两维局域互联神经网络的关联存储

本文将局域互联神经网络的新概念推广到两维情形,并对两维局域互联关联存储进行了理论分析和大量的计算机模拟.结果表明,两维局域互联神经网络的优点是,在满足存储容量限制的前提下,它与全局互联神经网络具有相同的关联存储能力,而其互联权重矩阵要比全局互联网络小得多.因而,有利于使用现有的空间光调制器实现两维大规模的人工神经网络.     

双极双向联想存储器的单通道光学实现

以附加神经元引入附加背景的方式获得将线性离散比极神经元的神经网络在单通道光学矢量-矩阵乘法器内实现的方法,给出了相应的光学系统的修正和非负光学模板的编码形式.以双极神经元的双向联想存储器为例进行了计算机和光电实验模拟.     

Convergence analysis of stochastic hybrid bidirectional associative memory neural networks with delays

The stability property of stochastic hybrid bidirectional associate memory (BAM) neural networks with discrete delays is considered. Without assuming the symmetry of synaptic connection weights and the monotonicity and differentiability of activation functions, the delay-independent sufficient conditions to guarantee the exponential stability of the equilibrium solution for such networks are given by using the nonnegative semimartingale convergence theorem.     

时滞双向联想记忆神经网络的全局稳定性

通过构造一个合适的Lyapunov泛函及应用不等式的分析技巧研究了具有时滞的双向联想记忆 神经网络的平衡点的全局稳定性问题-在对神经元激励函数较宽松的假设条件下（可以不满 足Lipschitz条件），获得了一个新的保证全局渐近稳定性的判定准则-结果可应用于包含非 Lipschitz的一类更加广泛的神经元激励函数的神经网络的设计中- 关键词： Lyapunov泛函 时滞 双向联想记忆神经网络 全局渐近稳定性     

A biologically constrained learning mechanism in networks of formal neurons

A new learning mechanism is proposed for networks of formal neurons analogous to Ising spin systems; it brings such models substantially closer to biological data in three respects: first, the learning procedure is applied initially to a network with random connections (which may be similar to a spin-glass system), instead of starting from a system void of any knowledge (as in the Hopfield model); second, the resultant couplings are not symmetrical; third, patterns can be stored without changing the sign of the coupling coefficients. It is shown that the storage capacity of such networks is similar to that of the Hopfield network, and that it is not significantly affected by the restriction of keeping the couplings' signs constant throughout the learning phase. Although this approach does not claim to model the central nervous system, it provides new insight on a frontier area between statistical physics, artificial intelligence, and neurobiology.     

A phenomenological memristor model for short-term/long-term memory

Memristor is considered to be a natural electrical synapse because of its distinct memory property and nanoscale. In recent years, more and more similar behaviors are observed between memristors and biological synapse, e.g., short-term memory (STM) and long-term memory (LTM). The traditional mathematical models are unable to capture the new emerging behaviors. In this article, an updated phenomenological model based on the model of the Hewlett–Packard (HP) Labs has been proposed to capture such new behaviors. The new dynamical memristor model with an improved ion diffusion term can emulate the synapse behavior with forgetting effect, and exhibit the transformation between the STM and the LTM. Further, this model can be used in building new type of neural networks with forgetting ability like biological systems, and it is verified by our experiment with Hopfield neural network.     

A macro-mechanical constitutive model of shape memory alloys

It is of practical interest to establish a precise constitutive model which includes the equations describing the phase transformation behaviors and thermo-mechanical processes of shape memory alloy (SMA). The microscopic mechanism of super elasticity and shape memory effect of SMA is explained based on the concept of shape memory factor defined by the author of this paper. The conventional super elasticity and shape memory effect of SMA are further unified as shape memory effect. Shape memory factor is redefined in order to make clear its physical meaning. A new shape memory evolution equation is developed to predict the phase transformation behaviors of SMA based on the differential relationship between martensitic volume fraction and phase transformation free energy and the results of DSC test. It overcomes the limitations that the previous shape memory evolution equations or phase transformation equations fail to express the influences of the phase transformation peak temperatures on the phase transformation behaviors and the transformation from twinned martensite to detwinned martensite occurring in SMA. A new macro-mechanical constitutive equation is established to predict the thermo-mechanical processes realizing the shape memory effect of SMA from the expression of Gibbs free energy. It is expanded from one-dimension to three-dimension with assuming SMA as isotropic material. All material constants in the new constitutive equation can be determined from macroscopic experiments, which makes it more easily used in practical applications. Supported by the National Natural Science Foundation of China (Grant No. 95505010), the National High Technology Research and Development Program of China (Grant No. 2006AA03Z109), the China Postdoctoral Science Foundation (Grant No. 20080430933), the Open Foundation of Institute of Engineering Mechanics of National Seism Bureau of China (Grant No. 2007B02), and the Harbin Talent Foundation of Scientific and Technical Innovation (Grant No. RC2009QN-017046)     

