基于残差网络的自然环境下蝴蝶种类识别

【目的】蝴蝶属鳞翅目(Lepidoptera)昆虫,其对生存环境敏感,能够作为区域生态环境的指示物种,自然环境下蝴蝶种类自动识别对生态系统稳定有重要意义。现有研究中蝴蝶种类和数量较少,且多以标本图像作为识别对象,鉴于此,本研究构建了自然环境下蝴蝶图像数据集,提出一种以残差网络为基础的蝴蝶种类识别模型LDResNet。【方法】首先,引入可变形卷积,增强网络对不同形状蝴蝶图像的特征提取能力,获得更细粒度的特征;其次,在可变形卷积后嵌入注意力机制,增大蝴蝶特征权重,降低冗余信息干扰;最后,利用改进的深度可分离卷积降低模型参数量。【结果】在自建数据集上实验,LDResNet模型取得了87.61%的平均识别准确率,较原始模型提升了3.14%,模型参数量仅为1.04 MB。【结论】LDResNet模型相较其他模型,在平均识别准确率和参数量方面均有明显优势,本研究模型可为自然环境下的蝴蝶种类自动识别提供技术支持。     

基于发育网络识别模型的大熊猫面部识别

个体识别是动物行为学与生态学研究工作的基础,也是制定珍稀野生动物保护政策的重要依据。为了丰富大熊猫个体识别和种群数量调查的方法,我们于2017年7月分别在四川省雅安市碧峰峡大熊猫基地和四川省汶川县耿达镇的中华大熊猫苑共计拍摄18只大熊猫个体,每只大熊猫拍摄6~13张高质量面部照片（共计131张）,利用发育网络（Developmental Network）建立大熊猫面部识别模型。利用此模型对存在部分背景的大熊猫面部照片进行识别检测,得到的个体识别率为79.41%,对完全去除背景的大熊猫面部照片进行识别检测,得到的个体识别率为58.82%。研究表明,发育网络具有足够的大熊猫个体识别能力,不同背景比例的照片对大熊猫个体识别的实际结果具有较大的影响。随着发育网络识别模型的发展,我们建议更多的野生动物保护研究者结合这一技术深入地开展珍稀野生动物（如大熊猫）个体识别研究,逐步提高识别准确度,并应用到关键区域大规模的动物调查中。     

基于双向长短时记忆网络和注意力机制的RNA m5C甲基化位点预测

RNA 5-甲基胞嘧啶(m5C)修饰在许多生物过程中发挥重要的作用,对m5C位点的准确识别有助于更好地理解其生物学功能,所以识别m5C甲基化位点十分必要。尽管已发展了多种识别m5C甲基化位点的机器学习方法,但预测能力仍有待提高。本文基于双向长短时记忆网络和注意力机制,提出了一种预测RNA m5C甲基化位点的深度学习算法。用该方法在人、小鼠、酿酒酵母和拟南芥共4种生物的RNA m5C数据集上进行实验,m5C位点预测AUC值分别达到92.5%、99.7%、93.6%和86.5%。与现有预测方法相比,该方法具有较好的预测性能,并且具有更优的泛化能力,为RNA m5C甲基化位点预测提供了一种新方法。     

基于深度残差神经网络迁移学习的牙形刺图像识别

为了探索基于深度神经网络模型的牙形刺图像智能识别效果,研究选取奥陶纪8种牙形刺作为研究对象,通过体视显微镜采集牙形刺图像1188幅,收集整理公开发表文献的牙形刺图像778幅,将图像数据集划分为训练集和测试集。通过对训练集图像进行旋转、翻转、滤波增强处理,解决了训练样本不足的问题。基于ResNet-18、ResNet-34、ResNet-50、ResNet-101、ResNet-152五种残差神经网络模型,采用迁移学习方法,对网络模型进行训练以获取模型参数,五种模型测试Top-1准确率分别为85.37%、85.85%、83.90%、81.95%、80.00%, Top-2准确率分别为94.63%、94.63%、94.15%、93.17%、93.66%,模型对牙形刺图像具有较好的识别效果。通过对比研究发现,ResNet-34识别准确率最高,说明对于特征简单的牙形刺属种,增加网络深度并不一定能提升准确率,而确定合适深度的模型则不仅可以提高识别准确率,还可以节约计算资源。通过ResNet-34模型的迁移学习训练和重新训练效果对比可以看出,迁移学习不仅可以获得较高的准确率,而且可以较快获取模型参...     

