International Business Machines Corporation

Salesforce, Inc.

Utopus Insights, Inc.

Palo Alto Research Center Inc.

Globalfoundries Inc.

University of Michigan

Guangzhou Luxvisions Innovation Technology Limited

Green Market Square Limited

Salesforce Com, Inc.

Rui Zhang

Tarun Kumar

Younghun Kim

Abhishek Raman

Harsh Chaudhary

Prasenjit Sarkar

Reshu Jain

Heiko Hary Ludwig

Zhenhua Hu

Yonggang Hu

Nathalie Baracaldo Angel

Hongzhi Wang

Subashini Balachandran

David Darden Chambliss

Jonathan Frederick Brunn

Ildar Khabibrakhmanov

Mark Alan Lavin

Young Min Lee

Kurt Edward Partridge

William W Owen

Rama Krishna Korlapati

Punit B Shah

Rachael Marie Huston Dickens

Tanveer Fathima Syeda-Mahmood

Wei Zhang

Min Li

Robert Roy Price

Joseph S Glider

Wendy Ann Belluomini

Oliver Brdiczka

Markus Ettl

James M A Begole

Maurice Kyojin Chu

William Griswold Sherman

Himabindu Pucha

Ali Anwar

Lukas Rupprecht

Thomas Steinke

Deepika Kakrania

Gopalkrishna Veni

Hongxia Yang

Stacey Truex

Fei Liu

Chandramouli Visweswariah

David Nahamoo

Dean Hildebrand

Rong Zhou

Wayne A Sawdon

Joseph S Hyde, Ii

Renu Tewari

Siyuan Lu

Ulrich Alfons Finkler

Fook-Luen Heng

Ersin Uzun

Bo Zhang

Richard Chow

David Shing-Ki Kung

Mohit Saxena

Xiaodong Cui

Kevin Wilson Warren

Prashant Pandey

Conrad M Albrecht

Marcus Oliver Freitag

Lazaros Polymenakos

Yi Zhou

Lei Liu

Honglak Lee

Ami H Dewar

Lianjun An

Dragomir Radkov Radev

Dimitrios Skourtis

Minwei Feng

Douglas Doole

Serge Philippe Rielau

Nathalie Baracaldo-Angel

Yi Xia

Prateek Swamy

Young T Chae

Abhijith Anilkumar

HaiChao Du

Prashasthi Prabhakar

JiJi Zhong

Dominique Jeremy Evans

Jefferson Huang

David Gershuni

Mikhail Lychagin

Lukas L Rupprecht

Muzammil Hussain

Emiliano DeCristofaro

Joan Melia-Segui

Techniques are disclosed relating to updating a database schema. A computer system may store database schema versions for a database schema. A given schema version may have an active state or one of multiple non-active states. One of the schema versions may be a first schema version that is in the active state and the remaining schema versions may be in the non-active states. The computer system may receive updates to apply to the database schema. The computer system may transition a second schema version from a first non-active state that does not permit structural or content updates to a second non-active state that permits structural and content updates. The computer system may apply the updates to the second database schema version and then transition the second database schema version into the active state and the first database schema version into the first non-active state.

Database object lifecycle states

An application server may identify a set of functions for the source code associated with an uninterruptable software service. In some examples, the subset of the set of functions may include one or more functions having a computational expense greater than a threshold. The application server may create a compilation template to compile a subset of the set of functions to a shared library, and may perform an instrumentation to identify the subset of the set of functions based on the shared library. The application server may create an optimized version of the shared library based on performing the instrumentation. The application server may then change one or more portions of the source code to repoint to the optimized version of the shared library based on creating the optimized version of the shared library.

Dynamically applying profile-guided optimization to a DBMS

Embodiments relate to an intelligent computer platform to build a federated learning framework including creating a hierarchy of machine learning models (MLMs). The hierarchy of MLMs has a primary MLM in a primary layer. Training the primary MLM includes capturing contributing model updates across at least one communication channel. A secondary MLM is created and logically positioned in a secondary layer of the hierarchy. The secondary MLM is operatively coupled to the primary MLM across the at least one communication channel. The created secondary MLM is initialized, including cloning weights and framework of the primary MLM into the secondary MLM, and populated with secondary data. The populated data has model updates local to the created secondary MLM. The secondary MLM is logically stored local to the secondary layer, and limits access to the secondary MLM to the secondary layer.

Building a federated learning framework

Provided are a computer program product, system, and method for generation of tasks and retraining machine learning modules to generate tasks based on feedback for the generated tasks. A machine learning module processes an input text message sent in the communication channel to output task information including an intended action and a set of associated users. A task message is generated including the output task information of a task to perform. The task message is sent to a user interface panel in a user computer. Feedback is received from the user computer on the output task information in the task message. The machine learning module is retrained to output task information from the input text message based on the feedback to reinforce likelihood correct task information is outputted and reinforce lower likelihood incorrect task information is outputted.

Automatic generation of tasks and retraining machine learning modules to generate tasks based on feedback for the generated tasks

Techniques are disclosed relating to managing data objects. A computer system may access a first application comprising a set of functions and may execute a second application to facilitate execution of the first application. The first application may be written in a different programming language than the second application. The second application may manage, on behalf of the first application, memory resources of data objects that are defined within ones of the set of functions of the first application. The second application may establish, for the data objects, ownership information that indicates which ones of the functions are utilizing the data objects. The ownership information may be usable to determine, upon completion of a function of the set of functions, whether a given one of the data objects can be deallocated based on whether the given data object is being utilized by another function.

