Sleep behaviour: sleep in continuously active dolphins

Sleep has been assumed to be necessary for development and to be a vital function in mammals and other animals. However, Lyamin et al. claim that in bottlenose dolphins (Tursiops truncatus) and killer whales (Orcinus orca), neonates and their mothers show almost no sleep behaviour for the first month after birth; this conclusion is based on their observation that the cetaceans keep swimming, avoid obstacles and rarely close their eyes for 24 hours a day throughout that period. Here we analyse the behaviour and eye closure of three neonate-mother pairs of bottlenose dolphins and find that, although the animals tend to open both eyes when surfacing to breathe, one or both eyes are closed during 'swim rest', an underwater sleeping behaviour that is associated with continuous activity. This observation calls into question the conclusions of Lyamin et al., who overlooked this type of sleep by analysing the animals' eye state only when they surfaced to breathe.     

Paleodietary changes by penguins and seals in association with Antarctic climate and sea ice extent

Positioned near the top of the food web, the dietary composition of Antarctic penguins and seals can be an excellent indicator of the regional food web and thus the status of the marine ecosystem. The dietary composition of modern penguins and seals has been well investigated; a long-term time series of data on penguin and seal diets, however, are rare. Such data, especially any predating the initiation of human harvesting of fish, whales and seals in Antarctica, are crucial for understanding and predicting responses of regional marine food webs to natural climate changes. Here we review recent progress on research of paleodietary change in Antarctic penguins and seals, specifically the Adelie penguin （Pygoscelis adeliae） and Antarctic fur seal （Arctocephalus gazella）. These studies indicate that the dietary changes of penguins correspond quite well with fluctuations in climate and sea ice extent during the Holocene. The depleted δ15N ratios found in modern Adelie penguins support the ＂krill surplus hypothesis＂ in relation to historic human depletion of krilleating fish, seals and whales.     

Dopamine neurons modulate pheromone responses in Drosophila courtship learning

Learning through trial-and-error interactions allows animals to adapt innate behavioural ‘rules of thumb’ to the local environment, improving their prospects for survival and reproduction. Naive Drosophila melanogaster males, for example, court both virgin and mated females, but learn through experience to selectively suppress futile courtship towards females that have already mated. Here we show that courtship learning reflects an enhanced response to the male pheromone cis-vaccenyl acetate (cVA), which is deposited on females during mating and thus distinguishes mated females from virgins. Dissociation experiments suggest a simple learning rule in which unsuccessful courtship enhances sensitivity to cVA. The learning experience can be mimicked by artificial activation of dopaminergic neurons, and we identify a specific class of dopaminergic neuron that is critical for courtship learning. These neurons provide input to the mushroom body (MB) γ lobe, and the DopR1 dopamine receptor is required in MBγ neurons for both natural and artificial courtship learning. Our work thus reveals critical behavioural, cellular and molecular components of the learning rule by which Drosophila adjusts its innate mating strategy according to experience.     

Environment: whale-call response to masking boat noise

Background noise can interfere with the detection and discrimination of crucial signals among members of a species. Here we investigate the vocal behaviour in the presence and absence of whale-watcher boat traffic of three social groups (pods) of killer whales (Orcinus orca) living in the nearshore waters of Washington state. We find longer call durations in the presence of boats for all three pods, but only in recent recordings made following a period of increasing boat traffic. This result indicates that these whales adjust their behaviour to compensate for anthropogenic noise once it reaches a threshold level.     

基于去噪的熵算子边界检测法

图像的熵用来刻划图像的平滑性或均匀性 在图像的局部窗口中计算时 ,如果窗口中存在边界 ,则窗口中的图像不均匀 ,其灰度变化急剧 ,计算出的熵小 ;反之熵大 设定熵的阈值 ,即可判断是否存在边界 由于熵算子对噪声很敏感 ,直接用它进行边界检测 ,效果很差 文中针对这一缺陷 ,提出将熵算子与去噪相结合的边界检测法 ,如果计算出的熵大于阈值 ,要判断是噪声的出现所引起 ,还是边界的出现所引起 ,这样 ,边检测边界边去噪声 用该方法对信噪比较低的图像进行实验 ,得到了满意的效果     

