Microsoft Technology Licensing, LLC

Huawei Technologies Co., Limited

Cccc Highway Consultants Co. Ltd.

Chinese Academy of Sciences

Other

Intel Corporation

Boe Technology Group Co., Ltd.

Lenovo (beijing) Limited

China Communications Construction Company Limited

Beijing Xiaomi Mobile Software Co., Ltd.

Petrochina Company Limited

Shanghai Zhenhua Heavy Industries Company Limited

Sichuan Shengnuo Oil. and Gas Engineering Technology Service Co., Ltd

Cccc Wuhan Harbour Engineering Design & Research Co., Ltd.

Crsc Research & Design Institute Group Co., Ltd.

Xiaodong Liu

Hao Zhang

Ming Lin

Wei Lin

Vinodh Gopal

Gang Wang

Wenbo Li

Qiang Li

Jianfeng Gao

Weizhu Chen

Pengcheng He

Feng Pan

Shuo Liu

Ying Jiang

Qiang Wang

Ningning Li

Xibin Shao

Li Deng

Yu Liu

Xiaodong Wang

Min Zhu

Yu Wang

Ye-Yi Wang

Wei Zhou

Jun Hu

Xiaodong He

Ziye Yang

Zhao Zhang

Gang Cao

Jian Yang

Zhigang Zhang

Ruofei Zhang

Weigang Li

Jingjing Chen

Yanshun Chen

Changpeng Liu

Jiaheng Wang

Hai Ji

Hoifung Poon

Qing Yu

Yuxiong He

Jian Jiao

Hailong Wang

Xiaopeng Wang

Xin Jin

Ke Deng

Wen Xu

Zhiyu Liu

Kui Li

Chao Sun

Weiqing Zhu

Haiqing Yin

Jibing Gao

Jianyu Tang

Jingwei Liu

Huihai Shen

Qian Cheng

Nanjian Wu

Huiyun Ma

Yaping Liu

Zhen Liu

Weiwei Guo

Guanshan Tong

Yimin Wang

Haiyong Wang

Xiangjun Zhang

Changhong Wang

Lifeng Zhang

Jun Xie

Yuan Jiang

Qihua Dai

Minjia Zhang

Wenhan Wang

Kuang Ting Chuang

Junxiong Sun

Chenggang Yu

Kexin Liu

Hao Cheng

Yonggang Lv

Lingfeng Liu

Huaiping Su

Yuchao Lin

Kevin Duh

Wenfeng Lou

Yuling Wang

Liuqing Tu

Yuanyuan Lv

Weibin Chen

Yukun Fu

Yiyuan Wu

Qiang Yin

Dezheng Yi

Huawen Wang

Yanyan Liu

Haifeng Zhong

Shanshui Zhou

Feng Ye

Fusong Zhang

A downhole throttling device based on wireless control includes an inlet nozzle, a throttling assembly, an electrical sealing cylinder, a gas guide cylinder, a lower adapter sleeve, an end socket, a female sleeve, and electrical components. The inlet nozzle is connected to the throttling assembly, the throttling assembly is connected to the electrical sealing cylinder and the gas guide cylinder, the electrical sealing cylinder and the gas guide cylinder are both connected to the lower adapter sleeve, the lower adapter sleeve is respectively connected to the end socket and the female sleeve, and the electrical components are arranged in the electrical sealing cylinder. A throttling effect is achieved by detecting the temperature and pressure in a tube by a temperature/pressure sensor in the electrical components and controlling a motor to rotate a movable valve in the throttling assembly by a circuit control assembly, thereby achieving wireless control over downhole throttling.

Downhole throttling device based on wireless control

Generally discussed herein are devices, systems, and methods for generating an embedding that is both local string dependent and global string dependent. The generated embedding can improve machine learning (ML) model performance. A method can include converting a string of words to a series of tokens, generating a local string-dependent embedding of each token of the series of tokens, generating a global string-dependent embedding of each token of the series of tokens, combining the local string dependent embedding the global string dependent embedding to generate an n-gram induced embedding of each token of the series of tokens, obtaining a masked language model (MLM) previously trained to generate a masked word prediction, and executing the MLM based on the n-based induced embedding of each token to generate the masked word prediction.

ML using n-gram induced input representation

This document relates to training of machine learning models. One example method involves providing a machine learning model having one or more mapping layers. The one or more mapping layers can include at least a first mapping layer configured to map components of pretraining examples into first representations in a space. The example method also includes performing a pretraining stage on the one or more mapping layers using the pretraining examples. The pretraining stage can include adding noise to the first representations of the components of the pretraining examples to obtain noise-adjusted first representations. The pretraining stage can also include performing a self-supervised learning process to pretrain the one or more mapping layers using at least the first representations of the training data items and the noise-adjusted first representations of the training data items.

Adversarial pretraining of machine learning models

An electronic device includes a first portion and a second portion, an electric power generating device and a load device. The electronic device is disposed in an opening of a connected space, and the electronic device and a wall have the opening divide the connected space into the first space and the second space. The first portion is located in a first space having a first ambient temperature, and the second portion is located in a second space having a second ambient temperature. The electric power generating device is configured to generate a thermo-electromotive force based on a temperature difference between the first ambient temperature and the second ambient temperature to generate electrical energy. The load device is configured to obtain the electrical energy and operate on the electrical energy.

Electronic device

Knowledge graphs can greatly improve the quality of content recommendation systems. There is a broad variety of knowledge graphs in the domain including clicked user-ad graphs, clicked query-ad graphs, keyword-display URL graphs etc. A hierarchical Transformer model learns entity embeddings in knowledge graphs. The model consists of two different Transformer blocks where the bottom block generates relation-dependent embeddings for the source entity and its neighbors, and the top block aggregates the outputs from the bottom block to produce the target entity embedding. To balance the information from contextual entities and the source entity itself, a masked entity model (MEM) task is combined with a link prediction task in model training.

