Patents by Inventor Yiwen Xu

Yiwen Xu 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).

  • Publication number: 20200242734
    Abstract: Methods and systems are disclosed using improved Convolutional Neural Networks (CNN) for image processing. In one example, an input image is down-sampled into smaller images with a smaller resolution than the input image. The down-sampled smaller images are processed by a CNN having a last layer with a reduced number of nodes than a last layer of a full CNN used to process the input image at a full resolution. A result is outputted based on the processed down-sampled smaller images by the CNN having a last layer with a reduced number of nodes. In another example, shallow CNN networks are built randomly. The randomly built shallow CNN networks are combined to imitate a trained deep neural network (DNN).
    Type: Application
    Filed: April 7, 2017
    Publication date: July 30, 2020
    Inventors: Shandong WANG, Yiwen GUO, Anbang YAO, Dongqi CAI, Libin WANG, Lin XU, Ping HU, Wenhua CHENG, Yurong CHEN
  • Publication number: 20200234411
    Abstract: Methods and systems are disclosed using camera devices for deep channel and Convolutional Neural Network (CNN) images and formats. In one example, image values are captured by a color sensor array in an image capturing device or camera. The image values provide color channel data. The captured image values by the color sensor array are input to a CNN having at least one CNN layer. The CNN provides CNN channel data for each layer. The color channel data and CNN channel data is to form a deep channel image that stored in a memory. In another example, image values are captured by sensor array. The captured image values by the sensor array are input a CNN having a first CNN layer. An output is generated at the first CNN layer using the captured image values by the color sensor array. The output of the first CNN layer is stored as a feature map of the captured image.
    Type: Application
    Filed: April 7, 2017
    Publication date: July 23, 2020
    Inventors: Lin XU, Liu YANG, Anbang YAO, dongqi CAI, Libin WANG, Ping HU, Shaodong WANG, Wenhua CHENG, Yiwen GUO, Yurong CHEN
  • Patent number: 10713148
    Abstract: The disclosure relate to testing software for operating an autonomous vehicle. For instance, a first simulation may be run using log data and the software to control a first simulated vehicle. During this, one or more characteristics of the simulated vehicle may be compared with one or more characteristics of a vehicle from the log data. The comparison may be used to determine a divergence point for starting a timer. In addition, a second simulation may be run using the log data and the software to control a second simulated vehicle. The divergence point may be used to determine a handover time to allow the software to take control of the second simulated vehicle. Whether the software is able to continue through the first simulation before the timer expires without a particular type of event occurring and/or the second simulation without the particular type of event occurring is determined.
    Type: Grant
    Filed: August 7, 2018
    Date of Patent: July 14, 2020
    Assignee: Waymo LLC
    Inventors: Carl Nygaard, Yiwen Xu, James Stout
  • Publication number: 20200050536
    Abstract: The disclosure relate to testing software for operating an autonomous vehicle. For instance, a first simulation may be run using log data and the software to control a first simulated vehicle. During this, one or more characteristics of the simulated vehicle may be compared with one or more characteristics of a vehicle from the log data. The comparison may be used to determine a divergence point for starting a timer. In addition, a second simulation may be run using the log data and the software to control a second simulated vehicle. The divergence point may be used to determine a handover time to allow the software to take control of the second simulated vehicle. Whether the software is able to continue through the first simulation before the timer expires without a particular type of event occurring and/or the second simulation without the particular type of event occurring is determined.
    Type: Application
    Filed: August 7, 2018
    Publication date: February 13, 2020
    Inventors: Carl Nygaard, Yiwen Xu, James Stout
  • Publication number: 20200026999
    Abstract: Methods and systems are disclosed for boosting deep neural networks for deep learning. In one example, in a deep neural network including a first shallow network and a second shallow network, a first training sample is processed by the first shallow network using equal weights. A loss for the first shallow network is determined based on the processed training sample using equal weights. Weights for the second shallow network are adjusted based on the determined loss for the first shallow network. A second training sample is processed by the second shallow network using the adjusted weights. In another example, in a deep neural network including a first weak network and a second weak network, a first subset of training samples is processed by the first weak network using initialized weights. A classification error for the first weak network on the first subset of training samples is determined.