Lifecycle tracking of data objects

Techniques are disclosed that relate to manipulating a chain of database objects without locking the chain. A computer system may maintain a chain that orders a set of database objects stored in a cache of the computer system. The computer system may receive a set of requests to perform database transactions. Based on those received set of requests, the computer system may determine to perform a plurality of chain operations that involve modifying the chain. The computer system may perform two or more of the plurality of chain operations at least partially in parallel using a set of atomic operations without acquiring a lock on the chain.

Mechanisms for maintaining chains without locks

Using a natural language analysis, a current message is classified into a current message class, the current message being a portion of an interaction in narrative text form. Using a trained message class prediction model, a probability of a previous message class having resulted in the current message class is determined. A previous message is extracted from the interaction using the probability, the previous message being a portion of the interaction occurring prior to the current message, the previous message being classified into the previous message class.

Identifying related messages in a natural language interaction

Techniques regarding privacy preservation in a federated learning environment are provided. For example, one or more embodiments described herein can comprise a system, which can comprise a memory that can store computer executable components. The system can also comprise a processor, operably coupled to the memory, and that can execute the computer executable components stored in the memory. The computer executable components can comprise a plurality of machine learning components that can execute a machine learning algorithm to generate a plurality of model parameters. The computer executable components can also comprise an aggregator component that can synthesize a machine learning model based on an aggregate of the plurality of model parameters. The aggregator component can communicate with the plurality of machine learning components via a data privacy scheme that comprises a privacy process and a homomorphic encryption process in a federated learning environment.

Private and federated learning

Techniques are disclosed relating to managing dependencies between shared data objects in a database system. In various embodiments, a data storage system may maintain a function cache that is operable to store functions that are accessible to one or more applications hosted by a server system. In response to a request to invalidate a first data type utilized by one or more of these functions, various embodiments may include accessing an inverted index that specifies a mapping between data types and corresponding functions, stored in the function cache, which depend on those data types. Using the inverted index, various embodiments include identifying a set of functions that depend on the first data type and invalidating the set of functions in the function cache such that the set of functions are no longer accessible by the applications.

Dependency management for shared data objects in a database system

A system for forecasting leaks in a fluid-delivery pipeline network. The system identifies a subsystem in the pipeline network that comprises a plurality of topologically connected stations. The system accesses historical temporal sensor measurements of a plurality of variables of the stations that are directly connected and generates a temporal causal dependency model for a first control variable at the first station in the subsystem, based on the plurality of time series of sensor measurements of a second variable of the first station, and temporal delay characteristics of the plurality of time series of sensor measurements of the second variable at the stations directly connected to the first station. The system automatically calculates a normal operating value of the first control variable at the first station and the deviations between actual measured values and the normal operating value and determines a threshold deviation that indicates a leak event.

System and method for forecasting leaks in a fluid-delivery pipeline network

Methods, computer readable media, and devices for identifying and preventing invalid memory access. A method may include defining a dynamic scope for an operation, receiving a request to allocate a portion of the range of shared memory, allocating a monotonically increasing portion of the range of shared memory such that a subsequent request to allocate memory is allocated a different portion of the range of shared memory, receiving a request to deallocate the allocated portion of the range of shared memory, deallocating the allocated portion of the range of shared memory by protecting the deallocated portion of the range of shared memory from any subsequent access, and in response to an access of the protected deallocated portion of the range of shared memory by one of the one or more threads or processes of the operation, trapping and terminating the one thread or process.

Identifying and preventing invalid memory access

Techniques for image processing and transformation are provided. A plurality of images and a plurality of maps are received, and a system of neural networks is trained based on the plurality of images and the plurality of maps. A first image is received, and a first map is generated by processing the first image using the system of neural networks.

Deep learning to correct map and image features

The present disclosure provides a camera module and an electronic device. The camera module comprises a bearing base, an image sensor, a baseplate, a lens bracket, and a lens group. The image sensor is disposed at the bearing base. The baseplate is disposed at the bearing base and is separated from the image sensor. The baseplate is disposed on the periphery of the image sensor. The baseplate is electrically connected with the image sensor. The lens bracket is disposed at the bearing base. The image sensor and a part of the baseplate are disposed in the lens bracket. The lens bracket comprises a photosensitive through hole. The image sensor corresponds to the photosensitive through hole. The lens group is disposed at the lens bracket and corresponds to the image sensor through the photosensitive through hole.

Camera module and electronic device

Technical solutions are described for predicting linepack delays. An example method includes receiving temporal sensor measurements of a first fluid-delivery pipeline network and generating a causality graph of the first fluid-delivery pipeline network. The method also includes determining a topological network of the stations based on the causality graph, where the topological network identifies a temporal delay between a pair of stations. The method also includes generating a temporal delay prediction model based on the topological network and predicting the linepack delays of a second fluid-delivery pipeline network based on the temporal delay prediction model, where a compressor station of the second fluid-delivery pipeline network compresses fluid based on the predicted linepack delays to maintain a predetermined pressure.

Linepack delay measurement in fluid delivery pipeline

Using a natural language analysis, a current message is classified into a current message class, the current message being a portion of an interaction in narrative text form. For the interaction using a state prediction model, an interaction outcome corresponding to the current message class is forecasted, the forecasting comprising computing a probability that the current message class will result in a successful message class. Using the state prediction model, a set of next message classes and a set of predicted interaction outcomes are determined, each message in the set of next message classes corresponding to the current message class, each predicted interaction outcome in the set of predicted interaction outcomes corresponding to a next message class in the set of next message classes. According to the corresponding predicted interaction outcome, the set of next message classes is ranked.

Outcome of a natural language interaction

Growing community of inventors

San Francisco, CA, United States of America

Rui Zhang