Patents by Inventor Song Ge
Song Ge 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: 20240076553Abstract: The present invention provides nonionic cleavable surfactants (NCS), specifically n-Decyl-disulfide-?-D-maltoside (DSSM), suitable for MS-based proteomics and analysis. These surfactants are designed to mimic the properties of a commonly used surfactant in structural biology, n-dodecyl-?-d-maltoside (DDM), but contain a disulfide bond that allows for facile cleavage and surfactant removal before analysis. DSSM and other NCS are compatible with native mass spectrometry, top-down and bottom-up proteomics, ESI-MS and other analytical techniques, and reduce signal suppression typically observed with other surfactants. DSSM and other NCS provide versatile surfactants that can facilitate protein sample preparation under non-denaturing conditions for a myriad of proteomic and structural biology applications and act as a general replacement for DDM.Type: ApplicationFiled: August 17, 2023Publication date: March 7, 2024Inventors: Kyle BROWN, Song JIN, Ying GE, Morgan GUGGER
-
Publication number: 20230368068Abstract: The present disclosure relates to systems, methods, and computer-readable media for training and implementing pipeline error detection models to facilitate automated detection of data quality (DQ) issues within recurring data pipelines. For example, systems described herein involve training a pipeline error detection model by first constructing a plurality of DQ constraints for a recurring data pipeline based on ranges of values observed over a history of pipeline executions. The systems may further train the model to predict DQ issues by synthetically applying data variants to historical executions of the recurring data pipeline or to data pipelines having similar characteristics thereto. Once trained, the pipeline error detection model(s) can be applied to new executions of the data pipeline as they become available to quickly and efficiently predict whether a given execution includes a predicted DQ issue therein.Type: ApplicationFiled: May 12, 2022Publication date: November 16, 2023Inventors: Yeye HE, Weiwei CUI, Song GE, Haidong ZHANG, Shi HAN, Dongmei ZHANG, Surajit CHAUDHURI
-
Publication number: 20230334891Abstract: The disclosure provides a method, a device and a computer readable medium for determining the penguin population relating to the technical field of image processing. The method of one embodiment of the disclosure comprises the following steps: firstly acquiring the penguin habitat image detected by the unmanned airborne sensor; and acquiring an interest region with penguins distributing, and a gray value of the interest region from the penguin habitat image; then for any one of gray value: determining a penguin state according to the gray value; finally determining the penguin population according to the penguin state.Type: ApplicationFiled: April 18, 2023Publication date: October 19, 2023Inventors: PENG ZHAO, HAO LIU, YUFEI DENG, PENGJIA LIU, JINMENG YAN, ZHENHUA CUI, SONG GE
-
Patent number: 11699093Abstract: Techniques for generating and executing an execution plan for a machine learning (ML) model using one of an edge device and a non-edge device are described. In some examples, a request for the generation of the execution plan includes at least one objective for the execution of the ML model and the execution plan is generated based at least in part on comparative execution information and network latency information.Type: GrantFiled: January 16, 2018Date of Patent: July 11, 2023Assignee: Amazon Technologies, Inc.Inventors: Nagajyothi Nookula, Poorna Chand Srinivas Perumalla, Aashish Jindia, Danjuan Ye, Eduardo Manuel Calleja, Song Ge, Vinay Hanumaiah, Wanqiang Chen, Safeer Mohiuddin, Romi Boimer, Madan Mohan Rao Jampani, Fei Chen
-
Publication number: 20220224801Abstract: According to implementations of the subject matter described herein, there is provided a solution for data embedding and data extraction in images. To perform data embedding, a target region for data embedding is determined from a source image, such as a chart image. Target data to be embedded is converted into a sequence of logical values represented in a predetermined format. Based on the sequence of logical values, image values of one or more image elements in the target region are changed such that the changed image values in the target region can be used to convey the sequence of logical values corresponding to the target data. The variations in image values are within a predetermined range such that no significant data distortion of perception distortion is caused by the data embedding. In a subsequent process, the embedded data can also be easily extracted from the image for use.Type: ApplicationFiled: May 3, 2020Publication date: July 14, 2022Inventors: Bin Zhu, Haidong Zhang, Yuanyuan Tang, He Huang, Song Ge
-
Patent number: 11372869Abstract: A system for frequent pattern mining uses two layers of processing: a plurality of computing nodes, and a plurality of processors within each computing node. Within each computing node, the data set against which the frequent pattern mining is to be performed is stored in shared memory, accessible concurrently by each of the processors. The search space is partitioned among the computing nodes, and sub-partitioned among the processors of each computing node. If a processor completes its sub-partition, it requests another sub-partition. The partitioning and sub-partitioning may be performed dynamically, and adjusted in real time.Type: GrantFiled: June 1, 2018Date of Patent: June 28, 2022Assignee: Microsoft Technology Licensing, LLCInventors: Shi Han, Yingnong Dang, Dongmei Zhang, Song Ge
-