基于改进鲸鱼算法优化支持向量机的故障诊断的研究与应用

故障诊断在工业生产过程中具有很重要的作用,尤其是对于要求比较高的分子蒸馏来说,微小的故障都会造成其提纯率,因此提出一种基于改进鲸鱼算法优化支持向量机的故障分类方法(improved whale optimization algorithm-support vector machine, IWOA-SVM),加入反向学习策略和对数权重因子到普通鲸鱼算法中。首先,用反向学习策略(opposition-based learning, OBL)代替随机初始种群,用反向学习策略选取出反向种群,对种群进行择优选择,一方面OBL能够高效地提高群智能算法的全局搜索能力,另一方面提高鲸鱼算法在重复迭代中的多样性,使其跳出局部最优解;其次,引入自适应权重因子并将其加入到鲸鱼优化算法中,利用权重因子的动态变化,很大程度上增强了全局搜索能力;最后,采用改进之后的鲸鱼算法对SVM的参数进行寻优,并利用优化之后的支持向量机对刮膜蒸发过程获得的故障数据进行诊断识别,将IWOA-SVM的结果与其他3种做对比。结果表明,IWOA-SVM算法分类准确率提升了2%,且其准确率保持在98%以上,在分类结果的准确性以及算法的鲁...     

Pathology: whales, sonar and decompression sickness

We do not yet know why whales occasionally strand after sonar has been deployed nearby, but such information is important for both naval undersea activities and the protection of marine mammals. Jepson et al. suggest that a peculiar gas-forming disease afflicting some stranded cetaceans could be a type of decompression sickness (DCS) resulting from exposure to mid-range sonar. However, neither decompression theory nor observation support the existence of a naturally occurring DCS in whales that is characterized by encapsulated, gas-filled cavities in the liver. Although gas-bubble formation may be aggravated by acoustic energy, more rigorous investigation is needed before sonar can be firmly linked to bubble formation in whales.     

深度学习在图像识别中的应用

深度学习通过建立深层神经网络来模拟人脑进行分析、学习和解释数据,被广泛用于图像识别领域.首先,简述了深度学习在图像识别中的研究现状;其次,介绍了卷积神经网络、深度置信网络、循环神经网络和生成对抗网络等几种常用于图像识别领域的深度学习网络模型;然后,从人脸识别、动作识别、跌倒检测等方面,论述了深度学习在图像识别领域的典型应用;最后,探讨了该领域的研究难点及发展前景.深度学习可以从不同的图像中自动提取相似的特征并进行分类,识别率高,鲁棒性强,推动了人工智能背景下图像识别的发展.无监督学习、对抗网络等将成为深度学习领域的热点.     

成年与幼仔斑海豹空气中声信号的特性比较

海豹科动物发声与其生命活动息息相关,记录其声信号并分析信号特征是海豹声学研究的一个重点.对斑海豹(Phoca largha)在空气中的发声及信号特性进行研究,对成年斑海豹与幼仔斑海豹的声信号特征进行比较,并利用Mat-lab R2011a等软件对这些信号进行时域和频域分析.结果表明:成年斑海豹在空气中声信号的持续时间分布在0.6～1.6s,平均值为1.15s,频谱能量集中分布在0.5～2kHz,峰值频率、3dB带宽和10dB带宽分别为(0.99±0.18),(0.22±0.17)和(0.69±0.16)kHz;幼仔斑海豹声信号平均持续时间为2.45s,比成年斑海豹长,频谱能量分布与成年斑海豹类似,峰值频率、3dB带宽和10dB带宽分别为(1.13±0.10),(0.11±0.10)和(0.61±0.44)kHz,其中峰值频率和3dB带宽变化范围均比成年斑海豹小.     

