Baidu Online Network Technology (beijing) Co., Ltd.

China University of Mining and Technology

Chinese Academy of Sciences

Shanghai Xiaodu Technology Co. Ltd.

Boe Technology Group Co., Ltd.

Shanghai Fudan Microelectronics Group Company Limited

Leland Stanford Junior University

Beijing Boe Optoelectronics Technology Co., Ltd.

Baidu USA LLC

Hefei Boe Optoelectronics Technology Co., Ltd.

Beijing Baidu Netcom Science and Technology Co., Ltd.

Baidu.com Times Technology (beijing) Co. Ltd.

Global Alliance for Tb Drug Development

Xuzhou Zhirun Mining Equipment Science and Technology Co., Ltd

Baidu Online Network Technology Co., Ltd.

Gang Zhang

Gongbo Zhou

Fan Jiang

Chaoquan Tang

Zhencai Zhu

Wei Li

Zongliang Huo

Tao He

Jingtuo Liu

Qiang Ju

Zhenzhi He

Hongwei Tang

Mengmeng Zhang

Li Wan

Shasha Wang

Xiaonan He

Zongliang Wen

Xue Dong

Haisheng Wang

Yingming Liu

Shengji Yang

Yang Liu

Hao Zhang

Kai Liu

Bo Zhang

Weijie Zhao

Wei Zhang

Shanhui Fan

Jia Liu

Meng Zhang

Guohua Cao

Ran Xu

Yuxing Peng

Chunlei Wang

Zhenkun Ma

Shuai Wang

Gang Shen

Hongyu Chen

Xin Zhang

Chun Li

Jun Yu

Yang You

Qing Yang

Jing Xue

Christopher Blair Cooper

Songyong Liu

Jiacheng Huang

Baopu Li

Zhihong Pan

Fenglei Wang

Dongfeng Zhang

Xiangtao Jiang

Lu Qian

Tianyu Wang

Wensong He

Ping Zhou

Xi Zhang

Yafei Miao

Xinya Zhang

Wenming Wang

Haihong Huang

Yanwen Fan

Liuhui Zhang

Zongfu Yu

Hualong Zu

Lingling Liu

Dayu Shen

Torbjorn Skauli

Hailiang Wang

Jingyu Ye

Xiaochen Jia

Renliang Weng

Zhongji Fan

Yuan Sun

Yixin Shi

Yunyun Zhang

Lingjin Xu

Jingbin Wang

Dali Yin

Yuan Hsin Peng

Yawen Diao

Qiang Zhu

Tingfei Wang

Tingsong Luo

Mingwei Li

Jiaguang Lu

Sunhonglu Dong

Wenxian Liu

Jin Lu

Qiao Gao

Nicholas Sergeant

Xiaoyan Zhao

Binna Liu

Yaohai Song

Zhihua Shen

Yuning Yang

Renquan Zhou

Yongxi Luo

Network architecture search (NAS) received a lot of attention. The supernet-based differentiable approach is popular because it can effectively share the weights and lead to more efficient search. However, the mismatch between the architecture and weights caused by weight sharing still exists. Moreover, the coupling effects among different operators are also neglected. To alleviate these problems, embodiments of an effective NAS methodology by similarity-based operator ranking are presented herein. With the aim of approximating each layer's output in the supernet, a similarity-based operator ranking based on statistical random comparison is used. In one or more embodiments, then the operator that possibly causes the least change to feature distribution discrepancy is pruned. In one or more embodiments, a fair sampling process may be used to mitigate the operators' Matthew effect that happened frequently in previous supernet approaches.

Neural architecture search via similarity-based operator ranking

A three-dimensional memory and a method for manufacturing the three-dimensional memory, the three-dimensional memory includes a storage unit and a logic control unit, a front of the storage unit and a front of the logic control unit are attached to each other, and the logic control unit is connected to a control circuit, wherein a second metal line of the storage unit and a first metal line of the storage unit are respectively disposed on upper and lower sides of a channel layer of the storage unit, and the first metal line and the second metal line are electrically connected to the control circuit.

