional derivatives, in which the past history is divided into several segments instead of discarded. The piecewise approximation is performed on each segment. Error estimation of piecewise memory principle is analyzed also. Numerical examples show that the contradiction of computing accuracy and complexity is effectively relaxed and the piecewise memory principle is superior to the existing short, variable and equal-weight memory principles. The impacts of the memory length, step size and segment size are also discussed.