基于小波变换和支持向量机的人脸检测系统

研究了基于灰度图像的人脸检测问题，采用小波变换方法提取人脸特征，大大地降低了特征矢量的维数。使用交叉检验方法有效地解决了支持向量机训练时的对数估计问题，所设计的系统可以分别进行离线训练和在线检测，并且具有学习的功能。实验结果表明，该系统具有较高的正检率和较低的虚警率，而且还可以通过再学习进一步提高系统的检测性能。     

基于滑动权值经验模式分解的医学图像增强研究

医学影像设备由于外在和内在因素的影响,所成图像对比度较低。运用小波、超小波增强算法,由于基函数需预先设定,无法自适应分析局部特性,导致增强后图像引入较大的噪声,图像细节出现畸变。二维经验模式分解(BEMD)可以数据自驱动分析局部数据;但由于插值算法或极值的影响,分解分量存在斑块或斑点的缺陷,影响后续处理。针对这种情况,提出了运用随局部数据自适应调整的滑动权值函数的滑动权值经验模式分解(SWEMD),解决了BEMD的缺陷。将SWEMD引入医学图像增强,并结合提出的非线性增强规则,能够更好地增强原始图像,增强图像的细节清晰无变形,对比度适中。实验证明该算法的增强效果优于目前的图像增强算法。     

一类食饵-捕食扩散模型分析

在两种群相互作用的Lotka-Voherra模型的基础上考虑了一类食饵种群分布在2个斑块：一个斑块上食饵和捕食者相互作用且对捕食者种群进行捕获;而另—个斑块属于食饵保护区．没有捕食者进入且不允许对食饵种群进行捕获．并且食饵种群可以在2个斑块间进行扩散的食饵—捕食模型。讨论了平衡点的存在性,利用Hurwitz判别法证明了平衡点的局部渐近稳定性和通过构造李雅普诺夫函数,得到了平衡点全局渐近稳定的结论。     

基于全局和局部特征的图像拼接方法

针对传统序列图像拼接算法中的误差累积问题,提出一种基于全局和局部特征的图像拼接方法. 同时拍摄大视场角、低分辨率全局图像和小视场角、高分辨率局部图像,利用深度学习替代传统算法提取两者匹配点,进而根据两者面积比等比例扩大全局图像的匹配点坐标,将局部图像无缩放地投影至全局图像所在平面,最后融合投影后局部图像的重叠区域,拼接形成一幅大视场角、高分辨率全景图像. 实验结果表明,该方法中深度学习快速且精准地实现了特征匹配,同时局部图像间相互独立,有效地解决了拼接顺序限制和拼接误差累积.      

Language evolution: semantic combinations in primate calls

Syntax sets human language apart from other natural communication systems, although its evolutionary origins are obscure. Here we show that free-ranging putty-nosed monkeys combine two vocalizations into different call sequences that are linked to specific external events, such as the presence of a predator and the imminent movement of the group. Our findings indicate that non-human primates can combine calls into higher-order sequences that have a particular meaning.     

高速公路团雾视频监测预警研究

随着高速公路建设发展,运行车辆的增加,高速公路团雾天气常造成重大交通事故。高速公路团雾监测预警工作尚未开展,为推进团雾分级预警相关标准实施、减少交通事故,利用公路沿线发光目标物的视频图像分析、前散式能见度仪资料校验,基于大气衰减定律,探究水平能见度与发光目标物光亮度参数的关系。结果表明：公路团雾天气水平能见度与发光目标物光亮度参数之间存在显著相关性;水平能见度与发光目标物光亮度部分参数存在线性关系,并可建立线性回归模型;结合“高速公路团雾预警等级”标准,可对高速公路团雾天气进行分级预警;使用高速公路团雾视频监测预警平台可向交通运营、管理部门、交通参与人发布团雾分级预警信息;发光目标物团雾视频图像监测预警方法具有科学性、创新性、可行性,可实现连续监测预警。发光目标物视频分析技术与(geographic information system, GIS)分析技术、导航技术相结合,将形成及时、准确的高速公路团雾分级预警体系。视频图像团雾监测预警模型将会成为高速公路团雾监测预警的主要手段之一。     

