优选伸缩比SCE-UA算法与新安江模型参数率定研究

洗牌复合形进化(SCE-UA)算法作为寻优能力较强的全局寻优算法已被广泛应用于水文模型参数率定领域。经典SCE-UA算法中的下山单纯形搜索采用固定伸缩比实现反射和收缩操作,其寻优效率还有待提升,实际应用中发现变动伸缩比能够改善算法的寻优效率。针对以上问题,结合数值实验与分析,发现伸缩比取值存在最优区间,在伸缩比最优区间内均匀采样,利用优选法在单纯形搜索过程中动态确定最适伸缩比,提出了基于优选伸缩比的SCE-UA算法。本研究利用改进的算法,基于多组人工降雨—径流资料开展了新安江模型参数率定研究。结果表明,相较于固定伸缩比,优选伸缩比算法在多组人工降雨—径流资料下的参数率定测试中均表现出相同迭代次数下目标函数值更优的结果,显著提升了优化效率,进一步提升了算法的深度搜索能力,具有推广应用价值。     

基于螺旋搜索机制的行星搜索算法

螺旋搜索机制的全局搜索能力强,广泛用于萤火虫算法及鲸鱼搜索算法,但其收敛速度慢,收敛精度低,局部搜索能力较差。通过改变收敛范围较小时的搜索模式,提出了局部螺旋搜索来提高其局部搜索能力,并引入变异操作提高其局部搜索能力,提出了行星搜索算法。通过对单峰及多峰值测试函数对该算法进行验证。结果表明行星搜索算法在收敛速度、搜索精度及局部搜索能力等方面较粒子群算法、萤火虫算法及鲸鱼搜索算法等有明显提升。     

Designing amino acid sequences to fold with good hydrophobic cores

We present two methods for designing amino acid sequences ofproteins that will fold to have good hydrophobic cores. Giventhe coordinates of the desired target protein or polymer structure,the methods generate sequences of hydrophobic (H) and polar(P) monomers that are intended to fold to these structures.One method designs hydrophobic inside, polar outside; the otherminimizes an energy function in a sequence evolution process.The sequences generated by these methods agree at the levelof 60–80% of the sequence positions in 20 proteins inthe Protein Data Bank. A major challenge in protein design isto create sequences that can fold uniquely, i.e. to a singleconformation rather than to many. While an earlier lattice-basedsequence evolution method was shown not to design unique folders,our method generates unique folders in lattice model tests.These methods may also be useful in designing other types offoldable polymer not based on amino acids     

实数编码量子共生演算法及其在云任务调度中的应用

针对共生演算法收敛慢和易陷入局部最优的问题,结合量子遗传算法理论,提出一种实数编码的量子共生演算法(real-coded quantum symbiotic organisms search,RQSOS)。首先依据三角模糊数提出差异度概念,并依此构造一个以自变量向量的分量和一对概率幅为等位基因的三倍染色体,使一条染色体携带更多信息并增强解的多样性;然后提出一种基于阿基米德螺旋线的探索学习模式,加强对解空间的探索精度;最后使用共生演算法更新差异度值并依据差异度值对种群进行学习和变异操作,促使整个种群快速向最优方向进化且减小了陷入局部最优的概率。利用数值优化问题和云任务调度问题对算法进行验证,仿真结果表明,RQSOS算法在收敛速度和寻优能力上均有明显提升,是一种可行有效的算法。     

基于引力搜索算法的分数阶变异时序回归GSA-TSGM(1,1)模型

为了利用分数阶累加算子在灰色短期预测中的高效性能,首次将分数阶累加算子引入变异时序回归模型以期取得更高的预测精度。主要方法如下：首先取湖北省链子崖某监测点1978—1987年的十年数据作为训练集并使用引力搜索算法确定最佳分数阶累加阶数,而1988—1993年的六年数据作为验证集验证提出的模型;其次对比了经典灰色模型GM(1,1)、分数阶累加灰色模型、变异时序回归模型TSGM(1,1)三种灰色模型。结果如下：首先修正了陈西江等人变异时序回归模型仿真时出现的错误,其次表明了相比于其他的模型,基于引力搜索算法的分数阶累加时序回归模型在进行灰色长期预测中具有较高的预测精度。因此,通过分数阶累加算子提高了灰色理论中长期预测模型的精度,为灰色长期预测提供了指导。     

Revisiting the two-stage choice model: an empirical study of consumer choice on brand website visits

