摘要：【中文】本申请提供了一种模型训练方法、图像特征提取方法、装置及电子设备，其中，所述模型训练方法包括：使用训练数据对待训练模型进行第一阶段训练，得到第一阶段模型，其中，所述第一阶段训练中使用第一损失函数作为监督信号；使用所述训练数据对所述第一阶段模型进行第二阶段训练，得到第二阶段模型，其中，所述第二阶段训练使用所述第一损失函数和第二损失函数作为监督信号，所述训练数据为一图像数据集；使用所述训练数据对所述第二阶段模型进行第三阶段训练，得到目标模型，其中，所述第三阶段训练中使用第二损失函数作为监督信号。通过三个阶段分别训练得到的目标模型能够更好地分类图像特征。 【EN】The application provides a model training method, an image feature extraction device and electronic equipment, wherein the model training method comprises the following steps: carrying out first-stage training on a model to be trained by using training data to obtain a first-stage model, wherein a first loss function is used as a supervision signal in the first-stage training; performing second-stage training on the first-stage model by using the training data to obtain a second-stage model, wherein the first loss function and the second loss function are used as supervision signals in the second-stage training, and the training data is an image data set; and performing third-stage training on the second-stage model by using the training data to obtain a target model, wherein a second loss function is used as a supervision signal in the third-stage training. The target models obtained through the respective training of the three stages can better classify the image characteristics.

摘要：【中文】本申请提供了一种车辆属性识别方法及其模型训练方法、装置和电子设备；其中，该车辆属性识别模型训练方法中，通过多层卷积层输出每层第一指定卷积层对应的特征图；通过每层第一指定卷积层所连接的组中的池化层和全连接层对第一指定卷积层对应的特征图进行降维处理；并通过每组中的损失函数，计算降维得到的特征向量的损失值；基于每组的损失值对初始模型进行训练，直至每组的损失值均收敛，得到目标模型。本申请实施例中，模型的网络结构较为简单，并且实现了各车辆属性对应特征的数据共享，减少了数据冗余，通过简单的模型结构即可实现车辆的多属性识别，提高了车辆属性识别的实时性，扩展了车辆属性识别的应用范围。 【EN】The application provides a vehicle attribute identification method, a model training method and a model training device thereof, and electronic equipment; in the vehicle attribute recognition model training method, a characteristic diagram corresponding to each first appointed convolutional layer is output through a plurality of convolutional layers; performing dimension reduction processing on the feature map corresponding to the first appointed convolutional layer through the pooling layer and the full-connection layer in the group connected with each first appointed convolutional layer; calculating loss values of the feature vectors obtained by reducing the dimensions through the loss functions in each group; and training the initial model based on the loss values of each group until the loss values of each group are converged to obtain the target model. In the embodiment of the application, the network structure of the model is simple, the data sharing of the corresponding characteristics of the vehicle attributes is realized, the data redundancy is reduced, the multi-attribute identification of the vehicle can be realized through the simple model structure, the real-time performance of the vehicle attribute identification is improved, and the application range of the vehicle attribute identification is expanded.

摘要：【中文】一种聚(亚苯基醚)组合物包括特定量的聚(亚苯基醚)、烯基芳族单体与共轭二烯的氢化嵌段共聚物、聚丙烯或聚乙烯、低分子量聚丁烯、阻燃剂、液体紫外吸收剂和聚(氧化烯)。该组合物可以用作电线和电缆的绝缘或护套材料，其中包括该组合物的绝缘和护套材料可以具有降低的表面析出。 【EN】A poly (phenylene ether) composition includes specific amounts of a poly (phenylene ether), a hydrogenated block copolymer of an alkenyl aromatic monomer and a conjugated diene, polypropylene or polyethylene, a low molecular weight polybutylene, a flame retardant, a liquid ultraviolet absorber, and a poly (alkylene oxide). The composition may be used as an insulation or jacketing material for wire and cable, wherein the insulation and jacketing material comprising the composition may have reduced surface precipitation.

