Abstract: A system including a transmission terminal, used for transmitting a target data packet obtained by encrypting and encapsulating a data flow; a network controller, deployed in a first Internet network and storing routing information from the transmission terminal to a receiving terminal; and a backbone transmission network device, deployed in a backbone transmission network and used for obtaining the routing information and the target data packet forwarded by the first Internet network and forwarding the target data packet to the receiving terminal based on the routing information. The present disclosure solves the technical problem of low efficiency and high cost of data transmission in the conventional techniques.

Abstract: A method and an apparatus for generating a video fingerprint are disclosed. The method includes: performing shot boundary detection on content of a video; determining a time duration of each shot according to positional points of the shot boundary, and compose the time duration of each shot into a shot boundary time slice sequence; and obtaining video fingerprint information according to the time slice sequence.

Abstract: Embodiments of the disclosure provide devices and methods for performing a top-k function. The device can include: a memory comprising a plurality of register files for storing the data elements, the plurality of register files comprising a parent register file and a first child register file associated with the parent register file, wherein the parent register file is associated with: first interface circuitry configured for reading a first parent data element from the parent register file and receiving a first child data element and a second child data element from the first child register file; and first comparison circuitry configured for updating the parent register file and the first child register file based on the first parent data element, the first child data element, and the second child data element according to a given principle.

Abstract: Methods and computer systems for storing a query plan in a cache are provided. The method can include: determining a use frequency of the query plan; determining a target storage layer for storing the query plan based on the determined use frequency, wherein the cache includes at least two storage layers and each of the storage layers is associated with a compression level of query plans; processing the query plan according to the compression level of the target storage layer; and storing the processed query plan in the target storage layer.

Abstract: A method for performing motion compensation includes: receiving a first wrap-around motion compensation flag associated with one or more pictures and indicating whether horizontal wrap-around motion compensation is enabled for the one or more pictures; and in response to the horizontal wrap-around motion compensation being enabled for the one or more pictures, receiving a parameter associated with a wrap-around motion compensation offset, the wrap-around motion compensation offset being associated with the one or more pictures, wherein a value of the wrap-around motion compensation offset is less than or equal to a difference minus 2, and the difference is obtained by a quotient of a picture width in luma samples divided by a minimum luma coding block size minus a quotient of a luma coding tree block size divided by the minimum luma coding block size.

Abstract: A micro kernel scheduling method and apparatus are disclosed in embodiments of this disclosure. The method is applied to a software platform and includes: receiving a scheduling instruction for a current micro kernel; and switching the current micro kernel to a target micro kernel. In some embodiments, a micro kernel is switched directly according to a scheduling instruction, and this is completed without any thread of the software platform, which solves the problems in the conventional system of high micro kernel switching cost and poor real-time performance caused by one-to-one correspondence between micro kernels and threads of the software platform.

Abstract: Methods and systems are provided for improving user authentication and access control by a network file system service in a multi-tenant public cloud environment by receiving a request for a connection to a file system from a file system client (client), sending an identification request for identification authentication of the client to a control system, receiving a response from the control system, establishing the connection to the file system upon determining that the connection to the file system is allowed based on cloud tenant information associated with the client, receiving an attempt to access the file system from the client by a sub-user, authenticating the sub-user based on the cloud tenant information, issuing a security token including a globally unique sub-user identifier of the sub-user, and using the security token to determine access rights of the sub-user to the file system for a subsequent request.

Abstract: The present disclosure provides systems and methods for processing video content. The method can include: partitioning, along a partitioning edge, a plurality of blocks associated with a picture into a first partition and a second partition; performing inter prediction on the plurality of blocks, to generate a first prediction signal for the first partition and a second prediction signal for the second partition; and blending the first and second prediction signals for edge blocks associated with the partitioning edge.

Abstract: The present disclosure provides methods, systems, and non-transitory computer readable media for managing a primary storage unit of an accelerator. The methods include assessing activity of the accelerator; assigning, based on the assessed activity of the accelerator, a lease to a group of one or more pages of data on the primary storage unit, wherein the assigned lease indicates a lease duration; and marking, in response to the expiration of the lease duration indicated by the lease, the group of one or more pages of data as an eviction candidate.

Abstract: Embodiments of the present disclosure provide a data migration method and apparatus. The method includes: receiving a migration task of migrating data in a first system to a second system; and calling upper-layer interfaces corresponding to the migration task, and calling underlying operation interfaces of the first system and the second system by the upper-layer interfaces through an abstract interface class to migrate the data of the first system to the second system. Bidirectional data transmission and migration can be implemented between any two data ends. Bidirectional data transmission and migration between any data terminals.

Abstract: The present application discloses a method, device, and system for processing a medical image. The method includes obtaining, by one or more processors, a target image, segmenting, by the one or more processors, a target region image from the target image, wherein the target region image comprises a target object region in the target image, analyzing, by the one or more processors, the target region image based at least in part on a machine learning model, and obtaining by the one or more processors, a recognition result based at least in part on the analysis of the target region image.