武汉东湖的磷-浮游植物动态模型

本文报道了东湖的一个富营养化模型。这个模型按照1年的时间标度描述东湖藻类的生长和磷循环,其状态变量包括浮游植物磷、藻类生物量、正磷酸盐、碎屑磷和沉积物磷。模型校准和检验结果表明,模型对于系统给定状态的描述是令人满意的,并且对于系统的强制函数的改变能给予合理的响应。根据东湖富营养化工程治理的初步设想,利用模型进行了东湖污水截流前后的水质预报,同时考查了截流后移出沉积物或引灌江水对于改善其水质的效果。模型所提供的各种预报可供拟定东湖治理方案时参考。     

基于注意力机制的RNA碱基关联图预测方法

目的 长链非编码RNA在遗传、代谢和基因表达调控等方面发挥着重要作用。然而，传统的实验方法解析RNA的三级结构耗时长、费用高且操作要求高。此外，通过计算方法来预测RNA的三级结构在近十年来无突破性进展。因此，需要提出新的预测算法来准确的预测RNA的三级结构。所以，本文发展可以用于提高RNA三级结构预测准确性的碱基关联图预测方法。方法 为了利用RNA理化特征信息，本文应用多层全卷积神经网络和循环神经网络的深度学习算法来预测RNA碱基间的接触概率，并通过注意力机制处理RNA序列中碱基间相互依赖的特征。结果 通过多层神经网络与注意力机制结合，本文方法能够有效得到RNA特征值中局部和全局的信息，提高了模型的鲁棒性和泛化能力。检验计算表明，所提出模型对序列长度L的4种标准（L/10、L/5、L/2、L）碱基关联图的预测准确率分别达到0.84、0.82、0.82和0.75。结论 基于注意力机制的深度学习预测算法能够提高RNA碱基关联图预测的准确率，从而帮助RNA三级结构的预测。     

哈尔滨二龙山水库浮游植物及水体污染状况的研究

报道了2000 年1 月、5 月、8 月、10 月四次对哈尔滨市境内的二龙山水库的浮游植物的调查结果。共发现浮游植物46 属55 种, 其中种类最多的是绿藻和硅藻。水库四季的平均浮游藻类数量为438.46 万个/升, 绿藻占46.98%, 硅藻占44.74%, 隐藻占6.76%。优势种是平裂藻、小环藻、直链藻、针杆藻、纤维藻、空星藻、十字藻、蹄形藻、栅藻、隐藻等。对8 月份水库及上游河流的水质进行了监测和分析。根据上述结果对水质进行了评价, 并对二龙山库区的水土保持及可持续性发展进行了探讨。     

罗非鱼对附着藻类和浮游植物的影响

鱼类通过牧食和营养盐排泄可以对水体生态系统产生影响,杂食性鱼类由于可摄食不同生境中的食物,可使生境之间的耦合作用发生变化。罗非鱼是我国南方很多水体的优势种,食物包括敞水生境的浮游植物和基质表层生境的附着藻类等。为了解罗非鱼对浮游植物和附着藻类的影响,实验在室外模拟条件下,分别设置罗非鱼组和无鱼对照组的两组处理,分析了罗非鱼对附着藻类及浮游植物生物量(叶绿素a)等的影响。结果表明:(1)罗非鱼显著地降低了附着藻类生物量,罗非鱼组中的附着藻类叶绿素a的平均值为0.15 mg·cm^-2,显著低于对照组中的1.26mg·cm^-2;(2)罗非鱼显著地增加了浮游植物的生物量,罗非鱼组中的浮游植物叶绿素a平均值为31.99μg·L^-1,显著高于对照组中的14.99μg·L^-1。研究结果显示,杂食性的罗非鱼可以促进系统的附着藻类向浮游植物转化。从控制浮游植物生物量的角度看,湖泊等水体的管理应该对罗非鱼密度加以有效控制。     

