FitzHugh-Nagumo神经元网络的联想记忆与分割

以广泛讨论的Fitz Hugh-Nagumo神经元节点组成脉动神经元网络,从神经系统空时模式编码理论研究网络的记忆(或模式)存储与时间分割问题.给定一个输入模式,它是几种模式的叠加,网络能够以一部分神经元同步发放的形式一个接一个地分割出每一种模式.如果输入的模式有缺损,系统能够把它们恢复成原型,即神经网络的联想记忆功能.模拟需要调节耦合强度和噪声强度等参数使得网络在特定的参数值和中等强度噪声达到最优的时间分割,与广泛讨论的随机共振现象一致.     

Self-organized memories in coupled map lattices

Spatially extended dynamical systems subjected to a periodic external input are able to store short-term memories as a result of self-organization. This phenomenon has been used to interpret sliding charge density wave experiments, and has been found in lattices of coupled maps and oscillators. In this paper, we aim to describe the influence of a stochastic perturbation in the self-organized memories in a lattice of coupled linear and weakly nonlinear maps. We consider the case in which the coupling between maps depends on the lattice distance as a power law.     

神经系统中随机和混沌感知信号的联想记忆与分割

为了模拟人与动物感知信息的真实环境，以脉动神经元节点组成神经元网络，研究在随机刺激和混沌刺激等极端条件下的记忆模式存储与时间分割问题.研究表明：网络对于若干种模式的叠加输入，能够以一部分神经元同步发放的形式在时间域上分割出每一模式. 如果输入模式是缺损的，系统能够把它们恢复到原型，即具有联想记忆功能.通过调节耦合强度和噪声强度等参数使得网络在中等强度噪声达到最优的时间分割，与广泛讨论的随机共振现象一致. 关键词： 神经网络 空时模式 联想记忆 随机共振     

Probabilistic quantum memories

Typical address-oriented computer memories cannot recognize incomplete or noisy information. Associative (content-addressable) memories solve this problem but suffer from severe capacity shortages. I propose a model of a quantum memory that solves both problems. The storage capacity is exponential in the number of qbits and thus optimal. The retrieval mechanism for incomplete or noisy inputs is probabilistic, with postselection of the measurement result. The output is determined by a probability distribution on the memory which is peaked around the stored patterns closest in Hamming distance to the input.     

栽培和野生中药材灯盏花的近红外光谱鉴别模型

分别在2台傅里叶变换近红外光谱仪上采集了43个栽培和野生中药材灯盏花样品的近红外漫反射光谱,提取光谱信息的15个主成分,方差贡献率达到99%以上。以20个灯盏花样品作为建模集,15个主成分作为网络学习输入层的15个节点,在2台仪器上用2套光谱分别建立了识别栽培和野生灯盏花样品的BP-神经网络模型,并对预测集的23个样品用于实际鉴别分析。两台仪器上的建模集样品模型回代正确识别率均为100%,预测集样品的正确识别率分别为100%和95.7%,结果表明,利用近红外光谱法进行栽培和野生中药材灯盏花的快速鉴别是可行的。     

无云情况下L波段微波辐射计快速大气校正方法

利用数值天气预报模式输出的大气温湿廓线数据与大气辐射传输模型计算了无云情况下全球海面大气在L波段的上行、下行辐射亮温及透射率, 建立三个参数与大气水汽含量及海表气压的回归关系模型——辐射-水汽模型, 利用该模型可快速计算大气辐射参数, 对L波段微波辐射计进行大气校正. 为了验证模型的实用性和可靠性, 利用SSM/I卫星水汽含量数据和数值模式地表气压数据通过模型计算大气辐射参数, 并与Aquarius卫星实测L波段微波辐射数据进行对比分析. 结果表明: 模型计算的辐射亮温比卫星观测数据偏低约0.335 K, 但改正系统性偏差后的均方根误差仅0.086 K, 且模型计算的大气透射率与卫星观测数据基本一致, 说明利用该模型对L波段微波辐射计进行大气校正具有较高的可靠性, 相比于传统利用大气辐射传输模型进行大气校正, 该模型更为简单快速, 输入参数更易获取, 更适于工程应用.     

Pair-correlated patterns in Hopfield model of neural networks

We study the retrieval properties of the Hopfield model of neural networks when the memorized patterns are statistically correlated in pairs. There is a finite correlationk between the memories of each pair, but memories of different pairs are uncorrelated. The analysis is restricted to the case of an arbitrary but finite number of memories in the thermodynamic limit. We find that there are two retrieval regimes: for 0<T<(1–k) the system recognizes the stored patterns and for (1–k)<T<(1+k) the system is able to recognize pairs, but it is not able to distinguish between its two patterns.     

