Abstract: Placement of a tenant database in an oversubscribed, database-as-a-service cluster comprised of a plurality of nodes is described. The placement may be based on per-node estimates of a probability of resource demand violation if the tenant database is placed on the node. Past resource usage of similar tenant databases subscribed to the cluster that are collected and stored as compressed traces may be used to obtain the estimates. In some examples, based on the estimates, a per-node expected number of resource violations is determined and compared across nodes, where the determined placement minimizes the number of resource violations. In other examples, when the tenant database is being placed in parallel with other tenant databases, a score assigned to each valid configuration for the placement may be modified based on the estimates, where the determined placement is the configuration having a lowest score.

Abstract: Methods, systems, and computer program products are provided for a compute cluster comprising placement and load balancing (PLB) logic that receives data (e.g., state metadata) relating to a service (e.g., database service) executing on the compute cluster, from a resource manager executing on the compute cluster, via a first API associated with the resource manager. The PLB logic receives second data from the service via a second API and determines whether a PLB action is indicated based on one of the second data or a combination of the first data and the second data. When a PLB action is indicated, the PLB logic sends a command to the resource manager to execute the PLB action. The PLB logic also receives queries from clients external to the compute cluster and may spawn a child PLB logic to offload PLB operations, respond to queries, or perform software validation in the child.

Abstract: Techniques for differential overbooking on a cloud database. These techniques may include determining a reservation amount of a multi-tenant resource for a first service of a based upon an overbooking characteristic of the first service, and determining that a total usage value of the multi-tenant resource by a plurality of services is greater than a threshold value. In addition, the techniques may include determining a service usage value of the multi-tenant resource by the first service, determining a first overage value of the first service based on the service usage value and the reservation amount, and performing a resource reclamation process over the multi-tenant resource based on the first overage value of the first service.

Abstract: Placement of a tenant database in an oversubscribed, database-as-a-service cluster comprised of a plurality of nodes is described. The placement may be based on per-node estimates of a probability of resource demand violation if the tenant database is placed on the node. Past resource usage of similar tenant databases subscribed to the cluster that are collected and stored as compressed traces may be used to obtain the estimates. In some examples, based on the estimates, a per-node expected number of resource violations is determined and compared across nodes, where the determined placement minimizes the number of resource violations. In other examples, when the tenant database is being placed in parallel with other tenant databases, a score assigned to each valid configuration for the placement may be modified based on the estimates, where the determined placement is the configuration having a lowest score.

Abstract: Techniques for differential overbooking on a cloud database. These techniques may include determining a reservation amount of a multi-tenant resource for a first service of a based upon an overbooking characteristic of the first service, and determining that a total usage value of the multi-tenant resource by a plurality of services is greater than a threshold value. In addition, the techniques may include determining a service usage value of the multi-tenant resource by the first service, determining a first overage value of the first service based on the service usage value and the reservation amount, and performing a resource reclamation process over the multi-tenant resource based on the first overage value of the first service.

Abstract: This application is directed to a geographic information display method. While a first mobile device is executing an instant messaging application, it obtains its own geographic location information, a second mobile device's geographic location information, and a first timestamp at which the geographic location information is recorded. The first mobile device then displays an interactive map associated with the first timestamp on a user interface. The displayed map includes two marks that represent the locations of the first and second mobile devices, respectively, and instant messages exchanged between these two mobile devices are displayed on the map adjacent to the two marks. As the first mobile device obtains updated geographic location information in association with a second timestamp, it updates the location of the second mobile device while the mark corresponding to the first mobile device remains at the same location on the user interface.

Abstract: A method including: receiving, from a first terminal, a request for establishing a chat group without receiving an identifier of an existing social network contact; monitoring incoming requests for additional users to join the first chat group; while monitoring, detecting a request from a second terminal to obtain information of other terminals located around the second terminal; in accordance with the request from the second terminal and in accordance with a determination that the second terminal and the first terminal are located within a preset distance, requesting authentication information corresponding to the first chat group; receiving the authentication information from the second terminal, wherein the authentication information is obtained by the second user independently of the server of the social network platform; and in accordance with a determination that the authentication information is correct, establishing the first chat group including the first user and the second user.

Abstract: A high parallelism computing system and instruction scheduling method thereof are disclosed. The computing system comprises: an instruction reading and distribution module for reading a plurality of types of instructions in a specific order, and distributing the acquired instructions to corresponding function modules according to the types; an internal buffer for buffering data and instructions for performing computation; a plurality of function modules each of which sequentially executes instructions of the present type distributed by the instruction reading and distribution module and reads the data from the internal buffer; and wherein the specific order is obtained by topologically sorting the instructions according to a directed acyclic graph consisting of the types and dependency relationships.