    Type: Application
    Filed: April 7, 2017
    Publication date: January 23, 2020
    Inventors: Libin Wang, Yiwen Guo, Anbang Yao, Dongqi Cai, Lin Xu, Ping Hu, Shangong Wang, Wenhua Cheng, Yurong Chen
  • Publication number: 20200026965
    Abstract: Methods and systems for budgeted and simplified training of deep neural networks (DNNs) are disclosed. In one example, a trainer is to train a DNN using a plurality of training sub-images derived from a down-sampled training image. A tester is to test the trained DNN using a plurality of testing sub-images derived from a down-sampled testing image. In another example, in a recurrent deep Q-network (RDQN) having a local attention mechanism located between a convolutional neural network (CNN) and a long-short time memory (LSTM), a plurality of feature maps are generated by the CNN from an input image. Hard-attention is applied by the local attention mechanism to the generated plurality of feature maps by selecting a subset of the generated feature maps. Soft attention is applied by the local attention mechanism to the selected subset of generated feature maps by providing weights to the selected subset of generated feature maps in obtaining weighted feature maps.
    Type: Application
    Filed: April 7, 2017
    Publication date: January 23, 2020
    Inventors: Yiwen GUO, Yuqing Hou, Anbang Yao, Dongqi Cai, Lin Xu, Ping Hu, Shangong Wang, Wenhua Cheng, Yurong Chen, Libin Wag
  • Publication number: 20200026988
    Abstract: Methods and systems are disclosed using improved training and learning for deep neural networks. In one example, a deep neural network includes a plurality of layers, and each layer has a plurality of nodes. For each L layer in the plurality of layers, the nodes of each L layer are randomly connected to nodes in a L+1 layer. For each L+1 layer in the plurality of layers, the nodes of each L+1 layer are connected to nodes in a subsequent L layer in a one-to-one manner. Parameters related to the nodes of each L layer are fixed. Parameters related to the nodes of each L+1 layers are updated, and L is an integer starting with 1. In another example, a deep neural network includes an input layer, output layer, and a plurality of hidden layers. Inputs for the input layer and labels for the output layer are determined related to a first sample. Similarity between different pairs of inputs and labels between a second sample with the first sample is estimated using Gaussian regression process.
    Type: Application
    Filed: April 7, 2017
    Publication date: January 23, 2020
    Inventors: Yiwen Guo, Anbang Yao, Dongqi Cai, Libin Wang, Lin Xu, Ping Hu, Shangong Wang, Wenhua Cheng, Wenhua Cheng, Yurong Chen
  • Publication number: 20200026499
    Abstract: Described herein are hardware acceleration of random number generation for machine learning and deep learning applications. An apparatus (700) includes a uniform random number generator (URNG) circuit (710) to generate uniform random numbers and an adder circuit (750) that is coupled to the URNG circuit (710). The adder circuit hardware (750) accelerates generation of Gaussian random numbers for machine learning.
    Type: Application
    Filed: April 7, 2017
    Publication date: January 23, 2020
    Inventors: Yiwen Guo, Anbang Yao, Dongqi Cai, Libin Wang, Lin Xu, Ping Hu, Shangong Wang, Wenhua Cheng
  • Publication number: 20200027015
    Abstract: Described herein are systems and methods for providing deeply stacked automated program synthesis. In one embodiment, an apparatus to perform automated program synthesis includes a memory to store instructions for automated program synthesis and a compute cluster coupled to the memory. The compute cluster supports the instructions for performing the automated program synthesis including partitioning sketched data into partitions, training diverse sets of individual program synthesis units each having different capabilities with partitioned sketched data and for each partition applying respective transformations, and generating sketched baseline data for each individual program synthesis unit.
    Type: Application
    Filed: April 7, 2017
    Publication date: January 23, 2020
    Inventors: Angang YAO, Dongqi CAI, Libin WANG, Lin XU, Ping HU, Shandong WANG, Wenhua CHENG, Yiwen GUO, Liu YANG, Yurong CHEN, Yuqing HOU, Zhou SU
  • Publication number: 20190197407
    Abstract: An apparatus and method are described for reducing the parameter density of a deep neural network (DNN). A layer-wise pruning module to prune a specified set of parameters from each layer of a reference dense neural network model to generate a second neural network model having a relatively higher sparsity rate than the reference neural network model; a retraining module to retrain the second neural network model in accordance with a set of training data to generate a retrained second neural network model; and the retraining module to output the retrained second neural network model as a fmal neural network model if a target sparsity rate has been reached or to provide the retrained second neural network model to the layer-wise pruning model for additional pruning if the target sparsity rate has not been reached.
