-
61.发明授权公开(公告)号：US11526774B2
公开(公告)日：2022-12-13
申请号：US17564071
申请日：2021-12-28
Applicant: ZHEJIANG LAB
Inventor： Hongsheng Wang , Haijun Shan , Jiaqing Fu
Abstract: Disclosed is a method for automatically compressing multi-task oriented pre-trained language model and a platform thereof. According to the method, a meta-network of a structure generator is designed, a knowledge distillation coding vector is constructed based on a knowledge distillation method of Transformer layer sampling, and a distillation structure model corresponding to a currently input coding vector is generated by using the structure generator; at the same time, a Bernoulli distribution sampling method is provided for training the structure generator; in each iteration, each encoder unit is transferred by Bernoulli distribution sampling to form a corresponding coding vector; by changing the coding vector input to the structure generator and a small batch of training data, the structure generator and the corresponding distillation structure are jointly trained, and a structure generator capable of generating weights for different distillation structures can be acquired.
-
公开(公告)号：US20220383565A1
公开(公告)日：2022-12-01
申请号：US17616161
申请日：2021-01-23
Applicant: ZHEJIANG LAB , MINFOUND MEDICAL SYSTEMS CO., LTD
Inventor： Wentao ZHU , Bao YANG , Long ZHOU , Hongwei YE , Ling CHEN , Fan RAO , Yaofa WANG
IPC: G06T11/00
Abstract: Disclosed is a three-dimensional low-dose PET reconstruction method based on deep learning. The method comprises the following steps: back projecting low-dose PET raw data to the image domain to maintain enough information from the raw data; selecting an appropriate three-dimensional deep neural network structure to fit the mapping between the back projection of the low-dose PET and a standard-dose PET image; after learning from the training samples the network parameters are fixed, realizing three-dimensional PET image reconstruction starting from low-dose PET raw data, thereby obtaining a low-dose PET reconstructed image which has a lower noise and a higher resolution compared with the traditional reconstruction algorithm and image domain noise reduction processing.
-
公开(公告)号：US20220344070A1
公开(公告)日：2022-10-27
申请号：US17695865
申请日：2022-03-16
Applicant: ZHEJIANG LAB , ZHEJIANG UNIVERSITY
Inventor： Cuihong LI , Zhiming CHEN , Yuanyuan MA , Xiaowen GAO , Zhenhai FU , Huizhu HU
Abstract: Disclosed is a multi-dimensional optical tweezers calibration device based on electric field quantity calibration and a method thereof. The polarization-dependent characteristics of a tightly focused optical trap are utilized to realize triaxial electric field force calibration of particles through a one-dimensional electric field quantity calibration device. The method of the present application enables a particle electric field force calibration system to be compatible with particle delivery and particle detection systems; the device is simplified and calibration complexity is reduced.
-
公开(公告)号：US20220265218A1
公开(公告)日：2022-08-25
申请号：US17571523
申请日：2022-01-09
Applicant: Zhejiang Lab
Inventor： Taihao LI , Guanxiong PEI , Yulong LIU
Abstract: The present invention discloses a real-time evaluation method and evaluation system for group emotion homogeneity. The method comprises the steps as follows: enabling testees to be in the same emotion induction environment, and collecting the original electroencephalograph (EEG) signals of multiple persons at the same time through online multichannel EEG equipment; and based on the average instantaneous phase per second of the beta frequency band and the energy value per second of the alpha frequency band obtained after wavelet transformation, calculating the time synchronization degree and the valence consistency degree in real time, and finally obtaining a group emotion homogeneity index for the objective evaluation of group emotion homogeneity.
-
公开(公告)号：US20220138414A1
公开(公告)日：2022-05-05
申请号：US17531813
申请日：2021-11-22
Applicant: ZHEJIANG LAB
Inventor： Hongsheng Wang , Haijun Shan , Shengjian Hu
Abstract: Disclosed is a meta-knowledge fine tuning method and platform for a multi-task language model. The method is to obtain highly transferable shared knowledge, that is, meta-knowledge, on different data sets of tasks of the same category, perform interrelation and mutual reinforcement on the learning processes of the tasks of the same category that correspond to different data sets and are in different domains, so as to improve the fine tuning effect of downstream tasks of the same category on data sets of different domains in the application of the language model, and improve the parameter initialization ability and the generalization ability of a general language model for the tasks of the same category.
-
Patent Agency Ranking
66.发明申请公开(公告)号：US20220093258A1
公开(公告)日：2022-03-24
申请号：US17543738
申请日：2021-12-07
Applicant: ZHEJIANG LAB
Inventor： Jingsong LI , Yu TIAN , Weiguo CHEN , Jing MA
Abstract: Provided is a multi-center synergetic cancer prognosis prediction system based on multi-source migration learning. The system includes a model parameter setting module, a data screening module, and a multi-source migration learning module, wherein the model parameter setting module is responsible for setting cancer prognosis prediction model parameters; the data screening module is arranged at a clinical center, and a management center transmits the set model parameter to each clinical center, such that each clinical center inquires a sample feature and prognosis index data from a local database according to the model parameter, so as to preprocess the data; and the multi-source migration learning module includes a source model training unit, a migration weight calculation unit, and a target model calculation unit.