In consumer choice behaviour literature, a two-stage choice model serves as a base theory where choice behaviour is decomposed into a consideration stage plus a choice stage. Yet, this default model has been increasingly challenged by the ‘limited consumer search’ school of thought because consumers do not necessarily review all products in a choice set before making decisions, but frequently draw on external information as mental short cuts. Methodologically, the choice behaviour in stage 1 has traditionally been latent because data that directly observe how consumers form their consideration sets by eliminating alternatives are lacking. This study continues the ‘limited consumer search’ line of reasoning and proposes an aided non-compensatory process in choice stage 1, where consumers reduce brand website alternatives by using online ratings to arrive at a consideration set. We use observed Web analytics data to unveil the stage 1 choice process and also the transition from stage 1 to stage 2, which is in need of further research based on extant literature. Lastly, we cross-validate our model with two types of websites (i.e. search vs. experience/credence) and find our model is contingent on the type of website content, where consumers’ inclination to use online ratings for decision-making varies.     

Enhanced user performance in an image gallery application with a mobile autostereoscopic touch display

In this study, we explored how stereoscopic depth affects performance and user experience in a mobile device with an autostereoscopic touch display. Participants conducted a visual search task with an image gallery application on three layouts with different depth ranges. The task completion times were recorded, and the participants were asked to rate their experiences. The results revealed that the image search times were facilitated by a mild depth effect and that too great a depth slowed search times and decreased user-experience ratings.     

Context matters: The structure of task goals affects accuracy in multiple-target visual search

Career visual searchers such as radiologists and airport security screeners strive to conduct accurate visual searches, but despite extensive training, errors still occur. A key difference between searches in radiology and airport security is the structure of the search task: Radiologists typically scan a certain number of medical images (fixed objective), and airport security screeners typically search X-rays for a specified time period (fixed duration). Might these structural differences affect accuracy? We compared performance on a search task administered either under constraints that approximated radiology or airport security. Some displays contained more than one target because the presence of multiple targets is an established source of errors for career searchers, and accuracy for additional targets tends to be especially sensitive to contextual conditions. Results indicate that participants searching within the fixed objective framework produced more multiple-target search errors; thus, adopting a fixed duration framework could improve accuracy for career searchers.     

基于MTPM和DPM的多无人机协同广域目标搜索滚动时域决策

传统的协同搜索决策方法在目标引导和机间协同方面存在不足. 研究建立了基于分布概率预测的目标概率图（Target probability map,TPM）初始化方法和基于贝叶斯准则的目标概率图动态更新方法,形成了修正目标概率图（Modified TPM,MTPM）及其运算机理.考虑对任务子区域进行可控回访,定义了数字信息素图（Digital pheromone map,DPM）,建立了数字信息素图使用方法及更新机理.设计了基于MTPM和DPM的寻优指标,建立了基于滚动时域控制的协同搜索决策方法（MTPM-DPM-RHC method,MDR）.仿真表明: 1） MTPM能有效降低对目标的虚警率和漏检率;2） DPM能有效实现对任务区域可控回访;3） MDR方法的遍历能力、重访能力和目标搜索效率均优于已有方法.     

用于提高谷歌图像搜索结果的二分类器在线学习方法

对于基于关键词的图像检索,利用检索结果的视觉相似性学习二分类器有望成为改善检索结果的最有效途径之一. 为改善搜索引擎的搜索结果,本文提出一种算法框架并且基于此框架着重研究训练数据选择这一关键问题. 训练数据选择过程由两个阶段组成:1）训练数据初始化以开始分类器学习过程;2）分类器迭代学习过程中的动态数据选择. 对于初始训练数据的选择,我们探讨了基于聚类和基于排序两种方法,并且对比了自动训练数据选择与人工标注的结果. 对于动态数据选择,我们比较了支持向量机和基于最大最小后验伪概率的贝叶斯分类器的分类效果. 组合上述两个阶段的不同方法,我们得到了8种不同的算法,并将其用于谷歌搜索引擎进行基于关键词的图像检索. 实验结果证明,如何从含有噪声的搜索结果中选择训练数据是搜索结果改善的关键问题. 实验显示我们的方法能够有效的改善谷歌搜索的结果,尤其是排序在前的结果. 尽早为用户提供更相关的结果能够更大程度的减少用户逐个翻页查看结果的工作. 另外,如何使自动训练数据选择与人工标注媲美仍是需要继续研究的一个问题.