Publication number: 20190220783Abstract: Techniques for generating and executing an execution plan for a machine learning (ML) model using one of an edge device and a non-edge device are described. In some examples, a request for the generation of the execution plan includes at least one objective for the execution of the ML model and the execution plan is generated based at least in part on comparative execution information and network latency information.Type: ApplicationFiled: January 16, 2018Publication date: July 18, 2019Inventors: Nagajyothi NOOKULA, Poorna Chand Srinivas PERUMALLA, Aashish JINDIA, Danjuan YE, Eduardo Manuel CALLEJA, Song GE, Vinay HANUMAIAH, Wanqiang CHEN, Safeer MOHIUDDIN, Romi BOIMER, Madan Mohan Rao JAMPANI, Fei CHEN
-
Publication number: 20180307732Abstract: A system for frequent pattern mining uses two layers of processing: a plurality of computing nodes, and a plurality of processors within each computing node. Within each computing node, the data set against which the frequent pattern mining is to be performed is stored in shared memory, accessible concurrently by each of the processors. The search space is partitioned among the computing nodes, and sub-partitioned among the processors of each computing node. If a processor completes its sub-partition, it requests another sub-partition. The partitioning and sub-partitioning may be performed dynamically, and adjusted in real time.Type: ApplicationFiled: June 1, 2018Publication date: October 25, 2018Inventors: Shi Han, Yingnong Dang, Dongmei Zhang, Song Ge
-
Patent number: 10013465Abstract: A system for frequent pattern mining uses two layers of processing: a plurality of computing nodes, and a plurality of processors within each computing node. Within each computing node, the data set against which the frequent pattern mining is to be performed is stored in shared memory, accessible concurrently by each of the processors. The search space is partitioned among the computing nodes, and sub-partitioned among the processors of each computing node. If a processor completes its sub-partition, it requests another sub-partition. The partitioning and sub-partitioning may be performed dynamically, and adjusted in real time.Type: GrantFiled: April 27, 2016Date of Patent: July 3, 2018Assignee: Microsoft Technology Licensing, LLCInventors: Shi Han, Yingnong Dang, Dongmei Zhang, Song Ge
-
Patent number: 9459947Abstract: Techniques for error report processing are described herein. Error reports, received by a developer due to program crashes, may be organized into a plurality of “buckets.” The buckets may be based in part on a name and a version of the application associated with a crash. Additionally, a call stack of the computer on which the crash occurred may be associated with each error report. The error reports may be “re-bucketed” into meta-buckets to provide additional information to programmers working to resolve software errors. The re-bucketing may be based in part on measuring similarity of call stacks of a plurality of error reports. The similarity of two call stacks—a measure of likelihood that two error reports were caused by a same error—may be based in part on functions in common, a distance of those functions from the crash point, and an offset distance between the common functions.Type: GrantFiled: January 12, 2015Date of Patent: October 4, 2016Assignee: Microsoft Technology Licensing, LLCInventors: Dongmei Zhang, Yingnong Dang, Song Ge
-
Publication number: 20160267700Abstract: Techniques and arrangements for creating and editing motion data stories are described herein. In some implementations, the techniques and arrangements may determine semantic differences between consecutive slides intended to be used as the basis for a motion data story, and use the determined differences to determine appropriate transitional animations and/or animation effects. In addition to determined semantic differences, templates may also be used to determine the transitional animations and/or animation effects.Type: ApplicationFiled: March 10, 2015Publication date: September 15, 2016Inventors: He Huang, Haidong Zhang, Zhitao Hou, Dongmei Zhang, Song Ge
-
Publication number: 20160239550Abstract: A system for frequent pattern mining uses two layers of processing: a plurality of computing nodes, and a plurality of processors within each computing node. Within each computing node, the data set against which the frequent pattern mining is to be performed is stored in shared memory, accessible concurrently by each of the processors. The search space is partitioned among the computing nodes, and sub-partitioned among the processors of each computing node. If a processor completes its sub-partition, it requests another sub-partition. The partitioning and sub-partitioning may be performed dynamically, and adjusted in real time.Type: ApplicationFiled: April 27, 2016Publication date: August 18, 2016Inventors: Shi Han, Yingnong Dang, Dongmei Zhang, Song Ge
-
Patent number: 9348852Abstract: A system for frequent pattern mining uses two layers of processing: a plurality of computing nodes, and a plurality of processors within each computing node. Within each computing node, the data set against which the frequent pattern mining is to be performed is stored in shared memory, accessible concurrently by each of the processors. The search space is partitioned among the computing nodes, and sub-partitioned among the processors of each computing node. If a processor completes its sub-partition, it requests another sub-partition. The partitioning and sub-partitioning may be performed dynamically, and adjusted in real time.Type: GrantFiled: April 27, 2011Date of Patent: May 24, 2016Assignee: Microsoft Technology Licensing, LLCInventors: Shi Han, Yingnong Dang, Song Ge, Dongmei Zhang
-
