公开(公告)号：US12229667B2

公开(公告)日：2025-02-18

申请号：US17209576

申请日：2021-03-23

Applicant: BEIJING BAIDU NETCOM SCIENCE AND TECHNOLOGY CO., LTD.

Inventor： Daxiang Dong ,  Wenhui Zhang ,  Zhihua Wu ,  Dianhai Yu ,  Yanjun Ma ,  Haifeng Wang

IPC: G06N3/08 ,  G06F9/50 ,  G06F18/214

Abstract: A method and an apparatus for generating a shared encoder are provided, which belongs to a field of computer technology and deep learning. The method includes: sending by a master node a shared encoder training instruction to child nodes, so that each child node obtains training samples based on a type of a target shared encoder included in the training instruction; sending an initial parameter set of the target shared encoder to be trained to each child node after obtaining a confirmation message returned by each child node; obtaining an updated parameter set of the target shared encoder returned by each child node; determining a target parameter set corresponding to the target shared encoder based on a first preset rule and the updated parameter set of the target shared encoder returned by each child node.

公开(公告)号：US20210216875A1

公开(公告)日：2021-07-15

申请号：US17218033

申请日：2021-03-30

Applicant: Beijing Baidu Netcom Science and Technology Co., Ltd.

Inventor： Tianjian He ,  Yi Liu ,  Daxiang Dong ,  Yanjun Ma ,  Dianhai Yu

IPC: G06N3/08 ,  G06N3/10

Abstract: A method for training a deep learning model may include: acquiring model description information and configuration information of a deep learning model; segmenting the model description information into at least two sections based on segmentation point variable in the configuration information, and loading the model description information to a corresponding resource to run; inputting a batch of training samples into a resource corresponding to a first section of model description information, then starting training and using obtained context information as an input of a resource corresponding to a subsequent section of model description information; and so on until an operation result of a resource corresponding to a final section of model description information is obtained; if a training completion condition is met, outputting a trained deep learning model; and otherwise, keeping on acquiring a subsequent batch of training samples and performing the above training steps until the condition is met.

公开(公告)号：US10762305B2

公开(公告)日：2020-09-01

申请号：US16005779

申请日：2018-06-12

Applicant: BEIJING BAIDU NETCOM SCIENCE AND TECHNOLOGY CO., LTD.

Inventor： Yi Liu ,  Daxiang Dong ,  Dianhai Yu

IPC: G06F16/35 ,  G06F40/56 ,  H04L12/58 ,  G06F17/18 ,  G06N3/04 ,  G06N3/08 ,  G06N3/00 ,  G06F40/35 ,  G06F40/49 ,  G06F40/279 ,  G06N5/04 ,  G06F16/583

Abstract: Embodiments of the present disclosure relate to a method for generating chatting data based on AI, a computer device and a computer-readable storage medium. The method includes: converting chatting data inputted by a user into an input word sequence; converting a tag of the user into a tag word sequence; based on a preset encoding-decoding model with an attention model, predicting according to the input word sequence and the tag word sequence to obtain a target word sequence; and converting the target word sequence into reply data of the chatting data.

公开(公告)号：US11954522B2

公开(公告)日：2024-04-09

申请号：US17076346

申请日：2020-10-21

Applicant: BEIJING BAIDU NETCOM SCIENCE AND TECHNOLOGY CO., LTD.

Inventor： Daxiang Dong ,  Haifeng Wang ,  Dianhai Yu ,  Yanjun Ma

IPC: G06F9/46 ,  G06F9/48 ,  G06F9/52 ,  G06N20/00

CPC classification number: G06F9/4881 ,  G06F9/52 ,  G06N20/00

Abstract: Embodiments of the present disclosure disclose a method for processing tasks in parallel, a device and a storage medium, and relate to a field of artificial intelligent technologies. The method includes: determining at least one parallel computing graph of a target task; determining a parallel computing graph and an operator scheduling scheme based on a hardware execution cost of each operator task of each of the at least one parallel computing graph in a cluster, in which the cluster includes a plurality of nodes for executing the plurality of operator tasks, and each parallel computing graph corresponds to at least one operator scheduling scheme; and scheduling and executing the plurality of operator tasks of the determined parallel computing graph in the cluster based on the determined parallel computing graph and the determined operator scheduling scheme.

公开(公告)号：US11651002B2

公开(公告)日：2023-05-16

申请号：US16193454

申请日：2018-11-16

Applicant: BEIJING BAIDU NETCOM SCIENCE AND TECHNOLOGY CO., LTD.

Inventor： Daxiang Dong ,  Jun Zhang ,  Dianhai Yu

IPC: G06F16/248 ,  G06F16/33 ,  G06F16/332 ,  G06F16/2457 ,  G06F7/16 ,  G06N3/04 ,  G06N3/08

CPC classification number: G06F16/248 ,  G06F7/16 ,  G06F16/24578 ,  G06F16/3326 ,  G06F16/3334 ,  G06N3/04 ,  G06N3/08

Abstract: A method for providing an intelligent service, an intelligent service system and an intelligent terminal based on artificial intelligence. The method comprises: receiving a first service request from a user (102); determining a search term and the weight thereof for the first service request (104); providing a first service result according to the search term and the weight thereof (106); and collecting feedback information for the first service result from the user, and adjusting, in real time, the search term and/or the weight thereof for the first service request, according to evaluation information in the feedback information (108).

公开(公告)号：US11238097B2

公开(公告)日：2022-02-01

申请号：US16000160

申请日：2018-06-05

Applicant: BEIJING BAIDU NETCOM SCIENCE AND TECHNOLOGY CO., LTD.

Inventor： Zhiliang Tian ,  Daxiang Dong ,  Dianhai Yu

IPC: G06N3/00 ,  G06F16/901 ,  G06N5/02 ,  G06N3/04 ,  G06F40/30

Abstract: A method and apparatus for recalling news based on artificial intelligence, a device and a storage medium. The method comprises: building an index repository according to candidate news, the index repository including M search trees, each search tree being a complete binary tree including at least two layers, each non-leaf node in each search tree corresponding to a semantic index vector, each piece of candidate news corresponding to a leaf node in each search tree; when news needs to be recommended to the user, generating a user's semantic index vector according to the user's interest tag; with respect to each search tree, respectively according to semantic index vectors corresponding to non-leaf nodes therein and the user's semantic index vector, determining a path from a first layer of non-leaf nodes to a leaf node, and regarding candidate news corresponding to the leaf node on the path as a recall result.