摘要：【中文】本发明涉及一种防撞式新能源汽车用充电装置，包括固定在地面上的基座、铰接安装在基座上的充电桩机体，在充电桩机体的左右两侧对称设置有后仰式缓冲装置，该后仰式缓冲装置可在充电桩机体受到正面的冲击并发生绕基座向后摆动时，对充电桩机体进行缓冲保护；在充电桩机体的左右两侧还设有与后仰式缓冲装置相连的缓冲气囊，该缓冲气囊的前端面超出充电桩机体的前端面，以吸收来自充电桩机体正面的冲击力。本发明能够在受到汽车撞击后，作适应性翻转而使充电桩免受损坏，还能进行移动防止大行程倒车的撞击，大大提高充电桩的安全性；本发明采用两级缓冲保护设计，对充电桩机体起到了良好的保护作用。 【EN】The invention relates to an anti-collision charging device for a new energy automobile, which comprises a base fixed on the ground and a charging pile body hinged on the base, wherein backstroke buffering devices are symmetrically arranged on the left side and the right side of the charging pile body, and can buffer and protect the charging pile body when the charging pile body is impacted from the front side and swings backwards around the base; the left and right sides that fills the electric pile organism still is equipped with the buffering gasbag that links to each other with formula buffer leans on backward, and the preceding terminal surface of this buffering gasbag surpasss the preceding terminal surface that fills the electric pile organism to the absorption comes from filling the positive impact force of electric pile organism. The invention can be adaptively turned over to prevent the charging pile from being damaged after the automobile is impacted, and can also be moved to prevent the impact of large-stroke reversing, thereby greatly improving the safety of the charging pile; the invention adopts a two-stage buffer protection design, and plays a good protection role on the charging pile body.

摘要：【中文】本发明涉及一种新能源电池用保护设备，包括与汽车车体固连且上端敞口的箱体以及铰接安装在箱体上端的盖板，还包括设置在箱体的内壳，该内壳用于盛放新能源电池，新能源电池的电线穿过内壳并从箱体上引出以为外界供电，该内壳的左右外侧壁及底部均与箱体之间通过缓冲组件隔离缓冲，箱体内还具有用于在盖板扣合时对盖板锁紧的锁紧机构；盖板的底部均设有弹簧，弹簧的底部固定有橡胶垫。本发明将新能源电池悬空隔离式的设计，有效避免了车辆行驶过程中的冲击对新能源电池的损坏，弹簧与橡胶垫的组合，以从上方对新能源电池予以束缚，从而将外界冲击能量从上下及左右方向快速耗散；锁紧机构对整体锁紧，同时具有自锁功能，安全性能高。 【EN】The invention relates to a protective device for a new energy battery, which comprises a box body fixedly connected with an automobile body and provided with an opening at the upper end, a cover plate hinged at the upper end of the box body, and an inner shell arranged on the box body, wherein the inner shell is used for containing the new energy battery; the bottom of apron all is equipped with the spring, and the bottom of spring is fixed with the rubber pad. The invention adopts the suspended isolated design of the new energy battery, effectively avoids the damage of the new energy battery caused by impact in the driving process of the vehicle, and the combination of the spring and the rubber pad restrains the new energy battery from the upper part, thereby quickly dissipating the external impact energy from the upper and lower directions and the left and right directions; the locking mechanism locks the whole body, and has a self-locking function and high safety performance.