Three-dimensional memory comprising storage unit and logic control unit and method for manufacturing the same

A storage unit, a method of manufacturing the storage unit, and a three-dimensional memory. The storage unit includes: a first conductivity-type substrate; a channel layer stacked on the first conductivity-type substrate in a first direction; a second conductivity-type conduction layer including a first part and a second part that are connected, the first part being located between the first conductivity-type substrate and the channel layer, and the second part being formed in a via hole passing through the channel layer; a channel passage layer penetrating the channel layer and the first part in a negative direction of the first direction, and extending into an interior of the first conductivity-type substrate; and an insulating layer located in the channel layer and surrounding a periphery of the channel passage layer. The first conductivity-type substrate and the second conductivity-type conduction layer provide carriers required for reading and erasing operations, respectively.

Storage unit and method of manufacturing the same and three-dimensional memory

Provided is an L-shaped stepped word line structure including: L-shaped word line units, each including a long side extending in a second direction and arranged adjacent to a gate line slit, and a short side extending in a first direction. A word line terminal included in the short side is formed in a stepped stacked layer structure including stacked layer pairs formed of an insulating material, a region close to the gate line slit in a stacked layer of each stacked layer pair serves as a replacement metal region including a short side region surface/internal metal layer respectively located on a surface/in an interior. In a first direction, a length of the short side region surface metal layer is greater than that of the short side region internal metal layer, and the word line terminal corresponds to the short side region surface metal layer.

L-shaped stepped word line structure, method of manufacturing the same, and three-dimensional memory

The present disclosure discloses a method and apparatus for generating an image. A specific embodiment of the method comprises: acquiring at least two frames of facial images extracted from a target video; and inputting the at least two frames of facial images into a pre-trained generative model to generate a single facial image. The generative model updates a model parameter using a loss function in a training process, and the loss function is determined based on a probability of the single facial generative image being a real facial image and a similarity between the single facial generative image and a standard facial image. According to this embodiment, authenticity of the single facial image generated by the generative model may be enhanced, and then a quality of a facial image obtained based on the video is improved.

Method and apparatus for generating image

A timing controller includes: a field programmable gate array configured to generate a reference clock signal, and obtain at least one group of clock signals according to the reference clock signal. Each group of clock signals includes at least two clock signals, and a waveform of each clock signal is same as a waveform of the reference clock signal, and active levels in different clock signals are provided with a delay of a preset duration. The reference clock signal includes a first clock sub-signal for first duration and a second clock sub-signal for a second duration. At least one output interface group is connected to the field programmable gate array. Each output interface group includes at least two output interfaces, and each of the at least two output interfaces is configured to output one clock signal of a group of clock signals corresponding to the output interface group.

Timing controller, display apparatus and display control method thereof

An electronic-cigarette cartridge and an electronic cigarette are provided. The electronic-cigarette cartridge includes: a first contact point and a second contact point which are connected to a base part of an electronic cigarette, wherein the first contact point is configured to connect a control signal, and the second contact point is configured to input a reference voltage signal; an authentication module, wherein one end of the authentication module is connected to the first contact point, another end is connected to the second contact point, and is configured to interact with the base part of the electronic cigarette to conduct anti-counterfeiting authentication; and a heating module which is connected in parallel with the authentication module and is configured to performing heating after the anti-counterfeiting authentication is passed; wherein a working frequency of the authentication module is at least 10 times that of the heating module.

Electronic-cigarette cartridge and electronic cigarette

A method and apparatus for compressing a neural network are provided. A specific embodiment of the method includes: acquiring a to-be-compressed trained neural network; selecting at least one layer from layers of the neural network as a to-be-compressed layer; performing following processing steps sequentially on each of the to-be-compressed layers in descending order of the number of level of the to-be-compressed layer: determining a pruning ratio based on a total number of parameters included in the to-be-compressed layer, selecting a parameter for pruning from the parameters included in the to-be-compressed layer based on the pruning ratio and a parameter value threshold, and training the pruned neural network based on a preset training sample using a machine learning method; and determining the neural network obtained after performing the processing steps on the selected at least one to-be-compressed layer as a compressed neural network, and storing the compressed neural network.