湖水含盐量和Cu2+浓度变化对Kinneret湖浮游植物的影响

通过人为改变湖水中的Cu^2 浓度和含盐量的方法,Cu^2 浓度和含盐量变化对Kinneret湖水中浮游植物可能造成的影响进行了分析。结果表明,湖水Cu^2 浓度增加会抑制Kinneret湖水中藻类等浮游植物的生长。这对改善湖水水质来7说是非常有利的,但高的Cu^2 浓度对农作物生长和人类健康是有害的,在另一方面,对含盐量较低的约旦河水来说,适当增加Cu^2 浓度则有利于藻类的生长。Kinneret湖中浮游植物的年平均生物量随湖水含盐量的下降而有增加的趋势,特别是当含盐量低于200μ左右时,浮游植物中蓝藻占的比例会升高,这对作为饮用水资源的Kinneret湖水来说是不利的,由此可见,适当控制和维持Kinneret湖水Cu^2 浓度和含盐量对湖水水质保护具有重要的意义。     

基于深度学习的葡萄砧木叶片识别研究

嫁接有利于增强树体对生物及非生物胁迫的适应能力,提高葡萄产量和品质。葡萄砧木品种多样复杂,识别难度较大,深度学习能够快速提取图像的深层特征,被广泛应用于植物图像分类识别领域。本研究以30份葡萄砧木成龄叶图像作为研究对象,通过采集叶片图像,构建了一个包含13547张的葡萄砧木叶片图像的数据集。采用GoogleNet、ResNet-50、ResNet-101以及VGG-16等4个卷积神经网络对其进行自动识别。结果表明：精度最高的分类网络为ResNet-101,在最优模型参数（学习率：0.005,最小批次：32,迭代次数：50）下精度达到97.5%。ResNet-101模型检测的30个品种中,平均预测精确率为92.59%,有7个品种的预测精确率达到100%;平均召回率为91.08%,有8个品种的召回率达到100%,叶片的叶面纹理、叶脉以及叶缘部分对品种识别的影响最大。以上结果证实,深度学习网络模型可以实现对葡萄砧木的自动实时识别,为葡萄砧木品种的保护、利用、分类研究以及其他农作物的品种识别提供参考。     

DCAMCP: A deep learning model based on capsule network and attention mechanism for molecular carcinogenicity prediction

The carcinogenicity of drugs can have a serious impact on human health, so carcinogenicity testing of new compounds is very necessary before being put on the market. Currently, many methods have been used to predict the carcinogenicity of compounds. However, most methods have limited predictive power and there is still much room for improvement. In this study, we construct a deep learning model based on capsule network and attention mechanism named DCAMCP to discriminate between carcinogenic and non-carcinogenic compounds. We train the DCAMCP on a dataset containing 1564 different compounds through their molecular fingerprints and molecular graph features. The trained model is validated by fivefold cross-validation and external validation. DCAMCP achieves an average accuracy (ACC) of 0.718 ± 0.009, sensitivity (SE) of 0.721 ± 0.006, specificity (SP) of 0.715 ± 0.014 and area under the receiver-operating characteristic curve (AUC) of 0.793 ± 0.012. Meanwhile, comparable results can be achieved on an external validation dataset containing 100 compounds, with an ACC of 0.750, SE of 0.778, SP of 0.727 and AUC of 0.811, which demonstrate the reliability of DCAMCP. The results indicate that our model has made progress in cancer risk assessment and could be used as an efficient tool in drug design.     