Abstract: A message forwarding method performed at an electronic device having one or more processors and memory storing a plurality of programs for forwarding messages using an instant messaging application, includes: displaying a dialog box including one or more chat messages associated with a first user account of the instant messaging application; selecting one or more chat messages in the dialog box; obtaining message content and associated information of each selected chat message, the associated information including one or more of: a message sender and a sending time of the chat message, a group name of a group corresponding to the dialog box, identifiers of participants of the group; and forwarding the message content and the associated information of each chat message to a second user account of the instant messaging application.

Abstract: The present invention relates to artificial neural networks, for example, deep neural networks. In particular, the present invention relates to a compression method considering load balance for deep neural networks and the device thereof. More specifically, the present invention relates to how to compress dense neural networks into sparse neural networks in an efficient way so as to improve utilization of resources of the hardware platform.

Abstract: The technical disclosure relates to artificial neural network. In particular, the technical disclosure relates to how to implement efficient data access control in the neural network hardware acceleration system. Specifically, it proposes an overall design of a device that can process data receiving, bit-width transformation and data storing. By employing the technical disclosure, neural network hardware acceleration system can avoid the data access process becomes the bottleneck in neural network computation.

Abstract: The present disclosure relates to a processor for implementing artificial neural networks, for example, convolutional neural networks. The processor includes a memory controller group, an on-chip bus and a processor core, wherein the processor core further includes a register map, an instruction module, a data transferring controller, a data writing scheduling unit, a buffer module, a convolution operation unit and a hybrid computation unit. The processor of the present disclosure may be used for implementing various neural networks with increased computation efficiency.

Abstract: The present invention relates to artificial neural networks, for example, deep neural networks. In particular, the present invention relates to a compression method for deep neural networks with proper use of mask and the device thereof. More specifically, the present invention relates to how to compress dense neural networks into sparse neural networks while maintaining or even improving the accuracy of the neural networks after compression.

Abstract: The present disclosure relates to a processor for implementing artificial neural networks, for example, convolutional neural networks. The processor includes a memory controller group, an on-chip bus and a processor core, wherein the processor core further includes a register map, an instruction module, a data transferring controller, a data writing scheduling unit, a buffer pool, a data reading scheduling unit and a computation module. The processor of the present disclosure may be used for implementing various neural networks with increased computation efficiency.

Abstract: The present technical disclosure relates to artificial neural networks, e.g., gated recurrent unit (GRU). In particular, the present technical disclosure relates to how to implement a hardware accelerator for compressed GRU based on an embedded FPGA. Specifically, it proposes an overall design processing method of matrix decoding, matrix-vector multiplication, vector accumulation and activation function. In another aspect, the present technical disclosure proposes an overall hardware design to implement and accelerate the above process.

Abstract: Provided in the present disclosure is a method of pruning a convolutional neural network based on feature map variation. The present invention enables compression of an entire network by means of removing a portion of filters in a convolutional layer, and such process is called pruning. A main contribution of the present invention is determining a pruning rule for filters in a single convolutional layer according to a feature map variation condition, using the rule to analyze network sensitivity, and pruning the entire network according to the network sensitivity.

Abstract: A multi-iteration method for compressing a deep neural network into a sparse neural network without degrading the accuracy is disclosed herein. In an example, the method includes determining a respective initial compression ratio for each of a plurality of matrices characterizing the weights between the neurons of the neural network, compressing each of the plurality of matrices based on the respective initial compression ratio, so as to obtain a compressed neural network, and fine-tuning the compressed neural network.

Abstract: The disclosure provides a compilation method and system for heterogeneous computing platform, and a runtime method and system for supporting program execution on the heterogeneous computing platform. Inputting a trained neural network model to a Neural Network (NN) optimizing compiler to generate an NN assembly file corresponding to the neural network; inputting the NN assembly file to an NN assembler to generate an NN binary file corresponding to the neural network; compilation and assembling a neural network application developed by users in a high-level language using a host compiler toolchain to generate a corresponding host assembly file and a host binary file in sequence; and linking the NN binary file and the host binary file using a host linker to generate a single hybrid linking executable file. The technical solution of the present disclosure has the advantages such as good computing performance, strong scalability, strong compatibility and high flexibility.

Abstract: Method and device of storing content objects within a social network application are disclosed. The method includes: a user device providing a user interface of the social network application on the user device, the user interface including an affordance for forming a content group including respective content objects of one or more object types; receiving a selection of a first plurality of images for inclusion in the first content group; providing a listing of context objects that are associated with at least one of the first plurality of images; receiving a user input indicating completion of the first content group; and forming the first content group containing the first plurality of images and the one or more content object selected from the listing of context objects information.

Abstract: The present invention relates to recurrent neural network. In particular, the present invention relates to how to implement and accelerate a recurrent neural network based on an embedded FPGA. Specifically, it proposes an overall design processing method of matrix decoding, matrix-vector multiplication, vector accumulation and activation function. In another aspect, the present invention proposes an overall hardware design to implement and accelerate the above process.