-
公开(公告)号：US20210383233A1
公开(公告)日：2021-12-09
申请号：US17101748
申请日：2020-11-23
Inventor： Weiyue SU , Shikun FENG , Zhifan ZHU , Weibin LI , Jingzhou HE , Shiwei HUANG
Abstract: The disclosure discloses a method for distilling a model, an electronic device, and a storage medium, and relates to the field of deep learning technologies. A teacher model and a student model are obtained. The second intermediate fully connected layer is transformed into an enlarged fully connected layer and a reduced fully connected layer based on a first data processing capacity of a first intermediate fully connected layer of the teacher model and a second data processing capacity of a second intermediate fully connected layer of the student model. The second intermediate fully connected layer is replaced with the enlarged fully connected layer and the reduced fully connected layer to generate a training student model. The training student model is distilled based on the teacher model.
-
2.发明申请公开(公告)号：US20210374174A1
公开(公告)日：2021-12-02
申请号：US17128047
申请日：2020-12-19
Inventor： Chaoxing CHEN , Zhengjie HUANG , Shikun FENG
IPC: G06F16/438 , G06F16/901 , G06F17/18 , G06F17/16
Abstract: Embodiments of the disclosure disclose a method and apparatus for recommending multimedia resources, an electronic device and a storage medium. The present disclosure relates to a nature language processing technology. The technology solution can be implemented by generating a relation graph based on first multimedia resources browsed by a plurality of user objects, taking two nodes between which a number of nodes in a path of the relation graph is less than a number threshold as a training sample, training a graph model to obtain a representative vector of each node in the training samples and recommending multimedia resources based on vector similarities between the representative vectors of respective nodes of the plurality of user objects and/or the representative vectors of respective nodes of the first multimedia resources.
-
公开(公告)号：US20180349350A1
公开(公告)日：2018-12-06
申请号：US15921386
申请日：2018-03-14
Inventor： Zhifan ZHU , Shikun FENG , Kunsheng ZHOU , Jingzhou HE
Abstract: This disclosure discloses an artificial intelligence based method and apparatus for checking a text. An embodiment of the method comprises: lexing a first to-be-checked text and a second to-be-checked text respectively, determining word vectors of the lexed words to generate a first word vector sequence and a second word vector sequence; inputting the first word vector sequence and the second word vector sequence respectively into a pre-trained convolutional neural network containing at least one multi-scale convolutional layer, identifying vector sequences in a plurality of vector sequences outputted by a last multi-scale convolutional layer as eigenvector sequences, to obtain eigenvector sequence groups respectively corresponding to the texts; combining eigenvector sequences in each eigenvector sequence group to generate a combined eigenvector sequence; and analyzing the generated combined eigenvector sequences to determine whether the first text and the second text pass a similarity check. The embodiment improves the flexibility in checking a text.
-
公开(公告)号：US20210390393A1
公开(公告)日：2021-12-16
申请号：US17128978
申请日：2020-12-21
Inventor： Zhengjie HUANG , Weibin LI , Shikun FENG
Abstract: A method for pre-training a graph neural network, an electronic device and a readable storage medium, which relate to the technical field of deep learning are proposed. An embodiment for pre-training a graph neural network includes: acquiring an original sample to be used for training; expanding the original sample to obtain a positive sample and a negative sample corresponding to the \original sample; constructing a sample set Corresponding to the original sample by using the original sample and the positive sample, the negative sample, and a weak sample corresponding to the original sample; and pre-training the graph neural network by taking the original sample and one of other samples in the sample set as input of the graph neural network respectively, until the graph neural network converges. The technical solution may implement pre-training of a graph neural network at a graph level.
-
公开(公告)号：US20210182498A1
公开(公告)日：2021-06-17
申请号：US16885358
申请日：2020-05-28
Inventor： Yu SUN , Haifeng WANG , Shuohuan WANG , Yukun LI , Shikun FENG , Hao TIAN , Hua WU
Abstract: The present disclosure provides a method, apparatus, electronic device and storage medium for processing a semantic representation model, and relates to the field of artificial intelligence technologies. A specific implementation solution is: collecting a training corpus set including a plurality of training corpuses; training the semantic representation model using the training corpus set based on at least one of lexicon, grammar and semantics. In the present disclosure, by building the unsupervised or weakly-supervised training task at three different levels, namely, lexicon, grammar and semantics, the semantic representation model is enabled to learn knowledge at levels of lexicon, grammar and semantics from massive data, enhance the capability of universal semantic representation and improve the processing effect of the NLP task.