摘要：【中文】本发明涉及应用于新能源汽车的储能系统的承载设备，包括底座以及分别固装在底座上的前立板，后立板；所述前立板中部区域安装有电机，电机连接有主套筒，主套筒在背离电机一侧滑动安装有伸缩杆，伸缩杆焊接有支承板。所述前立板，后立板二者的上端面均为圆弧面，前立板圆弧面上分别以支撑方式、前后滑动方式对应安装有活动杆、固定杆。本发明在具有保障太阳能光伏板最大限度的得到光能功能的基础上，能够在雨天来临时，通过控制活动杆的滑动，实现将幕布向右前侧展开，起到自动给予太阳能光伏板进行遮雨的效果，具有良好的保护作用。 【EN】The invention relates to a bearing device applied to an energy storage system of a new energy automobile, which comprises a base, a front vertical plate and a rear vertical plate, wherein the front vertical plate and the rear vertical plate are respectively fixedly arranged on the base; the motor is installed in preceding riser middle part region, and the motor is connected with the main sleeve, and the main sleeve has the telescopic link at deviating from motor one side slidable mounting, and the telescopic link welding has the bearing plate. The upper end surfaces of the front vertical plate and the rear vertical plate are both arc surfaces, and a movable rod and a fixed rod are correspondingly arranged on the arc surfaces of the front vertical plate in a supporting mode and a front-back sliding mode respectively. On the basis of ensuring that the solar photovoltaic panel obtains light energy to the maximum extent, the curtain can be unfolded towards the right front side temporarily in rainy days by controlling the sliding of the movable rod, so that the effect of automatically shielding the solar photovoltaic panel from rain is achieved, and the solar photovoltaic panel has a good protection effect.

摘要：【中文】本发明提供融合时间衰减因子的改进LFM协同过滤方法，包括如下步骤：S1)输入用户u，每个用户u评价过的物品i，用户对物品i评价的时间戳，物品分类数K；S2)从训练样本中获得用户u评价过的物品，数量为Sp，作为正样本；S3)从训练样本中获得用户u没有评价过的物品，数量为Sn,作为负样本，用正、负样本构建用户u的训练样本库；S4)构建引入时间衰减因子的LFM算法的损失函数，通过梯度下降法，用训练样本库进行求解，得出用户u的用户特征矩阵P和物品特征矩阵Q；S5)计算用户u对物品i的兴趣度；S6)对所有物品兴趣度进行排序，选出N个兴趣度最高且用户u没有评价过的物品，给用户u进行推荐。 【EN】The invention provides an improved LFM collaborative filtering method fused with time attenuation factors, which comprises the following steps: s1) inputting users u, the item i evaluated by each user u, the time stamp of the evaluation of the user on the item i and the item classification number K; s2) obtaining the items evaluated by the user u from the training samples, wherein the number of the items is Sp, and the items are used as positive samples; s3) obtaining articles which are not evaluated by the user u from the training samples, wherein the number of the articles is Sn, the articles are used as negative samples, and a training sample library of the user u is constructed by using the positive and negative samples; s4) constructing a loss function of an LFM algorithm introducing time attenuation factors, and solving by using a training sample library through a gradient descent method to obtain a user characteristic matrix P and an article characteristic matrix Q of a user u; s5) calculating the interest degree of the user u in the item i; s6), sorting the interest degrees of all the items, selecting the N items with the highest interest degrees and without evaluation by the user u, and recommending the N items to the user u.

摘要：【中文】本发明提供一种融合时间衰减因子的LFM训练样本构建方法，包括如下步骤：S1)获取正样本；S2)计算整个训练集中物品的流行度；S3)评估样本库数据的多样性；S4)给出不同样本库组成时的算法推荐准确度和召回率；S5)选择最优流行度占比的样本作为负样本；S6)将正、负样本共同组成用户的训练样本库。本发明的优点为：综合考虑物品流行度和样本多样性对推荐性能的影响并融合时间衰减因子，通过实验给出不同样本库组成时的算法推荐准确度和召回率，分析得出最优流行度占比，从而得到最优的负样本，获得较好的FC‑LFM算法训练效果。 【EN】The invention provides a construction method of an LFM training sample fused with a time attenuation factor, which comprises the following steps: s1) obtaining a positive sample; s2) calculating the popularity of the articles in the whole training set; s3) evaluating the diversity of the data of the sample library; s4) giving the algorithm recommendation accuracy and recall rate when different sample libraries are formed; s5) selecting a sample with the optimal popularity ratio as a negative sample; s6) the positive and negative examples are combined to form a training example library of the user. The invention has the advantages that: the influence of the popularity of the article and the diversity of the samples on the recommendation performance is comprehensively considered, time attenuation factors are fused, the algorithm recommendation accuracy and the recall rate when different sample banks are formed are given through experiments, the optimal popularity ratio is obtained through analysis, the optimal negative sample is obtained, and the better FC-LFM algorithm training effect is obtained.

