Patents by Inventor Guanhua Wang
Guanhua Wang has filed for patents to protect the following inventions. This listing includes patent applications that are pending as well as patents that have already been granted by the United States Patent and Trademark Office (USPTO).
-
Patent number: 11868880Abstract: An interconnect topology for communication between GPUs in a computing system is determined. A quantity of directed spanning trees are generated for transmitting data between the GPUs using the interconnect topology and packed. The directed spanning trees define the connections between GPUs that are to be utilized for the transmission and the amount of data to be transmitted on each connection. Program code is generated for implementing the data transfer defined by the directed spanning trees. When the program code is executed, the directed spanning trees are used to pipeline the transmission of chunks of data, such as model parameters used during data-parallel DNN training, between the GPUs. The program code can also determine an optimal chunk size for data to be transferred between the GPUs.Type: GrantFiled: February 14, 2019Date of Patent: January 9, 2024Assignee: MICROSOFT TECHNOLOGY LICENSING, LLCInventors: Nikhil Devanur Rangarajan, Jorgen Thelin, Amar Phanishayee, Guanhua Wang, Shivaram Venkataraman
-
Publication number: 20230410809Abstract: A method for interaction between a mobile terminal and an in-vehicle terminal, a related terminal, and a related system are provided. When a user command from an in-vehicle terminal is received, software and hardware of the in-vehicle terminal may be automatically invoked based on the user command to obtain information. The method includes: a mobile terminal receives first information sent by an in-vehicle terminal, wherein the first information indicates to execute a first function. The mobile terminal determines a first indication based on the first information, and sends the first indication to the in-vehicle terminal, wherein the first indication indicates the in-vehicle terminal to obtain second information. The mobile terminal receives the second information sent by the in-vehicle terminal. The mobile terminal executes the first function based on the second information.Type: ApplicationFiled: July 24, 2023Publication date: December 21, 2023Inventors: Qi Qi, Guanhua Wang
-
Publication number: 20230391193Abstract: This application provides a display method adaptive to a driving status of a vehicle and a related apparatus, are applied to a vehicle with a self-driving function. When the vehicle is switched to a manual driving state, a cockpit display screen in the vehicle presents a first display state, and a screen height of the cockpit display screen in the first display state is a first height. When the vehicle is switched to a self-driving state, the cockpit display screen in the vehicle presents a second display state, and a screen height of the cockpit display screen in the second display state is a second height. The second height is greater than the first height. When the vehicle is in different driving states, the cockpit display screen may present different heights, so that an entertainment requirement and a safe driving requirement that are of a user both can be met.Type: ApplicationFiled: August 22, 2023Publication date: December 7, 2023Inventors: Lei Huang, Guanhua Wang, Qi Qi, Xueyan Huang
-
Publication number: 20230236271Abstract: The following relates generally to accelerated magnetic resonance imaging (MRI) reconstruction. In some embodiments, a MRI machine learning algorithm is trained based on reference MRI data in non-Cartesian k-space. During the training, at each iteration of a plurality of iterations: (i) a non-Cartesian sampling trajectory ? may be optimized under the physical constraints, and/or (ii) an image reconstructor may be jointly iteratively optimized. Examples of the image reconstructor include a convolutional neural network (CNN) denoiser, a model-based deep learning (MoDL) image reconstructor, iterative image reconstructor, a regularizer, and an invertible neural network.Type: ApplicationFiled: January 17, 2023Publication date: July 27, 2023Inventors: Jeffrey Allen Fessler, Douglas Clair Noll, Guanhua Wang
-
Publication number: 20230204700Abstract: A computer system that performs a sparsity technique is described. During operation, the computer system accesses or obtains information associated with non-invasive measurements performed on at least an individual, historical non-invasive measurements, and a dictionary of predetermined features or basis functions associated with the historical non-invasive measurements. Note that the non-invasive measurements and the historical non-invasive measurements may include or correspond to magnetic resonance (MR) measurements. For example, the MR measurements may include magnetic resonance imaging (MRI) scans. Then, the computer system updates the dictionary of predetermined features based at least in part on the non-invasive measurements and the historical non-invasive measurements, where the updating includes performing a minimization technique with a cost function having an L2-norm term and an L0-norm term.Type: ApplicationFiled: February 22, 2023Publication date: June 29, 2023Inventors: Guanhua Wang, Matteo Alessandro Francavilla, Thomas Witzel, Jeffrey H. Kaditz
-
Patent number: 11614508Abstract: A computer system that performs a sparsity technique is described. During operation, the computer system accesses or obtains information associated with non-invasive measurements performed on at least an individual, historical non-invasive measurements, and a dictionary of predetermined features or basis functions associated with the historical non-invasive measurements. Note that the non-invasive measurements and the historical non-invasive measurements may include or correspond to magnetic resonance (MR) measurements. For example, the MR measurements may include magnetic resonance imaging (MRI) scans. Then, the computer system updates the dictionary of predetermined features based at least in part on the non-invasive measurements and the historical non-invasive measurements, where the updating includes performing a minimization technique with a cost function having an L2-norm term and an L0-norm term.Type: GrantFiled: October 25, 2021Date of Patent: March 28, 2023Assignee: Q Bio, Inc.Inventors: Guanhua Wang, Matteo Alessandro Francavilla, Thomas Witzel, Jeffrey H. Kaditz