Charged magneto-exciton states in semiconductor quantum dots

By embedding a layer of self-assembled quantum dots into a field-effect structure, we are able to control the exciton charge in a single dot. We present the results of photoluminescence experiments as a function of both charge and magnetic field. The results demonstrate a hierarchy of energy scales determined by quantization, the direct Coulomb interaction, the electron–electron exchange interaction, and the electron–hole exchange interaction. For excitons up to the triply charged exciton, the behavior can be understood from a model assuming discrete levels within the quantum dot. For the triply charged exciton, this is no longer the case. In a magnetic field, we discover a coherent interaction with the continuum states, the Landau levels associated with the wetting layer.     

Top-down and bottom-up modulation of language related areas – An fMRI Study

Background  

One major problem for cognitive neuroscience is to describe the interaction between stimulus and task driven neural modulation. We used fMRI to investigate this interaction in the human brain. Ten male subjects performed a passive listening and a semantic categorization task in a factorial design. In both tasks, words were presented auditorily at three different rates.     

A technique for generating Feynman diagrams

We present a simple technique that allows to generate Feynman diagrams for vector models with interactions of order2n and similar models (Gross-Neveu, Thirring model) using a bootstrap equation that uses only the free field value of the energy as an input. The method allows to find the diagrams to, in principle, arbitrarily high order and applies to both energy and correlation functions. It automatically generates the correct symmetry factor (as a function of the number of components of the field) and the correct sign for any diagram in the case of fermion loops. We briefly discuss the possibility of treating QED as a Thirring model with non-local interaction.     

Enhancement of electromagnetically induced transparency in room temperature alkali metal vapor

Electromagnetically induced transparency (EIT) has led to several quantum optics effects such as lasing without inversion or squeezed light generation. More recently quantum memories based on EIT have been experimentally implemented in different systems such as alkali metal atoms. In this system the excited state of the optical transition splits into several sublevels due to the hyperfine interaction. However, most of the theoretical models used to describe the experimental results are based on a Λ-system with only one excited state. In this article, we present a theoretical model for the Λ-type interaction of two light, fields and an atomic system with multiple excited state. In particular we show that if the control and probe fields are orthogonally circularly polarized the EIT effect in an alkali-metal vapor can almost disappears. We also identify the reasons of this reduction and propose a method to recover the transparency via velocity selective optical pumping.     

The role of sensitivity and uncertainty analysis in combustion modelling

Models play an important role in improving our understanding of combustion processes and more and more are able to assist in the design of advanced energy conversion devices. Due to constant improvements in computing power and techniques such as automatic kinetic mechanism generation, we have the ability to represent combustion processes with increasing levels of detail. This is particularly true for kinetic processes where complex mechanisms are being developed which describe the oxidation of both conventional and alternative fuels. These mechanisms may comprise of up to hundreds of species and thousands of reactions with thermo-kinetic data derived from a wide variety of sources including direct measurements, global combustion experiments, and theoretical calculations. However, significant uncertainties in the data used to parametrise combustion models still exist. These input uncertainties propagate through models of combustion devices leading to uncertainties in the prediction of key combustion properties. In order to improve confidence in these models to the extent where they can successfully be used in design, input uncertainties need to be reduced as far as possible. This requires focussing efforts on those parameters which drive predictive uncertainty, which may be identified through sensitivity analysis. The paper will describe the methodologies available for the sensitivity and uncertainty analysis of combustion models with examples focussed on chemical kinetics. It will then discuss how such techniques can be incorporated into strategies for model improvement and will try to provide some future perspectives on how we can proceed in this direction as a research community.     

Detection of Apple Marssonina Blotch with PLSR,PCA, and LDA Using Outdoor Hyperspectral Imaging

In this study, hyperspectral images were used to detect a fungal disease in apple leaves called Marssonina blotch(AMB). Estimation models were built to classify healthy, asymptomatic and symptomatic classes using partial least squares regression(PLSR), principal component analysis(PCA), and linear discriminant analysis(LDA) multivariate methods. In general, the LDA estimation model performed the best among the three models in detecting AMB asymptomatic pixels, while all the models were able to detect the symptomatic class. LDA correctly classified asymptomatic pixels and LDA model predicted them with an accuracy of 88.0%. An accuracy of 91.4% was achieved as the total classification accuracy. The results from this work indicate the potential of using the LDA estimation model to identify asymptomatic pixels on leaves infected by AMB.     