    Type: Application
    Filed: September 26, 2016
    Publication date: June 27, 2019
    Inventors: Anbang YAO, Yiwen GUO, Lin XU, Yan LIN, Yurong CHEN
  • Publication number: 20180344281
    Abstract: The present disclosure relates to systems and methods for configuring a medical device for a medical procedure. The systems may perform the methods to initialize a gantry angle of a medical device of a new scan. The systems may also perform the methods to obtain a pre-set gantry angle associated with the new scan. The systems may also perform the methods to determine whether the initialized gantry angle is consistent with the pre-set gantry angle. The systems may also perform the methods to adjust the initialized gantry angle of the medical device to the pre-set gantry angle in response to a determination that the initialized gantry angle is inconsistent with the pre-set gantry angle.
    Type: Application
    Filed: December 18, 2017
    Publication date: December 6, 2018
    Applicant: SHANGHAI UNITED IMAGING HEALTHCARE CO., LTD.
    Inventors: Yiwen XU, Yangyang LIN, Jiawen ZHOU, Jie YU
  • Publication number: 20180196411
    Abstract: A method of determining an energy-efficient operating point of a machine tool of a machine tool system with which identical workpieces for processing can be supplied to the machine tool sequentially in time. The machine tool has an operating point dependent machine cycle time and an operating point dependent power demand. The machine tool system has at least two machine tools and has a system cycle time, and the machine cycle time is shorter than the system cycle time. The method includes determining the energy-efficient operating point in accordance with a machine cycle time dependent characteristic energy demand function of the machine tool. The characteristic energy demand function represents a machine cycle time dependent energy demand of the machine tool over the system cycle time. A corresponding device and a machine tool system are also described.
    Type: Application
    Filed: May 23, 2016
    Publication date: July 12, 2018
    Inventors: Yiwen XU, Herman YAKARIA, Tobias KÖSLER, Thomas ACKERMANN, Johannes BAUER
  • Publication number: 20180169817
    Abstract: A method for reducing an energy demand of a machine tool (2, 3, 4) of a machine tool system (1), wherein the machine tool system (1) comprises at least a first machine tool (3), with a first machine cycle time, and a second machine tool (4), with a second machine cycle time. Identical workpieces (9) are transported sequentially in time for processing (101, 107), first to the first machine tool (3) and then to the second machine tool (4). The second machine cycle time is shorter than the first machine cycle time. The method according to the invention is characterized in that the workpieces (9), after being processed by the first machine tool (3), are collected (105, 106) before they are conveyed (107) to the second machine tool (4) for processing. The invention also concerns a related device (10) and a machine tool system (1).
    Type: Application
    Filed: May 24, 2016
    Publication date: June 21, 2018
    Inventors: Yiwen XU, Herman YAKARIA, Tobias KÖSLER, Thomas ACKERMANN, Johannes BAUER
  • Publication number: 20180133917
    Abstract: A method for controlling a cutting machining process on a machine tool by a P-controller that changes a controlled variable u(t) affecting the cutting machining process based on a control deviation e(t) between a control quantity y(t) and a guide quantity w(t). To improve the control, the control factor (K) of the P-controller is variable and determined depending on instantaneous value of the control quantity y(t) via load characteristic fields. Each load characteristic field specifies a predetermined control factor for a defined value or value range of the control quantity y(t). Further disclosed is a control device for a cutting machine tool, a cutting machine tool, and a process for the cutting machining of a workpiece.
    Type: Application
    Filed: April 26, 2016
    Publication date: May 17, 2018
    Inventors: Yiwen XU, Herman YAKARIA, Tobias KÖSLER, Thomas ACKERMANN, Falko FAHNAUER
  • Patent number: 9404989
    Abstract: A method for enhancing inspection of components of specific geometry based on Barkhausen noises. The method includes specifying a first calibration curve that is independent of the component geometry, which describes the relationship between surface hardness values and measured Barkhausen noise signals. A first noise signal is determined by the measuring device for a reference component having the specified geometry and a first hardness value. A second noise signal is determined for a second reference component, having the specified geometry and a second hardness value lower than the first. A second calibration curve is determined, in which the first calibration curve is fitted to the first noise signal at the first hardness value and to the second noise signal at the second hardness value, such that using the second calibration curve, the measured noise signal of a component having the specified geometry relates with a surface hardness value.
