Advanced New Technologies Co., Ltd.

Alibaba Group Holding Limited

Hangzhou Hikvision Digital Technology Co., Ltd.

Alipay (hangzhou) Information Technology Co., Ltd.

Dingtalk Holding (cayman) Limited

Netease (hangzhou) Network Co., Ltd.

Alipay Information Technology Co., Ltd.

Yuan Zhang

Xinying Yang

Yize Li

Benquan Yu

Wenyuan Yan

Shiliang Pu

Lili Zhang

Hang Chen

Yi Zhang

Zhenhao Wu

Qing Ren

Feng Bao

Chenlu Guo

Liying Yu

Qi Dong

Kaikai Yu

Junshao Zhang

Jiarui Shi

Jie Du

Di Xie

Ke Xu

Jingjing Yan

Haohua Zhong

Jia Guo

Weijie Chen

Jairui Shi

A deep neural network operation method and apparatus are provided. The method comprises: obtaining an input feature map of a network layer; displacing respectively, according to a preset displacement parameter, each of channels of the input feature map of the network layer along axes, to obtain a displaced feature map, wherein the preset displacement parameter comprises displacement amounts of the channel in the axes; and performing convolution operation on the displaced feature map with a 1×1 convolution kernel to obtain an output feature map of the network layer. The operation efficiency of the DNN can be improved through the above method.

Deep neural network operation method and apparatus

An operation method and apparatus for a network layer in a Deep Neural Network are provided. The method includes: acquiring a weighted tensor of the network layer in the Deep Neural Network, the weighted tensor comprising a plurality of filters; converting each filter into a linear combination of a plurality of fixed-point convolution kernels by splitting the filter, wherein a weight value of each of the fixed-point convolution kernels is a fixed-point quantized value having a specified bit-width; for each filter, performing a convolution operation on input data of the network layer and each of the fixed-point convolution kernels, respectively, to obtain a plurality of convolution results, and calculating a weighted sum of the obtained convolution results based on the linear combination of the plurality of fixed-point convolution kernels of the filter to obtain an operation result of the filter; determining output data of the network layer.

Operation method and apparatus for network layer in deep neural network

Provided in one or more embodiments of the description are an information importing method and apparatus. The method may comprise: a server receiving an information importing file uploaded by a user, wherein the information importing file includes information to be imported into a system; the server verifying the information included in the information importing file; the server marking information in the information importing file that does not pass the verification, so as to generate an error comparison file; and the server returning the error comparison file to the user.

Information importing method and apparatus

Implementations of this specification provide service identifier-based data indexing. An example method performed by a data indexing apparatus of a centralized database service provider that stores data by using a blockchain-type ledger includes obtaining a service identifier and a service label of a data record, the service identifier uniquely identifying a service attribute of the data record, and the service label representing a type of the data record; determining location information of the data record in the blockchain-type ledger, the location information including a block height of a data block that includes the data record, and a positional offset of the data record in the data block that includes the data record; establishing a mapping relationship among the service identifier, the service label, and the location information; and writing the mapping relationship into an index for which a primary key is the service identifier.

Service identifier-based data indexing

A method and an apparatus for quantizing an activation volume of a deep neural network are disclosed. The method includes: obtaining an activation volume of a network layer in the deep neural network (S), wherein elements in the activation volume are arranged in three directions: height, width, and depth; dividing depth segments in the activation volume in which a difference among element features is smaller than a preset threshold into a same slice group along the depth direction of the activation volume, so as to obtain a plurality of slice groups (S); quantizing each slice group respectively by using a quantization parameter corresponding to each slice group obtained through a quantization formula (S). The quantization error can be reduced through the above method.

Deep neural network-based method and device for quantifying activation amount

Embodiments of the present application provide a method, an apparatus, a computer device for convolution operation and a computer readable storage medium. The method includes: obtaining input data of a network layer in a convolutional neural work; extracting each time a plurality of data points from the input data according to a preset step size; mapping the plurality of data points extracted each time to the same position at different depth levels of three-dimensional data to obtain rearranged data; and performing convolution operation on the rearranged data with a convolution kernel of a preset size to obtain a convolution result. Through the present solution, the operation efficiency of the convolutional neural network can be improved.