Patent number: 9110769Abstract: Techniques for detecting, analyzing, and/or reporting code clone are described herein. In one or more implementations, clone-code detection is performed on one or more source code bases to find true and near clones of a subject code snippet that a user (e.g., a software developer) expressly or implicitly selected. In one or more other implementations, code clone is analyzed to estimate the code-improvement-potential (such as bug-potential and code-refactoring-potential) properties of clones. One or more other implementations present the results of code clone analysis with indications (e.g., rankings) of the estimated properties of the respective the clones.Type: GrantFiled: April 1, 2010Date of Patent: August 18, 2015Assignee: Microsoft Technology Licensing, LLCInventors: Dongmei Zhang, Yingnong Dang, Yingjun Qiu, Song Ge
-
Publication number: 20150127990Abstract: Techniques for error report processing are described herein. Error reports, received by a developer due to program crashes, may be organized into a plurality of “buckets.” The buckets may be based in part on a name and a version of the application associated with a crash. Additionally, a call stack of the computer on which the crash occurred may be associated with each error report. The error reports may be “re-bucketed” into meta-buckets to provide additional information to programmers working to resolve software errors. The re-bucketing may be based in part on measuring similarity of call stacks of a plurality of error reports. The similarity of two call stacks—a measure of likelihood that two error reports were caused by a same error—may be based in part on functions in common, a distance of those functions from the crash point, and an offset distance between the common functions.Type: ApplicationFiled: January 12, 2015Publication date: May 7, 2015Inventors: Dongmei Zhang, Yingnong Dang, Song Ge
-
Patent number: 8949675Abstract: Techniques for error report processing are described herein. Error reports, received by a developer due to program crashes, may be organized into a plurality of “buckets.” The buckets may be based in part on a name and a version of the application associated with a crash. Additionally, a call stack of the computer on which the crash occurred may be associated with each error report. The error reports may be “re-bucketed” into meta-buckets to provide additional information to programmers working to resolve software errors. The re-bucketing may be based in part on measuring similarity of call stacks of a plurality of error reports. The similarity of two call stacks—a measure of likelihood that two error reports were caused by a same error—may be based in part on functions in common, a distance of those functions from the crash point, and an offset distance between the common functions.Type: GrantFiled: November 30, 2010Date of Patent: February 3, 2015Assignee: Microsoft CorporationInventors: Dongmei Zhang, Yingnong Dang, Song Ge
-
Patent number: 8578213Abstract: Execution traces are collected from multiple execution instances that exhibit performance issues such as slow execution. Call stacks are extracted from the execution traces, and the call stacks are mined to identify frequently occurring function call patterns. The call patterns are then clustered, and used to identify groups of execution instances whose performance issues may be caused by common problematic program execution patterns.Type: GrantFiled: April 27, 2011Date of Patent: November 5, 2013Assignee: Microsoft CorporationInventors: Shi Han, Yingnong Dang, Song Ge, Dongmei Zhang, Bin Zhao, Feng Liang, Chao Bian, Xiangpeng Zhao, Cong Chen, Hang Li, Prashant Ratanchandani
-
Publication number: 20120278346Abstract: A system for frequent pattern mining uses two layers of processing: a plurality of computing nodes, and a plurality of processors within each computing node. Within each computing node, the data set against which the frequent pattern mining is to be performed is stored in shared memory, accessible concurrently by each of the processors. The search space is partitioned among the computing nodes, and sub-partitioned among the processors of each computing node. If a processor completes its sub-partition, it requests another sub-partition. The partitioning and sub-partitioning may be performed dynamically, and adjusted in real time.Type: ApplicationFiled: April 27, 2011Publication date: November 1, 2012Applicant: Microsoft CorporationInventors: Shi Han, Yingnong Dang, Song Ge, Dongmei Zhang
-
Publication number: 20120278658Abstract: Execution traces are collected from multiple execution instances that exhibit performance issues such as slow execution. Call stacks are extracted from the execution traces, and the call stacks are mined to identify frequently occurring function call patterns. The call patterns are then clustered, and used to identify groups of execution instances whose performance issues may be caused by common problematic program execution patterns.Type: ApplicationFiled: April 27, 2011Publication date: November 1, 2012Applicant: Microsoft CorporationInventors: Shi Han, Yingnong Dang, Song Ge, Dongmei Zhang, Bin Zhao, Feng Liang, Chao Bian, Xiangpeng Zhao, Cong Chen, Hang Li, Prashant Ratanchandani
-
Publication number: 20120278659Abstract: A call pattern database is mined to identify frequently occurring call patterns related to program execution instances. An SVM classifier is iteratively trained based at least in part on classifications provided by human analysts; at each iteration, the SVM classifier identifies boundary cases, and requests human analysis of these cases. The trained SVM classifier is then applied to call pattern pairs to produce similarity measures between respective call patterns of each pair, and the call patterns are clustered based on the similarity measures.Type: ApplicationFiled: April 27, 2011Publication date: November 1, 2012Applicant: Microsoft CorporationInventors: Shi Han, Yingnong Dang, Song Ge, Dongmei Zhang