光学并行高速缓冲存储列阵的实验研究

光学并行高速缓冲存储列阵和随机暂存列阵,是数字光学处理机中的重要部件。本文提出了一种简单的光学并行高速缓冲存储列阵系统,并在实验中完成了光信号列阵的随机存、取全过程。     

A high-fidelity memory scheme for quantum data buses

A novel quantum memory scheme is proposed for quantum data buses in scalable quantum computers by using adjustable interaction. Our investigation focuses on a hybrid quantum system including coupled flux qubits and a nitrogen–vacancy center ensemble. In our scheme, the transmission and storage(retrieval) of quantum state are performed in two separated steps, which can be controlled by adjusting the coupling strength between the computing unit and the quantum memory. The scheme can be used not only to reduce the time of quantum state transmission, but also to increase the robustness of the system with respect to detuning caused by magnetic noises. In comparison with the previous memory scheme, about 80% of the transmission time is saved. Moreover, it is exemplified that in our scheme the fidelity could achieve 0.99 even when there exists detuning, while the one in the previous scheme is 0.75.     

Computational complexity,learning rules and storage capacities: A Monte Carlo study for the binary perceptron

We examine the storage capacity for the binary perceptron using simulated annealing. In particular, we clarify the connection between the computational complexity of learning algorithms and the attained storage capacity. From finite-size studies we obtain a critical storage capacity, _c ()=0.8331±0.0016, in good agreement with the replica analysis of Krauth and Mézard. However, we demonstrate that a polynomial time cooling schedule yields a vanishing storage capacity in the thermodynamic limit as predicted by the dynamical theory of Horner. Nonetheless, we show these two results may be reconciled by explicitly verifying that the learning problem for the binary perceptron is NP-complete. This investigation has been made possible by the development of an accelerated annealing algorithm.     

A simultaneous generation of QKD and QDC via optical memory array for distributed network security

We propose a novel system of a simultaneous generation of continuous variable quantum key distribution (QKD) and quantum dense coding (QDC) via an optical memory array. The optical memory system is formed by using an array waveguide incorporating a nano-ring resonator, whereas the different spatial light modes can be generated and stored within an optical memory unit. The polarized photon is formed and stored within a storing device, i.e. a ring resonator, whereas the different time slot entangled photons can be generated, transmitted and detected by the different subscriber in the distributed networks. By using the optical memory concept, the continuous variable quantum key distribution is provided. Furthermore, the use of quantum dense coding via time division multiplexing paths, i.e. different time slot, is also plausible. The advantage of the proposed system is that the quantum key distribution can provide the network top security with high capacity and safety, which is the large demand of usage in the public networks.     

A computational study of entropy rules for charged fullerenes

Thermochemical data on fullerenes are relatively scarce. However, some thermochemical information can be derived from gas-phase experiments using the Knudsen cell mass spectrometry method. The third-law treatment can be carried out on the observed data, though one has to make the crucial presumption that the change in the thermodynamic potential ΔΦ _T ^o in the course of the reactions considered is negligible: ΔΦ _T ^o =0. It would be difficult to check the presumption directly in the experiment, but it can be checked computationally. Model reactions like C₆₀+ ₇₀ ^? = ₆₀ ^? +₇₀ are selected. The change in the thermodynamic potential ΔΦ _T ^o and the change in the standard entropy ΔS _T ^o are computed. For example, at a temperature of T=1000 K, the standard changes for the reaction evaluated using the SAM1 method are ΔΦ _T ^o =1.513 cal/(mol K) and ΔS _T ^o =?0.054 cal/(mol K). Overall, the computations support the critical thermodynamic presumption.     

Model architecture for associative memory in a neural network of spiking neurons

A synaptic connectivity model is assembled on a spiking neuron network aiming to build up a dynamic pattern recognition system. The connection architecture includes gap junctions and both inhibitory and excitatory chemical synapses based on Hebb’s hypothesis. The network evolution resulting from external stimulus is sampled in a properly defined frequency space. Neurons’ responses to different current injections are mapped onto a subspace using Principal Component Analysis. Departing from the base attractor, related to a quiescent state, different external stimuli drive the network to different fixed points through specific trajectories in this subspace.     