-
Patent Agency Ranking
公开(公告)号：US20210397947A1
公开(公告)日：2021-12-23
申请号：US17116291
申请日：2020-12-09
Inventor： Weibin LI , Zhifan ZHU , Shikun FENG , Jingzhou HE , Shiwei HUANG
Abstract: Embodiments of the present disclosure provide a method for generating a model for representing heterogeneous graph node. A specific implementation includes: acquiring a training data set, wherein the training data set includes node walk path information obtained by sampling a heterogeneous graph according to different meta paths; and training, based on a gradient descent algorithm, an initial heterogeneous graph node representation model with the training data set as an input of the initial heterogeneous graph node representation model, to obtain a heterogeneous graph node representation model.
-
公开(公告)号：US20190163737A1
公开(公告)日：2019-05-30
申请号：US16306488
申请日：2016-12-22
Inventor： Kunsheng ZHOU , Jingzhou HE , Lei SHI , Shikun FENG
Abstract: Disclosed are a method and an apparatus for constructing a binary feature dictionary. The method may include: extracting binary features from a corpus; calculating a preset statistic of each binary feature; and selecting a preset number of binary features in sequence according to the preset statistic to constitute the binary feature dictionary.
-
公开(公告)号：US20210201198A1
公开(公告)日：2021-07-01
申请号：US16945183
申请日：2020-07-31
Inventor： Weibin LI , Zhifan ZHU , Weiyue SU , Jingzhou HE , Shikun FENG , Yuhui CAO , Xuyi CHEN , Danxiang ZHU
IPC: G06N20/00 , G06F16/901
Abstract: A method for generating node representations in a heterogeneous graph, an electronic device, and a non-transitory computer-readable storage medium, and relates to the field of machine learning technologies. The method includes: acquiring a heterogeneous graph; inputting the heterogeneous graph into a heterogeneous graph learning model to generate a node representation of each node in the heterogeneous graph, in which the heterogeneous graph learning model generates the node representation of each node by actions of: segmenting the heterogeneous graph into a plurality of subgraphs, in which each subgraph includes nodes of two types and an edge of one type between the nodes of two types; and generating the node representation of each node according to the plurality of subgraphs.
-
公开(公告)号：US20210200806A1
公开(公告)日：2021-07-01
申请号：US16895996
申请日：2020-06-08
Inventor： Zhengjie HUANG , Chaoxing CHEN , Shikun FENG , Hua LU
IPC: G06F16/901 , G06F16/903 , G06N3/08 , G06F9/38
Abstract: Embodiments of the present disclosure relate to a method and apparatus for parallel processing of information. The method may include: detecting whether a source node in a graph structure processed by a graph neural network sends information to a target node; generating and recording, in response to detecting that the source node sends the information to the target node, a target information sequence and an index for the target information sequence, target information being determined based on a characteristic of the source node, a characteristic of the target node, and the sent information, and the index being used to group the target information sequence; using the index to group the target information sequence; and processing groups of target information in parallel.
-
公开(公告)号：US20210192288A1
公开(公告)日：2021-06-24
申请号：US16895242
申请日：2020-06-08
Inventor： Yuhui CAO , Shikun FENG , Xuyi CHEN , Jingzhou HE
Abstract: Embodiments of the present disclosure provide a method and apparatus for processing data. The method may include: acquiring a sample set; inputting a plurality of target samples in the sample set into a pre-trained first natural language processing model, respectively, to obtain prediction results output from the pre-trained first natural language processing model; determining the obtained prediction results as labels of the target samples in the plurality of target samples, respectively; and training a to-be-trained second natural language processing model, based on the plurality of target samples and the labels of the target samples to obtain a trained second natural language processing model, parameters in the first natural language processing model being more than parameters in the second natural language processing model.
-
-
-
-
-
-
-
-
-
Search Results
10
records
(took 0.02 seconds)