Method and apparatus for compressing neural network

A method and apparatus for compressing a neural network are provided. A specific embodiment of the method includes: acquiring a to-be-compressed trained neural network; selecting at least one layer from layers of the neural network as a to-be-compressed layer; performing the following processing steps sequentially on each of the to-be-compressed layers in descending order of the number of level of the to-be-compressed layer in the neural network: quantifying parameters of the to-be-compressed layer based on a specified number, and training the quantified neural network based on a preset training sample using a machine learning method; and determining the neural network obtained after performing the processing steps on the selected at least one to-be-compressed layer as a compressed neural network, and storing the compressed neural network. This embodiment achieves efficient compression of the neural network.

The present disclosure provides an intelligent interaction processing method and apparatus, a device and a computer storage medium. The method comprises: performing intention recognition for a preceding feedback item already returned to the user; continuing to return a subsequent feedback item to the user based on the intention of the preceding feedback item. According to the present disclosure, it is possible to guess the user's subsequent intention based on the preceding feedback item, and continue to return the desired subsequent feedback item to the user without the user's operations, so that the present disclosure is more intelligentized and richer and simplifies the user's operations.

Intelligent interaction processing method and apparatus, device and computer storage medium

Disclosed are a wellbore inspection system and method for an ultra-deep vertical shaft. The wellbore inspection system includes a wire rope moving system, inspection robots, a visual image acquisition system, a wireless communication module, a central control system, and an image post-processing system of an upper computer. The wire rope moving system includes a surface wire rope guide rail, an underground wire rope guide rail, a surface wire rope moving device, an underground wire rope moving device, and a wire rope. The visual image acquisition system includes explosion-proof cameras. After image information acquired by the explosion-proof cameras is processed by a lower computer, the processed image is transmitted by a wireless image transmission module to the image post-processing system of the upper computer. The central control system is connected to the inspection robots and the wire rope moving system, and the inspection robots are connected to the central control system.

Wellbore inspection system and method for ultra-deep vertical shaft

A method and apparatus for presenting a state of a voice interaction device are provided. The method includes: acquiring a current state of the voice interaction device; determining a current presentation mode corresponding to the current state, by searching a pre-stored correspondence relation between a state and a presentation mode according to the current state; and presenting the current state of the voice interaction device in the determined current presentation mode. In embodiments, a current state of a voice interaction device is presented to a user, and necessary usage recommendations and prompts are provided to a user, thereby improving user experience.

Method, apparatus and device for presenting state of voice interaction device, and storage medium

Embodiments of the present disclosure provides a voice processing method, apparatus and device, where the method includes: acquiring, by a terminal device, first voice information; and acquiring, by the terminal device, response information corresponding to the first voice information, and performing an operation corresponding to the response information according to a type of the response information, where the type of the response information is at least one of a voice type, a text type, an image type, a video type, and a program operation type, which improves the flexibility of the voice processing.

Voice processing method, apparatus and device

A query processing method. A server node parses a query request and generates a corresponding logical plan; judges the logical plan belonging to an OLAP scale query service or an OLTP scale query service; in response to judging that the logical plan belongs to the OLAP scale query service, segmenting the logical plan into logical subplans. A coordinator node generates a first physical plan corresponding to the logical plan, and feeds back the first physical plan to the server node. The server node assigns each of the logical subplans to the corresponding worker node according to the first physical plan. The worker node processes the corresponding logical subplan to obtain a query subresult, and reports the processing subresult layer by layer. The server node summarizes the received processing subresult, generates a final query result, and feeds back the final query result to a user.

Query processing method, query processing system, server and computer readable medium

A movement-synchronized wellbore inspection system and a movement-synchronization control method thereof are disclosed. The wellbore inspection system comprises a rope-climbing robot, a wire rope, a ground wire rope moving device, a ground wire rope moving track, an underground wire rope moving device, an underground wire rope moving track, an inertial sensor and a control device. An upper end of the wire rope is connected to the ground wire rope moving device, and an lower end of the wire rope passes through the rope-climbing robot and is then connected to the underground wire rope moving device. The control device controls the underground and ground wire rope moving devices to move in synchronization, and then the inertial sensor carried on the rope-climbing robot detects posture data of the wire rope and transmits the data to the control device.

Movement-synchronized wellbore inspection system and movement synchronization control method thereof

Growing community of inventors

Beijing, China

Gang Zhang