关于稀疏分布存贮模型中A连接权阵的讨论

Ｋａｎｅｒｖａ的稀疏分布存贮模型由于对寻址地址采用了稀疏编码，对数据采用了分布式存贮，从而解决了大维数向量的输入问题，ＳＤＭ实际上是一个由输入层，中间层和输出层组成的三层前向网络，其中神经元间的互迦权值在输入层与中间层是预置的（用矩阵Ａ表示），中间层与输出层的连接权阵Ｃ由外积法得到，文中假定在相同的学习规则下，就信噪比意义而言，Ａ的均匀预置能使ＳＤＭ获得最优性能，从而为Ａ的预置提供了理论依据。     

基于查表的开环MPM预失真器

提出了一种基于查表的开环记忆多项式模型(MPM,Memory Polynomial Model)预失真器.从MPM预失真器系数估计和执行信号预失真处理两方面加以考虑.首先,利用预失真器和高功率放大器(HPA,High Power Amplifier)传输特性的互补性,由采集到的HPA输入、输出基带数据进行单次最小二乘估计得到MPM预失真器的系数,避免了以往递归结构的收敛性和计算量大等问题.然后,利用MPM预失真器的特点,建立了基于输入信号强度索引的一维向量表,通过查表实现MPM预失真器的幂次累加项,极大地减少了实现预失真器所需的计算量及硬件开销.仿真显示,采用提出的预失真器可以有效补偿HPA对信号的记忆非线性失真,有效带宽内获得了近20 dB的功率改善,也显著改善了系统误码性能.     

Revisiting Perspective-Taking: Can People Maintain Imagined Perspectives?

Abstract

Three experiments were conducted to examine whether people can adopt and maintain imagined perspectives in the absence of target information. The task used entailed providing information about an imagined perspective in advance of target information to examine whether this would facilitate perspective-taking performance and reduce or eliminate alignment effects that are commonly reported in the literature. The three experiments employed different types of spatial environments: an environment learned from navigating a computer screen (Experiment 1), and an immersive environment that was either remote (Experiment 2) or immediate (Experiment 3) at the time of retrieval. Across the three experiments, results showed that information about an imagined perspective can be utilized ahead of target information. Furthermore, they suggested that alignment effects can be reduced as a result of processing information about perspective ahead of target information, but only when reasoning about specific nonimmediate spatial relations (Experiments 1 and 2). Results are discussed in connection with previous findings on spatial updating and the organizational structure of spatial memory.     

An Investigation into Vertical Bias Effects

ABSTRACT

In three experiments, after exploring a virtual environment (VE), adult participants made spatial judgments about the location of target objects that were higher and lower than their perceived test location within the VE. In Experiment 1, the locations of the target objects were inferred from verbal instructions. The main results were a tendency to judge objects as closer to the horizontal plane than their true locations, and more efficient downward than upward judgments. Both effects generally accord with findings reported by Wilson et al. (2004a Wilson, P. N., Foreman, N., Stanton, D. and Duffy, H. 2004a. Memory for targets in a multilevel simulated environment: Evidence for vertical asymmetry in spatial memory. Memory & Cognition, 32: 283–297. [Crossref] , [Google Scholar], 2004b Wilson, P. N., Foreman, N., Stanton, D. and Duffy, H. 2004b. Memory for targets in a multi-level simulated-environment: A comparison between able-bodied and physically disabled children. British Journal of Psychology, 95: 325–338.  [Google Scholar]). In Experiments 2 and 3, which were closely modeled on the design of the Wilson et al. studies, regression to the horizontal plane was noted but no downward bias was observed. A misperception in the viewing height between the floors and ceilings of the virtual rooms was apparent in both experiments. The results from the present study together with earlier investigations suggest different hierarchical encoding of between-axis and within-axis information.     

一种高超声速滑翔飞行器轨迹智能预测方法

针对高超声速滑翔飞行器(Hypersonic glide vehicle, HGV)机动性强、轨迹预测困难的问题,选取气动加速度作为预测参数,提出了一种基于集合经验模态分解和注意力长短时记忆网络的HGV轨迹智能预测方法。首先,以HGV六自由度运动方程为基础,分析了其机动特性和气动力变化规律,建立了动力学跟踪模型,对气动加速度进行实时估计;其次,利用集合经验模态分解对估计的气动加速度进行分解和重构,减弱噪声影响,避免对预测模型的干扰;最后,利用去噪后的气动加速度数据对注意力长短时记忆网络进行训练,进而预测未来气动加速度数据并重构HGV未来轨迹,实现轨迹的在线预测。实验仿真表明,该方法能有效预测HGV机动轨迹,预测精度高、稳定性好。     

嵌入式多核操作系统确定性研究

为解决嵌入式多核操作系统确定性问题,详细阐述了计算确定性,重点分析了多核结构引入的不确定性问题。通过研究嵌入式多核操作系统的结构,从共享存储的确定性访问、确定性任务调度、确定性通信等方面提出了解决多核结构带来的不确定性问题的方法。     

环氧树脂基形状记忆复合材料的制备与性能

以自制的形状记忆环氧树脂为基体,CCF300级碳纤维U-3160织物为增强材料,采用真空袋成型工艺制备形状记忆复合材料,分析碳纤维含量对复合材料的形状记忆性能的影响。结果表明,当碳纤维含量为0%~50%时,随着增强体碳纤维含量的增加,复合材料的形状固定率由98%降至91%,形状回复率也由98%降至91%。在有限次的循环实验中,循环次数对形状固定率和形状回复率没有明显的影响。     

MSMA振动发电机的建模与仿真

振动发电就是利用电磁感应、压电技术、智能材料等将外部的机械振动能量通过一定装置转换成电能,实现机械振动能量和电能的转换.在分析磁控形状记忆合金（Magnetic Shape MemoryAlloy,简称MSMA）振动发电原理的基础上,利用MSMA智能材料的维拉利效应对振动能量进行收集,建立了MSMA振动发电机的数学模型,求出振动发电机感应电动势与压应力及外加磁场的数学关系.分析了振动应力幅值、频率的响应特性,仿真结果验证了MSMA振动发电的可行性.     

聚氨酯改性环氧树脂形状记忆材料

以甲基四氢邻苯二甲酸酐(METHPA)为固化剂,利用自制的聚氨酯预聚体(PU)改性环氧树脂(TDE-85)制备形状记忆聚合物。研究了形状记忆环氧树脂的力学性能、Tg和形状记忆性能。研究表明,形状记忆环氧树脂的韧性得到明显的提高,Tg得到了相应的下降,形状记忆性能良好,形状回复率皆在96%以上,形状固定率约达100%。     

超高温度梯度定向凝固NiMnFeGa磁致形状记忆合金

利用超高温度梯度定向凝固装置制备出定向Ni51.2Mn20.0Fe13.0Ga15.8磁致形状记忆合金.沿晶体生长方向凝固组织由等轴晶逐渐过渡为柱状晶,马氏体变体取向性明显,γ相愈加细小,分布愈加均匀.Ni51.2Mn20.0-Fe13.0Ga15.8合金在加热和冷却过程中发生热致马氏体相变及逆相变,马氏体相变温度Ms＝121.3℃,Mf＝99.3℃,As＝118℃,Af＝147℃.