Due to the synapse-like nonlinearity and memory characteristics, memristor is often used to construct memristive neural networks with complex dynamical behaviors. However, memristive neural networks with multistructure chaotic attractors have not been found yet. In this article, a novel method for designing multistructure chaotic attractors in memristive neural networks is proposed. By utilizing a multipiecewise memristive synapse control in a Hopfield neural network (HNN), various complex multistructure chaotic attractors can be produced. Theoretical analysis and numerical simulation demonstrate that multiple multistructure chaotic attractors with different topologies can be generated by conducting the memristive synapse-control in different synaptic coupling positions. Differing from traditional multiscroll attractors, the generated multistructure attractors contain multiple irregular shapes instead of simple scrolls. Meanwhile, the number of structures can be easily controlled with the memristor control parameters. Furthermore, we design a module-based analog memristive neural network circuit and the arbitrary number of multistructure attractors can be obtained by selecting corresponding control voltages. Finally, based on the memristive HNNs, a novel image encryption cryptosystem with a permutation-diffusion structure is designed and evaluated, exhibiting its excellent encryption performances, especially the extremely high key sensitivity.

In electron beam lithography (EBL), the proximity effect seriously influences pattern resolution under high-precision conditions. Mainstream proximity effect correction (PEC) methods based on 2-D fast Fourier transform (2D-FFT) calculate a large number of unexposed points; thus, it may suffer from low efficiency especially when the exposure layout is unevenly distributed. This article proposes an efficient unequally spaced grid PEC method for EBL based on the fast multipole method (FMM). FMM in PEC just calculates the interaction between all the exposure points, and thus, it gets rid of the limitation of the equally spaced grid. Compared to the state-of-the-art PEC method based on 2D-FFT, the calculation speed of FMM will exceed the current fastest 2D-FFT convolution when the layout exposure density is below a certain proportion (approximately 80% under the 10-thread CPU parallel computing conditions). For the application of integrated circuit (IC) mask industry, the error of FMM is within the acceptable range of PEC. The PEC method in this article has been applied to a free software via software as a service (SaaS) mode, and a Windows-based EBL simulation and optimization software toolkit “HNU-EBL,” which is freely available athttp://www.ebeam.com.cn.

This paper proposes a special-purpose system to achieve high-accuracy and high-efficiency machine learning (ML) molecular dynamics (MD) calculations. The system consists of field programmable gate array (FPGA) and application specific integrated circuit (ASIC) working in heterogeneous parallelization. To be specific, a multiplication-less neural network (NN) is deployed on the non-von Neumann (NvN)-based ASIC (SilTerra 180 nm process) to evaluate atomic forces, which is the most computationally expensive part of MD. All other calculations of MD are done using FPGA (Xilinx XC7Z100). It is shown that, to achieve similar-level accuracy, the proposed NvN-based system based on low-end fabrication technologies (180 nm) is$1.6\times $faster and$10^{2}$-$10^{3}\times $more energy efficiency than state-of-the-art vN-based MLMD using graphics processing units (GPUs) based on much more advanced technologies (12 nm), indicating superiority of the proposed NvN-based heterogeneous parallel architecture.