    Type: Grant
    Filed: March 4, 2013
    Date of Patent: August 2, 2016
    Assignee: ZF Friedrichshafen AG
    Inventors: Oliver Bleicher, Herman Yakaria, Yiwen Xu
  • Patent number: 9383339
    Abstract: A method of inspecting a component (1) on the basis of Barkhausen noises in which a plurality of Barkhausen noise signals are processed, which have been or are determined at measurement positions (PS1, PS2, . . . , PS9) along the surface of the component (1) by a measuring device. According to the method, a computer forms a measurement matrix (M) from the Barkhausen noise signals, which matrix contains the Barkhausen noise signals detected as entries. A variety of characteristics are specified, each of which represents at least one cause of a manufacturing defect(s) of the component (1), each characteristic is associated with a processing procedure of the measurement matrix (M). The procedure is specific for the characteristic concerned. Finally, for each characteristic the measurement matrix (M) undergoes the associated processing procedure in which the intensity of the characteristic concerned is determined.
    Type: Grant
    Filed: March 4, 2013
    Date of Patent: July 5, 2016
    Assignee: ZF Friedrichshafen AG
    Inventors: Oliver Bleicher, Herman Yakaria, Yiwen Xu
  • Publication number: 20150061647
    Abstract: A method for enhancing inspection of components of specific geometry based on Barkhausen noises. The method includes specifying a first calibration curve that is independent of the component geometry, which describes the relationship between surface hardness values and measured Barkhausen noise signals. A first noise signal is determined by the measuring device for a reference component having the specified geometry and a first hardness value. A second noise signal is determined for a second reference component, having the specified geometry and a second hardness value lower than the first. A second calibration curve is determined, in which the first calibration curve is fitted to the first noise signal at the first hardness value and to the second noise signal at the second hardness value, such that using the second calibration curve, the measured noise signal of a component having the specified geometry relates with a surface hardness value.
    Type: Application
    Filed: March 4, 2013
    Publication date: March 5, 2015
    Applicant: ZF Friedrichshafen AG
    Inventors: Oliver Bleicher, Herman Yakaria, Yiwen Xu
  • Publication number: 20150054501
    Abstract: A method of inspecting a component (1) on the basis of Barkhausen noises in which a plurality of Barkhausen noise signals are processed, which have been or are determined at measurement positions (PS1, PS2, . . . , PS9) along the surface of the component (1) by a measuring device. According to the method, a computer forms a measurement matrix (M) from the Barkhausen noise signals, which matrix contains the Barkhausen noise signals detected as entries. A variety of characteristics are specified, each of which represents at least one cause of a manufacturing defect(s) of the component (1), each characteristic is associated with a processing procedure of the measurement matrix (M). The procedure is specific for the characteristic concerned. Finally, for each characteristic the measurement matrix (M) undergoes the associated processing procedure in which the intensity of the characteristic concerned is determined.
    Type: Application
    Filed: March 4, 2013
    Publication date: February 26, 2015
    Inventors: Oliver Bleicher, Herman Yakaria, Yiwen Xu
  • Publication number: 20130159324
    Abstract: A search query is received that is associated with two or more data sources so that a source-specific query is generated for each data source using a data source definition associated with the corresponding data source and the search query. Thereafter, searches are executed for a first set of data sources using the corresponding source-specific queries. These search results from the first set of data sources are consolidated. In addition, searches are executed for a second set of data sources using the corresponding source-specific queries. These search results from the second set of data sources are then consolidated with the consolidated search results from the first set of data sources. In some implementations, at least one search comprises a main search followed by a sub-search that filters results from the main search. Related apparatus, systems, techniques and articles are also described.
    Type: Application
    Filed: December 14, 2011
    Publication date: June 20, 2013
    Inventors: Yiwen Xu, Mu Shen, Evelyna Holban, Sebastien Phan, Xiaohua Xian, Ming Hao Xie, Bernd Reimann
  • Patent number: 7925515
    Abstract: A real-time directory access system provides the public with information of interest of an organization, and enabling mobile personal communication devices to access the organization's information of interest and to interactively communicate with the organization. This system includes an organization's central data server, a plurality of branch data servers for storing branch information of interest and for interactively communicating with mobile devices, a plurality of in-branch access points for distributing information of interest. The system also includes a plurality of mobile personal communication devices for receiving and displaying information of interest from access points and interactively communicating with branch data servers.
    Type: Grant
    Filed: October 22, 2007
    Date of Patent: April 12, 2011
    Assignee: Wenshine Technology Ltd.
    Inventors: Xinyi Xu, Yiwen Xu