基于Faster R-CNN的美国白蛾图像识别模型研究

昆虫监测中美国白蛾Hyphantria Cunea的人工辨识、分类费时费力,且主观性强。本文利用RPN人工神经网络模型对美国白蛾图像数据进行特征提取,并对比分析Inception_v2,ResNet50,ResNet101网络模型,设计了一种改进的美国白蛾人工神经网络识别模型IHCDM（Improved Hyphantria Cunea Artificial Neural Network Recognition Model,IHCDM）,采用端到端方法在GPU处理器上对该模型进行了训练,并对其进行了实验验证。结果表明：该模型对美国白蛾的识别准确率可达99.5%,相比于ResNet50与ResNet101网络模型,识别准确率提高了0.5%与0.4%。超参数微调后,在置信度阈值为0.85时,识别准确率99.7%,识别速度0.09 ms/张。IHCDM模型为美国白蛾的快速辨识、分类提供了一种新方法。     

Neural network modelling of the influence of channelopathies on reflex visual attention

This paper introduces a model of Emergent Visual Attention in presence of calcium channelopathy (EVAC). By modelling channelopathy, EVAC constitutes an effort towards identifying the possible causes of autism. The network structure embodies the dual pathways model of cortical processing of visual input, with reflex attention as an emergent property of neural interactions. EVAC extends existing work by introducing attention shift in a larger-scale network and applying a phenomenological model of channelopathy. In presence of a distractor, the channelopathic network’s rate of failure to shift attention is lower than the control network’s, but overall, the control network exhibits a lower classification error rate. The simulation results also show differences in task-relative reaction times between control and channelopathic networks. The attention shift timings inferred from the model are consistent with studies of attention shift in autistic children.     

Long Term Effect of Cyprinid Fishes on Phytoplankton and Zooplankton Communities in a Shallow Water Protection Reservoir

The effects of fish kill and different fish stocks on the phytoplankton and zooplankton dynamics were studied in a shallow hypertrophic reservoir system. When fish stock was below 100 kg ha⁻¹, nutrient availability was not the main limiting factor for growth of phytoplankton. Consequently top‐down forces controlled phytoplankton. In the years with high fish stock (>100 kg ha⁻¹) the bottom‐up forces dominated as nutrient availability was the main limiting factor for growth of phytoplankton. We can conclude that significant water quality improvement can be achieved in the reservoir system by decreasing fish stock below 100 kg ha⁻¹. Although clear‐water phase could be stabilised temporary by macrophytes, stabilisation of good water quality requires continuous regulation of fish community. (© 2004 WILEY‐VCH Verlag GmbH & Co. KGaA, Weinheim)     

The Effect of Artificial Destratification on Phytoplankton in a Reservoir

The effects of artificial destratification on limnological conditions and on phytoplankton were surveyed for 6 years (1995-2000) in Lake Dalbang (South Korea), a water supply reservoir receiving nutrients from agricultural non-point sources. In order to reduce odor problems caused by cyanobacterial blooms, six aerators were installed in 1996 and operated regularly during the warm season. Aeration destratified the water column of the reservoir and produced homogeneous physical and chemical parameters. The maximum surface temperature in summer decreased from 28.9 °C before aeration to 20.0-26.4 °C after aeration, whereas the maximum hypolimnetic temperature increased from 8.0 to 17.0-23.7 °C. Despite these changes, surface water concentrations of total phosphorus (TP) and chlorophyll a(CHLA) and their seasonal patterns did not change with destratification. Phosphorus loading was concentrated in heavy rain events during the summer monsoon, and TP and CHLA reached maximal concentrations in late summer after the monsoon. Because the hypolimnion was never anoxic prior to aeration, internal loading did not seem to be substantial. Cyanobacteria were the dominant phytoplankton in summer before aeration, but diatoms replaced them after operation of the aerator. Cyanobacteria blooms were eliminated. In contrast, total algal biomass in the water column (as CHLA integrated over depth) increased from 190 mg m^–2 in 1995 to 1150, 300, 170, and 355 mg m^–2 in 1997, 1998, 1999, and 2000, respectively. The increased ratio of mixing depth to euphotic depth to 2.5 may have resulted in a net reduction in the amount of underwater irradiance experienced by phytoplankton cells, and this may have favored the switch to diatom dominance. Furthermore, the mixing may have allowed diatoms to flourish in summer by lowering their settling loss that would be critical in stratified water columns. In conclusion, the destratification in this reservoir was effective in preventing cyanobacteria blooms, but not in reducing the total algal standing crop.     