Abstract: This application discloses methods, apparatus, and electronic devices for face pose estimation and three-dimensional face reconstruction. The face pose estimation method comprises: acquiring a two-dimensional face image for processing, constructing a three-dimensional face model corresponding to the two-dimensional face image, and determining a face pose of the two-dimensional face image based on face feature points of the three-dimensional face model and face feature points of the two-dimensional face image. With this approach, the face pose estimation is performed based on the three-dimensional face model corresponding to the two-dimensional face image, instead of only based on a three-dimensional average face model. As a result, a high accuracy pose estimation can be obtained even for a face with large-angle and exaggerated facial expressions. Thus, robustness of the pose estimation can be effectively improved.

Abstract: A method for storing application program information including segmenting the application program information into program sub-information, with an information segmentation algorithm of the application program information having undergone algorithm obfuscation processing; and storing the program sub-information. The present disclosure significantly reduces the possibility of application program information being acquired by hackers and the like, and improves the security of application program information and electronic devices.

Abstract: Embodiments of the present disclosure provide interaction methods and devices. The method can include obtaining a display object comprising a plurality of collage pictures for display; acquiring, in response to a switch operation for at least one collage picture of the plurality of collage pictures of the display object, a first target collage picture, wherein a first target content type of the first target collage picture matches a content type of the at least one collage picture; and replacing the at least one collage picture with the first target collage picture.

Abstract: The present disclosure provides data-driven methods and apparatuses for predicting user inquiries. One exemplary method includes: collecting user behavior data and pre-processing the user behavior data when a user inquiry is received; extracting candidate user behavior data that is contributive to the user inquiry from the pre-processed user behavior data; screening the candidate user behavior data based on a set target behavior data set, and selecting candidate user behavior data that is contained in the target behavior data set; inputting the screened candidate user behavior data into a trained classifier model; and predicting an inquiry category to which the user inquiry belongs. One exemplary apparatus includes a pre-processing module, an extraction module, and a prediction module. The method and the apparatus embodiments of the present disclosure can improve the efficiency and accuracy of the prediction.

Abstract: The present disclosure discloses a system and method for distributed neural network training. The method includes: computing, by a plurality of heterogeneous computation units (HCUs) in a neural network processing system, a first plurality of gradients from a first plurality of samples; aggregating the first plurality of gradients to generate an aggregated gradient; computing, by the plurality of HCUs, a second plurality of gradients from a second plurality of samples; aggregating, at each of the plurality of HCUs, the aggregated gradient with a corresponding gradient of the second plurality of gradients to generate a local gradient update; and updating, at each of the plurality of HCUs, a local copy of a neural network with the local gradient update.

Abstract: Embodiments of the present disclosure present a hyper-square interconnect topology and advanced ring-based AllReduce operations. In some embodiments, a topology is provided that is an improvement over conventional interconnect topologies by eliminating delays associated with long wirings. In some embodiments, computing nodes are divided into sub-sections to better allocate computing tasks, and the system can be optimized to divide up the computing nodes by maximizing the number of square sub-sections in the topology. In some embodiments, the system can be optimized to select square sub-sections first for each computing task. Each sub-section can comprise some computing nodes or all computing nodes in the hyper-square interconnect topology. This flexibility allows the hyper-square interconnect topology to utilize the computing nodes more efficiently by assigning appropriate numbers of computing nodes to each computing task based on the computing need of the computing task.

Abstract: Methods and devices, the method including receiving a matrix of a neural network model; classifying at least a portion of the matrix as a first section based on a first distribution pattern of non-zero elements of the portion of the matrix; and identifying memory addresses of the non-zero elements in the first section of the matrix for loading, according to a first order determined based on the first distribution pattern, the non-zero elements in the first section into one or more vector registers.

Abstract: Methods and systems are provided for sequence-to-sequence conversion from unstructured search queries to structured database queries, so that lay persons may retrieve information from relational databases without specialized knowledge of database query languages. An encoder module and a decoder module of a learning model are trained to convert an unstructured search query to an intermediate feature vector by computing co-attention and self-attention based on a context string and a database schema, encoding the database schema in the context string by application of self-attention between the context string containing tokens of the database schema with learned structural attention heads which relate the token to logic of the database. Training is performed using labeled training datasets which include structured database queries which are normalized by parsing into a semantic representation thereof, followed by linearization.

Abstract: Methods and apparatus for reducing a size of a neural network model, the method including: compressing data of the neural network model; identifying structure information of a vector register, wherein the structure information includes a number of registers included in the vector register; comparing a number of elements in the compressed data with a first condition, wherein the first condition is determined based on the number of registers in the vector register; and in response to the number of elements satisfying the first condition, associating the compressed data with the vector register to enable loading the compressed data to the vector register.