International Business Machines Corporation

David E Cox

David Bruce Lindquist

David Louis Kaminsky

John Ryan McGarvey

Kent F Hayes, Jr

Abdi Salahshour

Anna Wanda Topol

Jinjun Xiong

Chuang Gan

James Corvin Fletcher

Yada Zhu

Rajesh Radhakrishnan

Yang Zhang

Jessie Carrigan Rosenberg

Shiyu Chang

John Patrick Whitfield

Craig M Lawton

Kent Fillmore Hayes, Jr

Carl Shawn Kessler

Kaizhi Qian

Jonathan Scott Greenfield

Mathew James Brozowski

John Ferguson Linton

Karl D Gottschalk

Akash Srivastava

Sundari M Revanur

Dan Gutfreund

Anne Renee Reidelbach

Zixuan Yuan

John J Lambert

Unsupervised cross-domain data augmentation techniques for long-text document based prediction and explanation are provided. In one aspect, a system for long-document based prediction includes: an encoder for creating embeddings of long-document texts with hierarchical sparse self-attention, and making predictions using the embeddings of the long-document texts; and a multi-source counterfactual augmentation module for generating perturbed long-document texts using unlabeled sentences from at least one external source to train the encoder. A method for long-document based prediction is also provided.

Unsupervised cross-domain data augmentation for long-document based prediction and explanation

A computer-implemented method is provided of using a machine learning model for disentanglement of prosody in spoken natural language. The method includes encoding, by a computing device, the spoken natural language to produce content code. The method further includes resampling, by the computing device without text transcriptions, the content code to obscure the prosody by applying an unsupervised technique to the machine learning model to generate prosody-obscured content code. The method additionally includes decoding, by the computing device, the prosody-obscured content code to synthesize speech indirectly based upon the content code.

Global prosody style transfer without text transcriptions

A generator network of a variational autoencoder can be trained to approximate a simulator and generate a first result. The simulator is associated with input data, based on which the simulator outputs output data. A training data set for the generator network can include the simulator's input data and output data. Based on the simulator's output data and the first result of the generator network, an inference network of the variational autoencoder can be trained to generate a second result. The second result of the trained inference network inverts the first result of the generator and approximates the simulator's input data. The trained inference network can function as an inverted simulator.

Simulator-assisted training for interpretable generative models

A method, data processing system, and computer program product for caching information. A computer determines whether at least one of a client data processing system and a hardware device in a plurality of hardware devices has performed a number of operations that have been identified as being associated with a request that is sent to a first hardware device in the plurality of hardware devices. The first hardware device is configured to obtain information from a second hardware device in the plurality of hardware devices for use in generating a response to the request. The computer caches the information in a cache for the plurality of hardware devices prior to a subsequent request for the information being sent to the first hardware device in response to a determination that the at least one of the client data processing system and the hardware device has performed the number of operations.

Information caching system

Methods, systems and computer program products for management of application programs on a network including a server supporting client stations are provided. The server provides applications on-demand to a user logging in to a client supported by the server. Mobility is provided to the user and hardware portability is provided by establishing a user desktop interface responsive to a login request which presents to the user a desktop screen through a web browser interface which accesses and downloads selected application programs from the server responsive to a request from the user on the user desktop screen at the client. The application program is then provided from the server and executed at the client. The application program may further be customized to conform to the user's preferences and may also provide for license use management by determining license availability before initiating execution of the application program. Finally, software distribution and installation may be provided from a single network management server.

Methods, systems and computer program products for distribution of application programs to a target station on a network

Methods, systems and computer program products are provided for controlling a user session in a network by defining rules for controlling user sessions based on characteristics of an operating environment, determining the characteristics of an instance of an operating environment associated with a user session in the network and applying the defined rules to the determined characteristics to control the user session based on the characteristics of the instance of the operating environment. In particular, the type of network connection, the type of device connected to the network, the user identification of the user and/or the identification of an application executed by the user may be determined and the session controlled based on this determination. Preferably, the rules control the content of communications provided during the user session based on the characteristics of the operating environment.

Methods, systems and computer program products for policy based network control of characteristics of user sessions

A method, system, and computer program product code for dynamically refreshing user credentials in a distributed processing environment. The present invention provides for fast, local refresh of credentials by a server if the credentials expire during an on-going secure operation. This technique avoids the need for rolling back the operation and requiring the client to restart after acquiring fresh credentials. The ability for a systems administrator to invalidate credentials which have been compromised is maintained.

Dynamic credential refresh in a distributed system

Management of configurable applications programs on a network is provided by using two program files for each configurable application program. The two program files are provided to a network server station which operates as the on-demand server for software deployment and may also act as the application server. The on-demand server makes the first (configuration manager) program available to an administrator to obtain preferences for the configurable preferences of the application program which have been designated as administrator preferences. The on-demand server also provides a second (application launcher) program to its client stations. The application launcher program not only provides for a user interface to execute the application program itself but also allows a user to specify one or more of the configurable parameters of the application program. An icon is displayed on the screen of the client station through which a user may initiate execution of the application program. The application launcher program provides the identity of the user to the server along with the request to initiate execution of the application program. The on-demand server then initiates execution of the application program using stored values for the user and administrator set preferences or, if no preferences have yet been obtained for the particular user, obtains user preferences before initiating execution of the application program.

Methods, systems and computer program products for license use management on a network

Methods, systems and computer program products for centralized management of application programs on a network

Methods, systems and computer program products for management of configurable application programs on a network

In a network composed of communicating resources, the status of an aggregate resource may be determined by calculations based upon the status assumed by the real resources contained within the aggregate rather than being determined directly. The status of real resources may be propagated to still higher aggregate resources whose own status is based upon the status of the real resources contained within it. An aggregate resource may have multiple potential statuses and a decision as to which status to report for an aggregate resource at a given time is calculated based upon the status of the underlying real resources which it contains. Each real resource has associated with it another parameter which is its aggregation priority value indicating to what degree a change in its status will affect higher level nodes or aggregates in a network hierarchy. In assigning a new aggregate status to a given aggregate resource, the calculation for aggregate status determines the status of the real underlying resources that are descendents of the aggregate resource, retrieves the appropriate threshold levels for status change for the aggregate, and checks the aggregation priority of the real resources that are descendents of the aggregate and are currently in an unsatisfactory state and then assigns a new aggregate resource status.

Method and system for reporting the status of an aggregate resource

Growing community of inventors

Raleigh, NC, United States of America

David E Cox