黟县宏村水系浮游植物调查与水质评价

2002年9月至2004年4月对黟县宏村水系进行四季调查,共采集到浮游植物122属种。依据调查结果选用“污染指示种”、“指示群落”两种方法对宏村水系水质污染状况进行了评价。研究表明,宏村水系属微污——甲型中污水体。     

Phytoplankton and zooplankton as indicators of water quality in Discovery Bay,Jamaica

Several investigations exist which use planktonic communities as indicators of water quality in Jamaican and Caribbean Bays, however, few are conducted before there are obvious effects of eutrophication. Therefore, most of our ‘baseline’ data are for bays already severely affected by pollution. This study was conducted to assess water quality in Discovery Bay, Jamaica, before there were severe signs of eutrophication. The bay was monitored over a 12-month period (October 1995–September 1996) using 10 stations. Physicochemical data indicated a well mixed upper 5 m of water column, below which discontinuities in temperature/salinity profiles indicated the influence of colder, more saline waters associated with deep offshore currents. Physicochemical variables were within the range for oligotrophic systems with a tendency towards mesotrophic in localized areas close to the shoreline. Signs of anthropogenic stress were associated with the eastern, southwestern and western sections of the bay. Of the over 120 species of phytoplankton found in the waters of Discovery Bay, most were neritic/oceanic and diatoms dominated while 11 were found to be potentially harmful species. While these harmful species occurred at all stations they occurred most frequently at stations on the eastern side of the bay. About 107 zooplankton species were identified, 52 of which were copepods. The species also represented a mix of neritic and oceanic taxa and mean abundances for the area ranged from 1077 m⁻³ at the mouth of the bay to 3794 m⁻³ close to the south shore (station 6). Generally stations closest to shore had greater zooplankton abundances than centrally located bay stations and stations close to oceanic influence. Acartia tonsa and Lucifer faxoni showed greatest densities at shoreline areas of the bay while Oithona plumifera, Undinula vulgaris and Temora stylifera were important at stations closest to oceanic influences. These species were thus considered as indicators of these different areas within the bay. From physicochemical data and the planktonic assemblage, Discovery Bay cannot be considered polluted, it is still more accurately classified as generally pristine with mesotrophic zones in the eastern and southeastern sections of the bay. These data therefore provide a real baseline of conditions for similar tropical coastal embayments.     

Visual and buying sequence features-based product image recommendation using optimization based deep residual network

A recommendation system is an imaginative resolution for managing the restrictions in e-commerce services with item details and user details. Also, it is used to determine the user preferences to recommend the items they expected to buy. Several conventional collaborative filtering techniques are devised in the recommender model, but it has some complexities. Hence, an innovative optimization-driven deep residual network is devised in this paper for a product recommendation system. Here, the product of images is used for extracting features where the Convolutional neural network (CNN) features are computed, and then it is given as input to the deep residual network aimed at product recommendation. The deep residual network is trained using developed Elephant Herding Feedback Artificial Optimization (EHFAO), which is obtained by integrating Elephant Herding optimization (EHO) into the Feedback Artificial Tree (FAT). Here, the item grouping is carried out on input data based on K-means clustering. After item grouping, Cosine similarity is used to perform matching of groups, where the best group is acquired among all the available groups. Extraction of list of visitors is done from the best group. Then, the list of items is obtained from the sequence of best visitor. Next, the corresponding binary sequence is obtained for the applicable sequence of visitor. From this sequence of best visitor, the recommended product is acquired. Then, the recommended product is subjected to the sentiment analysis for which the score is determined. Here, the sentiment analysis helps to decide whether the product is recommended or not recommended. If the score is positive, then the same product is recommended; otherwise, the new product is recommended. The proposed EHFAO-based deep residual network attained better performance in comparison to the other techniques with a maximal F-measure at 84.061%, 84.061% precision, 87.845% recall along with minimal Mean Squared Error (MSE) of 0.216.