Convolution operation method and apparatus, computer device, and computer-readable storage medium

An information processing method and apparatus, a storage medium and an electronic device are provided. A first preset area is provided on a graphical user interface (GUI). A virtual joystick control is provided in the first preset area. In response to a first touch operation in the first preset area, the virtual joystick control is controlled according to the first touch operation. In a process of executing the first touch operation, in response to at least one second touch operation in the first preset area, an initial position of the at least one second touch operation is acquired. When the initial position is located in a second preset area, it is determined that the at least one second touch operation is an invalid operation. When the initial position is located in a third preset area, the virtual joystick control is controlled according to the at least one second touch operation to control the movement of the virtual character. The present disclosure improves the operation efficiency of a game as well as the experience of a player during game playing.

Information processing method and apparatus, storage medium, and electronic device

Disclosed herein are methods, systems, and apparatus, including computer programs encoded on computer storage media, for managing transaction requests in ledger systems. One of the methods includes: receiving a plurality of transaction requests by a computing system, each of the plurality of transaction requests including content information of a corresponding transaction and authentication information of the corresponding transaction, storing each of the plurality of transaction requests in a request stream by the computing system, obtaining storage information of a transaction request of the plurality of transaction requests in the request stream by the computing system, and storing the storage information of the transaction request in an index of an index stream and authentication information of a transaction corresponding to the transaction request in a storage element of a storage stream by the computing system, the storage element corresponding to the index, the index stream being associated with the storage stream.

Managing transaction requests in ledger systems

Implementations of this specification provide signature verification methods and apparatuses for a blockchain ledger. An example method includes receiving by a server, a signature verification instruction that comprises a verification object parameter and a hash value. The verification object parameter includes a third-party parameter, a platform parameter, or a time service certificate parameter, the verification object parameter indicates a type of a to-be-verified object, and the server is configured to store data by using the blockchain ledger. The server obtains the to-be-verified object based on the verification object parameter and the hash value. The type of the to-be-verified object includes a third-party digital signature, a server digital signature, or a time service certificate. The server sends the to-be-verified object to a client for verification by the client.

Signature verification for a blockchain ledger

Disclosed herein are methods, systems, and apparatus, including computer programs encoded on computer storage media, for managing blockchain-based centralized ledger systems. One of the methods includes: transmitting a timestamp request for a to-be-timestamped block in a blockchain to a trust time server by a ledger server in a centralized ledger system that stores data in the blockchain, the trust time server being associated with a trust time authority and independent from the centralized ledger system, the blockchain including a plurality of blocks storing transaction data, receiving a timestamp and associated signature for the to-be-timestamped block from the trust time server by the ledger server, and storing information of the timestamp and the associated signature for the to-be-timestamped block in the blockchain by the ledger server.

Managing blockchain-based centralized ledger systems

This disclosure relates to blockchain-type storage of receipt data. In one aspect, a method includes obtaining a to-be-stored data record including a first service attribute. Execution information of the to-be-stored data record is determined. A previous data record is obtained from a blockchain-type ledger. The previous data record includes a second service attribute that is the same as the first service attribute. Each data block in the blockchain-type ledger includes a block header and a block body. A hash value of the previous data record is determined. A receipt record is generated. The receipt record includes the hash value of the previous data record and the execution information of the to-be-stored data record. The to-be-stored data record and the receipt record are written into a same block body in the blockchain-type ledger.

Blockchain-type data storage

A client device displays a user interface, which includes a menu item associated with forced reminders. In response to a user selecting the menu item, the client device displays a forced-reminder user interface, which includes a content display area configured to display one or more forced-reminder messages.

Communication and messaging system

A computer-implemented method for performing authentication includes: determining, by a database server storing data in a blockchain ledger, a target ledger segment on which time service authentication is to be performed; generating a Merkle tree corresponding to the target ledger segment; determining a root hash of the Merkle tree, the root hash of the Merkle tree being based on a block hash of each data block in a set of one or more data blocks; executing a predetermined time capture process in a trusted execution environment to obtain a trusted time from an interface provided by a trusted time service organization; generating a digital signature for the trusted time and the root hash in the trusted execution environment; and generating a time service certificate including the trusted time, the root hash, and the digital signature.

Growing community of inventors

Hangzhou, China

Yuan Zhang