A study of the structure of G-P zones in Ti-rich TiNi shape memory melt-spun ribbons

A hybrid structure composed of ordered C11_b Ti₂Ni and Ti₅Ni₃ substructures for the G-P zones precipitated in Ti-rich Ti–Ni shape memory ribbons is proposed to explain the appearance of extra diffuse scattering streaks in transmission electron microscopy observations. The point defects of Ni vacancies and Ti antisites, resulting in the lattice fluctuant displacement, also cause the incommensurate extra diffuse scattering streaks.     

The influence of thermally assisted tunneling on the performance of charge trapping memory

We evaluate the influence of the thermally assisted tunneling (TAT) mechanism on charge trapping memory (CTM) cell performance by numerical simulation, and comprehensively analyse the effects of the temperature, trap depth, distribution of trapped charge, gate voltage and parameters of TAT on erasing/programming speed and retention performance. TAT is an indispensable mechanism in CTM that can increase the detrapping probability of trapped charge. Our results reveal that the TAT effect causes the sensitivity of cell performance to temperature and it could affect the operational speed, especially for the erasing operation. The results show that the retention performance degrades compared with when the TAT mechanism is ignored.     

温稠密钛电导率计算

温稠密物质是惯性约束核聚变、重离子聚变、Z箍缩动作过程中物质发展和存在的重要阶段. 其热力学性质和辐射输运参数在聚变实验和内爆驱动力学模拟过程中有至关重要的作用. 本文通过建立非理想Saha方程, 结合线性混合规则的理论方法模拟了温稠密钛从10^-5-10 g·cm^-3, 10⁴ K到3×10⁴ K区间的粒子组分分布和电导率随温度密度的变化, 其中粒子组分分布由非理想Saha方程求解得到. 线性混合规则模型计算温稠密钛的电导率时考虑了包括电子、原子和离子之间的多种相互作用. 钛的电导率的计算结果与已有的爆炸丝实验数据相符. 通过电导率随温度密度变化趋势判断, 钛在整个温度区间, 密度0.56 g·cm^-3时发生非金属相到金属相相变. 对于简并系数和耦合系数的计算分析, 钛等离子体在整个温度和密度区间逐渐从弱耦合、非简并状态过渡到强耦合部分简并态.     

用于形状记忆材料研究的中子衍射原位温度加载系统

配合中国工程物理研究院的中子衍射应力谱仪开展形状记忆材料的相关研究,设计了一套原位温度加载系统。该系统可为测量样品提供25~800 ℃温度环境,温度控制器采用大林改进算法,有效地消除了温度过冲问题。该系统已开展了多次带束测试,结果表明其结构设计适用于中子衍射实验,温度控制在上升阶段无超调、稳态误差为1 ℃,满足形状记忆材料的温度加载实验要求。     

Flow version of statistical neurodynamics for oscillator neural networks

We consider a neural network of Stuart–Landau oscillators as an associative memory. This oscillator network with N$N$ elements is a system of an N$N$-dimensional differential equation, works as an attractor neural network, and is expected to have no Lyapunov functions. Therefore, the technique of equilibrium statistical physics is not applicable to the study of this system in the thermodynamic limit. However, the simplicity of this system allows us to extend statistical neurodynamics [S. Amari, K. Maginu, Neural Netw. 1 (1988) 63–73], which was originally developed to analyse the discrete time evolution of the Hopfield model, into the version for continuous time evolution. We have developed and attempted to apply this method in the analysis of the phase transition of our model network.     

Quantitative structure-plasticity relationship in metallic glass:A machine learning study

The lack of the long-range order in the atomic structure challenges the identification of the structural defects, akin to dislocations in crystals, which are responsible for predicting plastic events and mechanical failure in metallic glasses(MGs). Although vast structural indicators have been proposed to identify the structural defects, quantitatively gauging the correlations between these proposed indicators based on the undeformed configuration and the plasticity of MGs upon external loads is still lacking. Here, we systematically analyze the ability of these indicators to predict plastic events in a representative MG model using machine learning method. Moreover, we evaluate the influences of coarse graining method and medium-range order on the predictive power. We demonstrate that indicators relevant to the low-frequency vibrational modes reveal the intrinsic structural characteristics of plastic rearrangements. Our work makes an important step towards quantitative assessments of given indicators, and thereby an effective identification of the structural defects in MGs.