-
Publication number: 20220212690Abstract: A vehicle-mounted device information display method is provided. The method is applied to the Internet of Vehicles or self-driving field and includes: obtaining information about lane lines of a road surface on which a first vehicle is located, where the lane lines are at least two lines on the road surface that are used to divide different lanes; and displaying, based on the information about the lane lines, virtual lane lines whose types are consistent with those of the lane lines. This method can be applied to a self-driving interface in an intelligent car, so that a driver can obtain, from the self-driving interface, types of the lane lines of the traveling road surface in this case. This enriches display content of the self-driving interface.Type: ApplicationFiled: March 24, 2022Publication date: July 7, 2022Inventors: Weixi ZHENG, Guanhua WANG, Zijie CHEN, Xueyan HUANG
-
Publication number: 20200160171Abstract: Technologies are disclosed herein for dynamically generating communication primitives for use in model parameter synchronization during data-parallel DNN training by packing directed spanning trees. An interconnect topology for communication between GPUs in a computing system is determined. A quantity of directed spanning trees are generated for transmitting data between the GPUs using the interconnect topology and packed. The directed spanning trees define the connections between GPUs that are to be utilized for the transmission and the amount of data to be transmitted on each connection. Program code is generated for implementing the data transfer defined by the directed spanning trees. When the program code is executed, the directed spanning trees are used to pipeline the transmission of chunks of data, such as model parameters used during data-parallel DNN training, between the GPUs. The program code can also determine an optimal chunk size for data to be transferred between the GPUs.Type: ApplicationFiled: February 14, 2019Publication date: May 21, 2020Inventors: Nikhil Devanur RANGARAJAN, Jorgen THELIN, Amar PHANISHAYEE, Guanhua WANG, Shivaram VENKATARAMAN
-
Patent number: 9715074Abstract: An access terminal box includes a protective cover, a base, and a base plate, where the protective cover includes a cover plate and side walls that are disposed around the cover plate, the cover plate and the side walls form an open-mouthed box structure, the base is fastened inside the open-mouthed box structure of the protective cover by using a first detachable structure, the base is fastened to the base plate by using a second detachable structure. In embodiments of the present invention, when the base is to be detached from the base plate, it is not required to first detach the protective cover; therefore, a fiber coiled on the base can be better protected, damage to the fiber can be avoided, and it is suitable for non-professional personnel to detach the access terminal box.Type: GrantFiled: May 5, 2016Date of Patent: July 25, 2017Assignee: HUAWEI TECHNOLOGIES CO., LTD.Inventors: Deyi Shi, Yeren Wang, Guanhua Wang
-
Publication number: 20160327766Abstract: An access terminal box includes a protective cover, a base, and a base plate, where the protective cover includes a cover plate and side walls that are disposed around the cover plate, the cover plate and the side walls form an open-mouthed box structure, the base is fastened inside the open-mouthed box structure of the protective cover by using a first detachable structure, the base is fastened to the base plate by using a second detachable structure. In embodiments of the present invention, when the base is to be detached from the base plate, it is not required to first detach the protective cover; therefore, a fiber coiled on the base can be better protected, damage to the fiber can be avoided, and it is suitable for non-professional personnel to detach the access terminal box.Type: ApplicationFiled: May 5, 2016Publication date: November 10, 2016Applicant: HUAWEI TECHNOLOGIES CO., LTD.Inventors: Deyi SHI, Yeren WANG, Guanhua WANG
-
Patent number: 8642395Abstract: A process for assembling a Chip-On-Lead packaged semiconductor device includes the steps of: mounting and sawing a wafer to provide individual semiconductor dies; performing a first molding operation on a lead frame; depositing epoxy on the lead frame via a screen printing process; attaching one of the singulated dies on the lead frame with the epoxy, where the die attach is done at room temperature; and curing the epoxy in an oven. Throughput improvements may be ascribed to not including a hot die attach process. An optional plasma cleaning step may be performed, which greatly improves wire bonding quality and a second molding quality. In addition, since a first molding operation is performed before the formation of epoxy to avoid the problem of the epoxy hanging in the air, the delamination risk between the epoxy and the die is avoided.Type: GrantFiled: March 19, 2010Date of Patent: February 4, 2014Assignee: Freescale Semiconductor, Inc.Inventors: Zhe Li, Qingchun He, Guanhua Wang, Zhijie Wang, Nan Xu
-
Publication number: 20100248426Abstract: A process for assembling a Chip-On-Lead packaged semiconductor device includes the steps of: mounting and sawing a wafer to provide individual semiconductor dies; performing a first molding operation on a lead frame; depositing epoxy on the lead frame via a screen printing process; attaching one of the singulated dies on the lead frame with the epoxy, where the die attach is done at room temperature; and curing the epoxy in an oven. Throughput improvements may be ascribed to not including a hot die attach process. An optional plasma cleaning step may be performed, which greatly improves wire bonding quality and a second molding quality. In addition, since a first molding operation is performed before the formation of epoxy to avoid the problem of the epoxy hanging in the air, the delamination risk between the epoxy and the die is avoided.Type: ApplicationFiled: March 19, 2010Publication date: September 30, 2010Applicant: FREESCALE SEMICONDUCTOR, INCInventors: Zhe Li, Qingchun He, Guanhua Wang, Zhijie Wang, Nan Xu