基于集成迁移学习的细粒度图像分类算法

针对现有的大部分细粒度图像分类算法都忽略了局部定位和局部特征学习是相互关联的问题,提出了一种基于集成迁移学习的细粒度图像分类算法。该算法的分类网络由区域检测分类和多尺度特征组合组成。区域检测分类网络通过类别激活映射(class activation mapping,CAM)方法获得局部区域,以相互强化学习的方式,从定位的局部区域中学习图像的细微特征,组合各局部区域特征作为最终的特征表示进行分类。该细粒度图像分类网络在训练过程中结合提出的集成迁移学习方法,基于迁移学习,通过随机加权平均方法集成局部训练模型,从而获得更好的最终分类模型。使用该算法在数据集CUB-200-2011和Stanford Cars上进行实验,结果表明,与原有大部分算法对比,该算法具有更优的细粒度分类结果。     

以行动学习法提升干部执行力

行动学习法是对以往经验的总结与反思,并在反思中获得经验,求得发展的方法。它既是构建学习型组织的可行方法,也是提升干部执行力的有效途径。干部执行力是一种综合能力,能够通过培养来提升。行动学习法通过群策群力的团队合作模式对实际问题进行深入探析,并经过讨论、反思、质疑后,制定解决问题的方案,然后付诸实施,在整个过程中可以不断提升干部执行力。正确运用行动学习法,不仅可以改善干部知识结构,也可以提升团队合作能力以及干部对政策理论的理解力。     

Conservation of polymorphic simple sequence loci in cetacean species

Length polymorphisms within simple-sequence loci occur ubiquitously in non-coding eukaryotic DNA and can be highly informative in the analysis of natural populations. Simple-sequence length polymorphisms (SSLP) in the long-finned pilot whale Globicephala melas (Delphinidae) have provided useful information on the mating system as well as on the genetic structure of populations. We have therefore tested whether the polymerase chain reaction primers designed for Globicephala could also be used to uncover variability in other whale species. Homologous loci could indeed be amplified from a diverse range of whales, including all toothed (Odontoceti) and baleen whales (Mysticeti) tested. Cloning and sequencing these loci from 11 different species revealed an unusually high conservation of sequences flanking the simple-sequence stretches, averaging 3.2% difference over 35-40 Myr. This represents the lowest divergence rate for neutral nucleotide positions found for any species group so far and raises the possible need for a re-evaluation of the age of the modern whales. On the other hand, the high conservation of non-coding sequences in whales simplifies the application of SSLP DNA fingerprinting in cetacean species, as primers designed for one species will often uncover variability in other species.     

基于局部结构和视觉显著特征的红外和可见光图像泊松重构融合算法

研究了一种基于泊松重构的红外和可见光图像融合算法,算法在梯度域内实现图像信息的融合,可有效避免传统空域和变换域方法在融合图像中由于局部亮度不一致而产生伪边缘。另外,提出的算法在源图像梯度融合时,同时考虑了图像的局部结构显著和视觉显著特征,能够在保留源图像更多细节的同时突出输入图像的视觉显著目标信息。通过与其他最新融合算法的对比实验结果显示本文算法获得的融合图像既有突出的红外目标,又有清晰的可见光背景细节,并且不会产生伪影和噪声,同时客观评价指标也有显著的优势。     

Consideration of the Local Correlation of Learning Behaviors to Predict Dropouts from MOOCs

Recently, Massive Open Online Courses(MOOCs) have become a major online learning methodology for millions of people worldwide. However, the dropout rates from several current MOOCs are high. Usually, dropout prediction aims to predict whether a learner will exhibit learning behaviors during several consecutive days in the future. Therefore, the information related to the learning behaviors of a learner in several consecutive days should be considered. After in-depth analysis of the learning behavior patterns of the MOOC learners, this study reports that learners often exhibit similar learning behaviors on several consecutive days, i.e., the learning status of a learner for the subsequent day is likely to be similar to that for the previous day. Based on this characteristic of MOOC learning,this study proposes a new simple feature matrix for keeping information related to the local correlation of learning behaviors and a new Convolutional Neural Network(CNN) model for predicting the dropout. Extensive experimental validations illustrate that the local correlation of learning behaviors should not be neglected. The proposed CNN model considers this characteristic and improves the dropout prediction accuracy. Furthermore, the proposed model can be used to predict dropout temporally and early when sufficient data are collected.