International Business Machines Corporation

Chinese Academy of Sciences

Analog Devices International Unlimited Company

Ke Wei Wei

Xin Yang

Shuo Li

Ping Liang

Heng Liu

Xiang Sean Zhou

Jian Wei Zhang

Chen Xin Yu

Miao Zheng

Li Ping Yang

Xinyu Liu

Shuang Ys Yu

Xinhua Wang

Sen Huang

Xiaobo Wang

Hong Min

You-Chin Gene Fuh

Jing Zhou

Xiao Li

Hao Wu

Chao Zhao

Serge Bourbonnais

Gong Su

Maryela Evelin Weihrauch

Wenwu Wang

Yi Dai

Dong Jie Wei

Meichi M Lin

Xi Long Zheng

Ting Xu Guan

Wei Liu

Dong Chen

Jing James Xu

Thomas Abel Beavin

Qi Zhang

ShengYan Sun

Meng Wan

Peter Bendel

Yuan Feng

Junfeng Li

Nan Jiang

Jun Wang

Shu Hua Liu

Patrick Dooling Bossman

Chi Liu

Guang Ming Zhang

DongJie Wei

Chu Yun Tong

Chun Guang Zeng

Tian Tian

Ling Zhuo

Ge Song

Shuo Wu

Zhenyu Shi

Aine McCarthy

Yan Jia

Michael Schenker

Guo Ran Sun

Xinke Wang

Ji Gao Fu

Yao Zhao

Shuchun Xie

Zhang Hao

Yuankun Wang

Haibo Yin

Zhen Hua Dong

En Zhong Wang

Qilong Bao

Jean-François Puget

Qi Wen

An approach is provided in which a method, system, and program product analyze, while training a machine learning model, a set of first data transformation operators in a first data preparation pipeline that generates a plurality of constructed features from a set of training data. The method, system, and program product create a plurality of second data preparation pipelines from the first data preparation pipeline, wherein the set of first data transformation operators are converted to a set of second data transformation operators and each assigned to one of the plurality of second data preparation pipelines. The method, system, and program product deploy the plurality of second data preparation pipelines to a runtime system.

Auto feature preparation for high performance online inferencing

An embodiment generates an initial set of training data from monitoring data. The initial set of training data is generated by combining outputs from a plurality of pretrained classifiers. The embodiment trains a new classification model using the initial set of training data to identify anomalies in monitoring data. The embodiment performs a multiple-level clustering of the data samples resulting in a plurality of clusters of sub-clusters of data samples, and generates a review list of data samples by selecting a representative data sample from each of the clusters. The embodiment receives an updated data sample from the expert review that includes a revised target classification for at least one of the data samples of the expert review list. The embodiment then trains another replacement classification model using a revised set of training data that includes the updated data sample and associated revised target classification.

Intelligent expansion of reviewer feedback on training data

A method for processing a plurality of queries is provided according to embodiments of the present disclosure. In this method, based on a plurality of queries and an execution plan for the plurality of quires, a plurality of record identification (ID) numbers can be stored into a pool in a numerical order. Each of the plurality of record ID numbers can identify a data record in a database. Then, the execution plan can be performed to batch a plurality of data records corresponding to the plurality of record ID numbers in the database based on a distribution of the plurality of record ID numbers in the pool.

Query processing

A method includes generating a plurality of vectors representing words in a plurality of documents about an information technology (IT) system and clustering the plurality of vectors to produce a plurality of clusters. The method also includes identifying a cluster of the plurality of clusters that contains a plurality of clustered vectors, generating a feature based on a plurality of words represented by the plurality of clustered vectors, and training a machine learning model to identify an anomaly in the IT system based on the feature.

Automatic anomaly detection

Aspects of the invention include a computer-implemented method that receives, by a processor, an ensemble decision tree and generates, by the processor, native code from the ensemble decision tree. The method compiles, by the processor, the native code into machine language and scores, by the processor, the execution time of the native code. The method dynamically reoptimizes, by the processor, portions of the native code corresponding to the most traversed portion of the ensemble decision tree.

Generating native code with dynamic reoptimization for ensemble tree model prediction

A GaN-based superjunction vertical power transistor and a manufacturing method thereof. The transistor includes: a N-GaN layer; a first P-GaN layer as a current blocking layer, formed on the N-GaN layer and having a gate region window; and a thin barrier Al(In, Ga)N/GaN heterostructure conformally formed on the current blocking layer and filling the bottom and one or more sidewalls of the gate region window, wherein the N-GaN layer has an etched groove completely or partially filled with a second P-type GaN layer, an N-GaN layer is formed under the second P-type GaN layer, and the N-GaN layer is in direct contact with the second P-type GaN layer and the N-GaN layer to form a superjunction composite structure.