Learning graph representations using hierarchical transformers for content recommendation

Video playback quality detection methods, systems, and apparatuses are described. The method can include obtaining a first video data packet of a target video from a bearer network and obtaining a data byte quantity of first video data included in the first video data packet. The method may also include determining a buffer byte quantity stored in a video quality detection buffer. Furthermore, the method can include obtaining video frame data rate information corresponding to the target video and reducing the buffer byte quantity stored in the video quality detection buffer based on a video frame rate of the target video and the video frame data rate information of the target video. The method may further include determining a playback status of playing the target video on a video player, and determining video playback quality of the target video based on the playback status. According to the embodiments of the present invention, user experience and accuracy of the video playback quality detection method can be improved.

Video playback quality detection method and apparatus

Systems and methods are provided for facilitating the building and use of natural language understanding models. The systems and methods identify a plurality of tokens and use them to generate one or more pre-trained natural language models using a transformer. The transformer disentangles the content embedding and positional embedding in the computation of its attention matrix. Systems and methods are also provided to facilitate self-training of the pre-trained natural language model by utilizing multi-step decoding to better reconstruct masked tokens and improve pre-training convergence.

Efficient transformer language models with disentangled attention and multi-step decoding

A configuration apparatus includes a processor and a memory. The processor is configured to obtain first data of a first device, the first data indicating that the first device is ready for network configuration. The memory is configured to store network configuration data of a second device. The processor is further configured to: obtain the network configuration data of the second device, and send the network configuration data to the first device, the network configuration data being used to configure the first device to be in a same network connection state as the second device. A power path exists between the first device and the second device, and at least one of the first data or the network configuration data is transmitted through the power path.

Configuration device and method

An apparatus is described. The apparatus includes an accelerator to be coupled to a memory region that the accelerator shares with a virtualization environment comprising a guest OS, a guest VM and an SSD device driver. The accelerator is to forward a submission queue doorbell setting made by the SSD device driver in the shared memory to a corresponding submission queue doorbell in an SSD controller.

Shared memory mechanism to support fast transport of SQ/CQ pair communication between SSD device driver in virtualization environment and physical SSD

An image processing method and device are disclosed. The method is applicable to an image processing device having an operating system, and the method includes: receiving, by an image processing module of the operating system, an instruction of a first application program to call the image processing module of the operating system, where the instruction carries a to-be-displayed image and a resolution of the to-be-displayed image; and when the resolution of the to-be-displayed image is less than a first threshold, performing, by the image processing module, super-resolution processing on the to-be-displayed image, and displaying an image obtained after the super-resolution processing, to resolve a problem of a low image definition in a conventional display method.

Image processing method and device using multiple layers of an operating system

Screwdriver handle

The present application discloses a bearing structural member, a support, a joint assembly and a tube section assembly, wherein the maximum counter force provided by the bearing structural member under an external load is an own designed threshold value, namely when the external load is greater than the designed threshold value of the bearing structural member, the bearing structural member deforms, and provides a counter force equal to the designed threshold value; the support includes at least one bearing structural member; and the joint assembly and the tube section assembly are both equipped with the support. When the external load is not high and is less than the designed threshold value, the bearing structural member may effectively suppress the deformation just like a rigid structural member; when the external load exceeds the designed threshold value, the bearing structural member may deform, and provide a stable counter supporting force less than the external load. The bearing structural member may be applied to the support, the joint assembly and the tube section assembly, plays a role in protecting the structures or key structural components, and may be widely applied to the fields of design of bridges, design of building structures, design of tunnels and the like.

Bearing structural member, support, joint assembly and tube section assembly

Technologies for adaptive processing of multiple buffers is disclosed. A compute device may establish a buffer queue to which applications can submit buffers to be processed, such as by hashing the submitted buffers. The compute device monitors the buffer queue and determines an efficient way of processing the buffer queue based on the number of buffers present. The compute device may process the buffers serially with a single processor core of the compute device or may process the buffers in parallel with single-instruction, multiple data (SIMD) instructions. The compute device may determine which method to use based on a comparison of the throughput of serially processing the buffers as compared to parallel processing the buffers, which may depend on the number of buffers in the buffer queue.

Technologies for adaptive processing of multiple buffers

Systems, methods, and computer-executable instructions for approximating a softmax layer are disclosed. A small world graph that includes a plurality of nodes is constructed for a vocabulary of a natural language model. A context vector is transformed. The small world graph is searched using the transformed context vector to identify a top-K hypothesis. A distance from the context vector for each of the top-K hypothesis is determined. The distance is transformed to an original inner product space. A softmax distribution is computed for the softmax layer over the inner product space of the top-K hypothesis. The softmax layer is useful for determining a next word in a speech recognition or machine translation.

Graph representations for identifying a next word

A vehicle-mounted system and a vehicle are disclosed. The vehicle-mounted system includes a vehicle lamp system, the vehicle lamp system includes a direction changing unit or/and a light intensity adjusting unit and a vehicle lamp; the vehicle lamp is arranged on a front side of a vehicle body and configured to emit light; the direction changing unit is arranged on a light emergent side of the vehicle lamp, and configured to change a transmission direction of light that is incident on the direction changing unit and emitted by the vehicle lamp; and the light intensity adjusting unit is configured to adjust intensity of the light emitted by the vehicle lamp.

Vehicle-mounted system and vehicle

Growing community of inventors

Beijing, China

Xiaodong Liu