摘要：【中文】本发明提供基于深度学习的文案生成方法、装置及电子设备，所述方法包括步骤：获取用户输入标题；采用jieba中文分词方法对所述标题进行分词操作；采用TF‑IDF算法对经过分词操作的所述标题进行关键词提取，以获得关键词集合；将所述关键词集合作为输入输送至预设文本生成算法模型中，将得到的输出作为文案。该装置采用上述方法，通过对文案数据做处理，提取出文案中重点描述的关键词作为模型的输入，构建新的训练数据，使得输入和输出有很强的对应关系，从而提高了模型生成的文案和输入间的相关性，大幅提高了生成文案的质量。 【EN】The invention provides a deep learning-based file generation method, a deep learning-based file generation device and electronic equipment, wherein the method comprises the following steps: acquiring a user input title; performing word segmentation operation on the title by adopting a jieba Chinese word segmentation method; extracting keywords from the titles subjected to word segmentation operation by adopting a TF-IDF algorithm to obtain a keyword set; and conveying the keyword set as input to a preset text generation algorithm model, and taking the obtained output as a file. By adopting the method, the device extracts key words which are mainly described in the file as the input of the model by processing the file data, constructs new training data, and ensures that the input and the output have strong corresponding relation, thereby improving the correlation between the file generated by the model and the input and greatly improving the quality of the generated file.

摘要：【中文】本发明公开一种基于局部测量的封闭空间面板声学贡献度识别方法,包括：在封闭空间内部设置两个半封闭的全息测量面，并测量全息测量面上的声压值；基于声学等效源原理，在封闭空间内部建立外声场虚源面和向内声场虚源面，建立全息测量面与声场虚源面之间的传递关系，计算向内声场虚源面上的源强列向量；利用待测振动面板表面的导纳边界条件及向内声场虚源面上的源强列向量，计算向内声场虚源面入射到待测振动面板表面所产生的散射声场，还原出自由场条件下待测振动面板上的声压和法向振速；将待测振动面板划分为若干个离散单元，计算待测振动面板对封闭声场的声学贡献度。本发明能够灵活快速地对封闭空间结构面板局部的声学贡献度进行识别。 【EN】The invention discloses a closed space panel acoustic contribution degree identification method based on local measurement, which comprises the following steps: two semi-closed holographic measuring surfaces are arranged in the closed space, and the sound pressure value on the holographic measuring surfaces is measured; based on the acoustic equivalent source principle, establishing an external sound field virtual source surface and an inward sound field virtual source surface in a closed space, establishing a transfer relation between a holographic measuring surface and the sound field virtual source surface, and calculating a source intensity column vector on the inward sound field virtual source surface; calculating a scattered sound field generated by the fact that the virtual source surface of the inward sound field is incident to the surface of the vibrating panel to be tested by utilizing the admittance boundary condition of the surface of the vibrating panel to be tested and the source strength column vector on the virtual source surface of the inward sound field, and reducing the sound pressure and the normal vibration velocity on the vibrating panel to be tested under the condition of a free field; dividing the vibration panel to be tested into a plurality of discrete units, and calculating the acoustic contribution degree of the vibration panel to be tested to the closed sound field. The method can flexibly and quickly identify the local acoustic contribution degree of the panel of the closed space structure.