A multineuron interaction model for neural networks

A multineuron interaction model (RS model) with an energy function given by the product of the squared distances in phase space between the state of the net and the stored patterns is studied in detail within a mean-field approach. Two limits are considered: when the patterns and antipatterns are stored (as in the Hopfield model), PAS case, and when only the patterns are taken into account, OPS case. TheT=0 solutions for the proper memories are exactly obtained for all finite values of, as a consequence of the energy function: whenever one of the overlaps is exactly one the corresponding equations decouple and no configuration average is required. Special interest is focused on the OPS situation, which presents a peculiar phase space topology. On the other hand, the PAS configuration recovers the Hopfield model in the appropriate limit, while keeping associative memory abilities far beyond the critical values of other models when the full Hamiltonian is considered.     

Models of random graph hierarchies

We introduce two models of inclusion hierarchies: random graph hierarchy (RGH) and limited random graph hierarchy (LRGH). In both models a set of nodes at a given hierarchy level is connected randomly, as in the Erd?s-Rényi random graph, with a fixed average degree equal to a system parameter c. Clusters of the resulting network are treated as nodes at the next hierarchy level and they are connected again at this level and so on, until the process cannot continue. In the RGH model we use all clusters, including those of size 1, when building the next hierarchy level, while in the LRGH model clusters of size 1 stop participating in further steps. We find that in both models the number of nodes at a given hierarchy level h decreases approximately exponentially with h. The height of the hierarchy H, i.e. the number of all hierarchy levels, increases logarithmically with the system size N, i.e. with the number of nodes at the first level. The height H decreases monotonically with the connectivity parameter c in the RGH model and it reaches a maximum for a certain c _max in the LRGH model. The distribution of separate cluster sizes in the LRGH model is a power law with an exponent about ? 1.25. The above results follow from approximate analytical calculations and have been confirmed by numerical simulations.     

What to do? Does science have a role?

A new generation of integrated assessment models of climate change policies is needed to capture the basic dynamical processes that govern the required transformation of the present fossil-based global economic system to a sustainable decarbonized system. After an overview of the abatement technologies and policy instruments that are already available and able today to achieve the transformation, three examples are presented of typical actor-based, system-dynamical models that are able to simulate some of the key dynamics of the transition processes. In addition to developing a new hierarchy of integrated assessment models, scientists need also to better educate the public and policy makers on the wide-reaching implications of the inherent inertia of the climate system.     

Measuring information transfer

An information theoretic measure is derived that quantifies the statistical coherence between systems evolving in time. The standard time delayed mutual information fails to distinguish information that is actually exchanged from shared information due to common history and input signals. In our new approach, these influences are excluded by appropriate conditioning of transition probabilities. The resulting transfer entropy is able to distinguish effectively driving and responding elements and to detect asymmetry in the interaction of subsystems.     

双耳幅值差确定声源方向的神经信息处理机理研究

神经信息系统实质上是定量系统, 应引起足够重视. 关于神经系统的定量研究方面的报道比较少见. 这一问题将会影响进一步的研究, 如双耳声音定向. 双耳定向是定量测量, 用定性分析的方法无法满足要求. 已有的生理实验发现声音输入信号强度与听觉神经的输出频率存在单调递增关系, 所以本文中声音强度的变化被简化成神经脉冲频率的变化. 本文基于圆映射和符号动力学原理, 建立了神经回路定量模型, 模型中对同侧输入回路采用兴奋性耦合, 对侧输入回路采用抑制性耦合, 并考虑神经元间突触连接的量子释放特征, 采用化学耦合模型实现连接, 用耦合系数表示神经元间的耦合程度. 采用Hodgkin-Huxley模型仿真研究听觉神经回路的输入/输出脉冲序列关系. 在已经仿真过的参数范围, 模型在输入信号变化与输出脉冲频率变化间存在单调递增/递减的关系. 对于单输入单输出的神经元, 采用符号动力学方法进行符号化; 对于多输入单输出的神经元, 采用分析各输出脉冲的产生时间, 判断其变化位置, 从神经脉冲序列中得到对应的两耳声音幅值差变化, 以此定位声源. 随着输出脉冲数的增加, 符号序列的长度增加, 符号序列对输入信号变化敏感, 能够得到较高的测量精度. 仿真结果表明这个模型是定量的, 神经脉冲序列能够区分信号的大小.     

Attention Mechanisms and Their Applications to Complex Systems

Deep learning models and graphics processing units have completely transformed the field of machine learning. Recurrent neural networks and long short-term memories have been successfully used to model and predict complex systems. However, these classic models do not perform sequential reasoning, a process that guides a task based on perception and memory. In recent years, attention mechanisms have emerged as a promising solution to these problems. In this review, we describe the key aspects of attention mechanisms and some relevant attention techniques and point out why they are a remarkable advance in machine learning. Then, we illustrate some important applications of these techniques in the modeling of complex systems.