GaN-based superjunction vertical power transistor and manufacturing method thereof

Methods, computer program products, and systems are presented. The method computer program products, and systems can include, for instance: receiving, by a production database, a query statement for processing by the production database; the production database in performing runtime execution processing of the query statement identifying one or more query statement value of the query statement; the production database in performing runtime execution processing of the query statement establishing, using a predictive model, an access path based on the identified one or more query statement value; and the production database using the established access path in performing runtime execution processing of the query statement.

Multiple access path selection by machine learning

Database configuration tuning is provided. A set of database nodes having similar data factors is selected in a centerless network of database nodes. Configuration models corresponding to the set of database nodes are trained using data parallelism. Trained configuration models corresponding to the set of database nodes are combined to form a federated configuration model. It is determined whether performance indicators corresponding to the set of database nodes are greater than a performance threshold level. In response to determining that the performance indicators corresponding to the set of database nodes are greater than the performance threshold level, a database configuration corresponding to the federated configuration model is recommended to a new database node. The new database node is joined to the centerless network.

Database tuning using a federated machine learning system of a centerless network

In an approach to determining an access method for a partition in a partition table, a computer receives a query and determines if there is a partition table utilized by the query. When there is a partition table utilized by the query, then the computer determines that a partition in the partition table meets the plurality of conditions of the query. The computer collects a plurality of partition level statistics for the partition that meets the plurality of conditions of the query. Additionally, the computer determines, based, at least in part, on the plurality of partition level statistics, a cost for one or more access methods for the partition that meets the plurality of conditions of the query. Furthermore, the computer determines, based, at least in part, on the cost for each access method, an access method for the partition that meets the plurality of conditions of the query.

Partition access method for query optimization

A computer-implemented method includes identifying a primary database interaction statement and determining one or more intermediate database interaction statements based on the primary database interaction statement. The computer-implemented further includes determining, for each of the one or more intermediate database interaction statements, an intermediate statement transformation indicator. The computer-implemented method further includes determining one or more transformed database interaction statements based on each intermediate statement transformation indicator.

Syntactical transformation of database interaction statements

An electronic circuit comprises an input voltage circuit, an analog-to-digital converter (ADC) circuit, and logic circuitry. The input voltage circuit is configured to generate multiple input voltages. The ADC circuit is configured to convert the multiple input voltages to first digital values using the first longer ADC acquisition time and convert the multiple input voltages to second digital values using the second shorter ADC acquisition time. The logic circuitry is configured to determine calibration information for the ADC circuit using the first digital values and the second digital values, and scale analog-to-digital (A/D) conversion results of the ADC circuit using the calibration information.

Calibration method for precision signal chain linearity

A computer-implemented method includes identifying a primary database interaction statement and determining one or more intermediate database interaction statements based on the primary database interaction statement. The computer-implemented further includes determining, for each of the one or more intermediate database interaction statements, an intermediate statement transformation indicator. The computer-implemented method further includes determining one or more transformed database interaction statements based on each intermediate statement transformation indicator. A corresponding computer program product and computer system are also disclosed.

A computer-implemented method includes receiving a batch processing job. The method identifies multiple access paths for the batch processing job. The method determines one or more runtime costs for each of the multiple access paths. The method determines whether the batch processing job includes comparable access paths. The method responds to the batch processing job not including comparable access paths by receiving a query and selecting a desired access, based on the one or more runtime costs. The method responds to the batch processing job including comparable access paths by determining one or more context costs for each of the one or more access paths, receiving a query, identifying a context for the query, and selecting a desired access path for the query. The desired access path is based on the one or more runtime costs, the one or more context costs, and the context for the query.

Context based access path selection

A GaN-based enhancement-mode power electronic device and a method for manufacturing the same. The GaN-based enhancement-mode power electronic device comprises: a substrate; a thin barrier Al(In,Ga)N/GaN heterostructure formed on the substrate; a gate, a source, and a drain formed on the thin barrier Al(In,Ga)N/GaN heterostructure. An AlN or SiNx passivation layer is formed on access regions between the gate and the source and between the gate and the drain, respectively, such that two dimensional electron gas is recovered in channels of the thin barrier Al(In,Ga)N/GaN heterostructure below the MN passivation layer by utilizing the MN passivation layer having polarization characteristics, or by using the SiNx passivation layer with positive fixed bulk/interface charges, so as to reduce on-resistance of the device and inhibit high-voltage current collapse in the device.

Semiconductor device and method for manufacturing the same

Aspects of the present disclosure relate to an approach for generating query execution plans for a database query. A computer-implemented method comprises determining whether a size of a file to be generated during execution of a database query is variable in response to obtaining the database query. The method further comprises determining a plurality of ranges for the size of the file in response to determining that the size of the file is variable. The method further comprises generating a plurality of query execution plans corresponding to the plurality of ranges. Accordingly, a plurality of query execution plans corresponding to different file sizes can be generated for the database query, and an efficient and cost effective query execution plan may be selected based on the actual file size during the execution of the database query.

Generation of query execution plans

Growing community of inventors

Beijing, China

Ke Wei Wei