Huawei Technologies Co., Limited

Spreadtrum Communications (Shanghai) Co., Ltd.

Chinese Academy of Sciences

Rda Microelectronics (Shanghai) Co., Ltd.

Rda Microelectronics Technologies (Shanghai) Co., Ltd.

Qinghua Yu

Long Wang

Li Zhou

Yanchao Xu

Sihai Ye

Xinhua Yang

Jauh Yuh-Ren

Yu Wang

Chi Zhang

Hongrui Jiang

Peng Wei

Lei Zhao

Jiayi Zhuang

Zhicheng Sui

Mohan Liu

Wenqiang Tian

Deming Yu

Xingpeng Zhao

Xiangyu Wang

Mingfang Pan

Qingchang Yang

Wenhua Fan

Yinzheng Zhou

Lixun Bai

This application relates to the field of artificial intelligence technologies, and discloses a machine learning model search method, a related apparatus, and a device. In the method, before model search and quantization, a plurality of single bit models are generated based on a to-be-quantized model, and evaluation parameters of layer structures in the plurality of single bit models are obtained. Further, after a candidate model selected from a candidate set is trained and tested, to obtain a target model, a quantization weight of each layer structure in the target model may be determined based on a network structure of the target model and evaluation parameters of all layer structures in the target model, a layer structure with a maximum quantization weight in the target model is quantized, and a model obtained through quantization is added to the candidate set.

Machine learning model search method, related apparatus, and device

A neural network model processing method includes obtaining a first low-bit neural network model through training, where the model includes a first operation layer and a second operation layer. Each operation layer includes at least one operation. Values/a value of a parameter and/or data used for the operation are/is represented by using N bits, and N is a positive integer less than 8. The neural network model processing method further includes compressing the model to obtain a second low-bit neural network model, where the compressed model includes a third operation layer. The third operation layer is equivalent to the first operation layer and the second operation layer, and an operation layer other than the third operation layer in the at least one operation layer is the same as an operation layer other than the first operation layer and the second operation layer in the at least two operation layers.

Neural network model processing method and apparatus

An audio output device, a method for eliminating sound leakage, and a storage medium are provided in the disclosure. The audio output device includes a processor, a first speaker, and a second speaker. The processor is configured to generate a second audio file based on a first audio file and a parameter adjustment value. The processor is configured to control the first speaker to play the first audio file to generate a first audio and synchronously control the second speaker to play the second audio file to generate a second audio. The first audio is transmitted to an ear of a user for the user to listen, and the second audio is used to eliminate an audio leaked to an outside of the audio output device during transmission of the first audio to the ear of the user.

Audio output device, method for eliminating sound leakage, and storage medium

A checkerboard imager comprises an aperture pair array in a rectangular shape, a 2D optical waveguide grating array, a 3D optical waveguide beam transmission array, a 2D optical waveguide quadrature modulation coupler array, and a photoelectric conversion data acquisition and image processing module. An object light is converged by sub-apertures of the aperture pair array, collected by the 2D optical waveguide grating array, and split into narrow-spectrum beams which are output to the 3D optical waveguide beam transmission array for cross-pairing, modulated and coupled by the 2D optical waveguide quadrature modulation coupler array, and reach the photoelectric conversion data acquisition and image processing module to obtain an object image. A method for implementing the checkerboard imager is provided where each module is independently manufactured and then integrated to improve yield of the modules and imager's optical efficiency, expand equivalent apertures, and improve working capability.

Checkerboard imager and implementation method therefor

A wireless roaming method and an apparatus, a storage medium and an access point device are disclosed. The wireless roaming method comprises: receiving, from an overlapping access point, first signal strength indication information that describes signal strength when a station transmits a data packet to the overlapping access point via a current channel, wherein the coverage of the overlapping access point overlaps with that of a current access point, the station accesses to the current access point, and the current access point operates on the current channel; measuring second signal strength information that describes signal strength when the station transmits a data packet to the current access point via the current channel; and sending the first signal strength indication information and the second signal strength indication information to the station, and/or generating a switching instruction based on the first signal strength indication information and the second signal strength indication information and sending at least the switching instruction to the station. The technical solution of the present disclosure can improve the efficiency of wireless roaming.