-
公开(公告)号：US20250086022A1
公开(公告)日：2025-03-13
申请号：US18569295
申请日：2023-10-11
Applicant: ZHEJIANG LAB
Inventor： Yong LI , Laiping ZHAO , Jie LI , Wen CHENG , Guang CHEN , Lingfang ZENG
Abstract: A method for data processing is provided, and includes: obtaining each piece of to-be-processed data, determining whether a set amount of the to-be-processed data is capable to be processed under a current processing process by a data processing model, if not, obtaining data processing periods of the data processing model under multiple configuration combinations; for a data processing period of each of the multiple configuration combinations, determining an amount of data that is capable to be processed by the data processing model within the data processing period, as a target data amount; by taking the data processing model to be capable to process the set amount of the to-be-processed data as a target, according to the target data amount for a data processing period of each of the multiple configuration combinations, selecting a target configuration combination from the multiple configuration combinations.
-
公开(公告)号：US12223650B2
公开(公告)日：2025-02-11
申请号：US18796239
申请日：2024-08-06
Applicant: ZHEJIANG LAB
Inventor： Yu Zhang , Chaoliang Sun , Zhichao Wang , Huan Zhang , Haotian Qian , Tianzi Jiang
Abstract: A system for predicting disease with graph convolutional neural network based on multimodal magnetic resonance imaging, which extracts the radiomics information of multiple brain regions across modals as the features of nodes from multimodal magnetic resonance data, and extracts the connectomics information between brain regions to form an adjacency matrix. T1-weighted structural images extract cortical information through cortical reconstruction, and resting-state magnetic resonance data are used to calculate amplitude of low frequency fluctuations, regional homogeneity and functional connectivity. Through multimodal data preprocessing, image index extraction and structured data integration, multimodal unstructured magnetic resonance image data are integrated into unified graph-structured data, and the disease is predicted by a graph convolutional neural network method. The system can better integrate the cross-modal physiological indexes of multiple brain regions and the correlation between brain regions and improve prediction ability of the model and generalization ability of the model with different diseases.
-
公开(公告)号：US20250030509A1
公开(公告)日：2025-01-23
申请号：US18549553
申请日：2023-05-25
Applicant: ZHEJIANG LAB , ZHEJIANG UNIVERSITY
Inventor： Huizhu HU , Zhenhai FU , Xiaowen GAO , Xingfan CHEN , Nan LI , Cheng LIU , Zhiming CHEN , Jinsheng XU , Shaochong ZHU , Yingying WANG , Peitong HE
Abstract: The present disclosure provides a method and device for performing signal communication based on a levitated particle. In one example, the method includes: preparing a levitated state of the particle; regulating and measuring a net charge quantity carried by the levitated particle; calibrating electromagnetic response characteristics of the levitated particle; applying an electromagnetic communication signal; obtaining and demodulating the electromagnetic communication signal. In an example, the device includes: a levitation trapper; a charge measure-regulator; an electromagnetic response calibrator, configured to obtain, in advance, a background noise and an electromagnetic response transfer function of the levitated particle; a communication signal detect-demodulator, configured to detect a motion response of the levitated particle under an electromagnetic communication signal; based on the background noise and the electromagnetic response transfer function of the levitated particle, recover the applied electromagnetic communication signal from the detected motion response, and demodulate symbols of the electromagnetic communication signal.
-
公开(公告)号：US20250005109A1
公开(公告)日：2025-01-02
申请号：US18689934
申请日：2023-06-30
Applicant: ZHEJIANG LAB
Inventor： Hongyang CHEN , Rong CAO
IPC: G06F18/25 , G06N3/042 , G06N3/045 , G06N3/0464
Abstract: Predicting parking space vacancy rate methods and apparatuses, storage media and devices, acquiring parking space vacancy rates of each of parking lots in the area to be predicted at a plurality of moments before the moment to be predicted as historical vacancy rates of each of the parking lots; obtaining a first feature by inputting the historical vacancy rates of each of the parking lots into the feature extraction network, wherein the first feature is used to characterize a relationship between the historical vacancy rates of each of the parking lots and time; obtaining a fusion feature by inputting the spatial relationship diagram and the first feature into the graph fusion network; and obtaining a parking space vacancy rate of each of the parking lots in the area to be predicted at the moment to be predicted by inputting the fusion feature into the result prediction network.
-
-
-
-
-
-
-
-
-
Search Results
332
records
(took 0.02 seconds)