Wireless roaming method and device, storage medium, and access point device

A method and an apparatus for Bluetooth connection are disclosed. A first Bluetooth connection can be established between a first node of multiple data receiving nodes and a data source, the multiple data receiving nodes can share a data packet transmitted by the data source via the first Bluetooth connection, and a second node can be selected according to a preset rule from among the multiple data receiving nodes to feed back a response signal to the data source. In a method and an apparatus, the multiple data receiving nodes can be regarded as a whole, which share a data packet transmitted from the data source and jointly maintain the first Bluetooth connection, and only one of the multiple data receiving nodes is selected to feed back a response signal. Conforming to Bluetooth communication protocols, a method and an apparatus can meet the demanding requirements on hardware resources of a Bluetooth connection among a data source and multiple data receiving nodes. A one-to-many Bluetooth connection can be realized which saves power consumption without additionally upgrading the data source.

Bluetooth connection method and device

A wireless roaming method, an access point apparatus and a mobile station are disclosed. The wireless roaming method comprises: determining whether or not a key for a target access point corresponding to a mobile station is stored; if the key is not stored, performing key exchange on behalf of the mobile station with the target access point so as to obtain a key corresponding to the mobile station; and sending the key to the mobile station. In the technical solution of the present disclosure, before the mobile station connects to a target access point, a current access point, on behalf of the mobile station, obtains a key corresponding to the mobile from the target access point. Therefore, the mobile station does not need to perform key exchange with the target access point after accessing to the target access point, which may reduce roaming delays caused by key exchange and improve the efficiency of wireless roaming, and thereby improve user experience.

Wireless roaming method, access point device, and mobile station

A routing method and a device to provide a routing path that satisfies a quality of service (QoS) requirement, where the routing method includes performing QoS measurement on a path between the first routing node and a reachable adjacent routing node, to obtain a QoS parameter set of the path between every two adjacent routing nodes, reporting the QoS parameter set to the control node, receiving one or more routing tables of routing performed from the first routing node to the other routing node that are delivered by the control node, and receiving a packet, determining, from the one or more routing tables, a routing table corresponding to an identifier of the packet, and forwarding the received packet based on the routing table.

Routing method and device

An attack data packet processing method, an apparatus, and a system are provided. The method includes receiving, by a management node, description information of an attack data packet and an attack type of the attack data packet, where the description information and the attack type are sent by an awareness node; determining a processing policy on the attack data packet of the attack type according to the attack type; and sending the description information and the processing policy to a switch using a software-defined networking controller, so that the switch performs an operation indicated by the processing policy on the attack data packet with the description information.

Attack data packet processing method, apparatus, and system

A software defined networking (SDN) controller, a data center system, and a routing connection method relates to the field of communications technologies, where the method may use to resolve a technical problem in the prior art that a data center network cannot complete recovery of a communication fault. The SDN controller includes a receiver configured to receive a route calculation request sent by an endpoint device manager, where the route calculation request carries information about endpoint devices of both communication parties that need to perform route calculation, and a processor configured to perform route calculation for the endpoint devices of the both communication parties that need to perform route calculation, and control, according to calculated route information, a forwarding device to perform routing connection for the endpoint devices of the both communication parties.

SDN controller, data center system, and routing connection method

A method for implementing a session border controller (SBC) pool and a SBC device are provided. In the present invention, an SBC pool is formed by at least two SBCs, the at least two SBCs are mutually backed up for disaster recovery and use a same Internet Protocol (IP) address for a terminal device, and whether the terminal device is registered in the SBC pool is determined; and if the terminal device has been registered in the SBC pool, a service message is forwarded to an SBC with which the terminal device is registered. In this way, it is avoided that SBC device disaster recovery places a special requirement for the terminal device, and the terminal device does not need to be configured with two IP addresses. Furthermore, all devices in the SBC pool can process the service message of the terminal device, thereby increasing a resource utilization rate.

Method for implementing session border controller pool, and session border controller

Embodiments of the present invention provide a method, a device, and a system for realizing disaster tolerance backup. The method includes the following steps: a device with an active-standby function determines a current state of the device with the active-standby function, after it is determined that the current state is an active state, the device with the active-standby function issues reachable route information to a connected routing device, where the device with the active-standby function and a device of which a current state is a standby state have a same IP address, the reachable route information includes route information relevant to the IP address, so that the IP address is reachable, and the device with the active-standby function and the device of which the current state is the standby state are backup devices for each other. The present invention can realize remote disaster tolerance backup and avoid service interruption.

Growing community of inventors

Shanghai, China

Qinghua Yu