Microsoft Technology Licensing, LLC

Other

Adam Clyde Eversole

Michael Dean Stokes

Joseph Spencer King

Joseph Ternasky

Robert L C Parker

Michael M Byrd

Patrick Markus Baudisch

Paul Stephen Hellyar

Avronil Bhattacharjee

Kapil Sharma

Bo Qin

Pulin Thakkar

Quinn Hawkins

Dong Yu

Gang Li

Oliver H Foehr

Daniel Cory

Maneesh Agrawala

Frank Torsten Bernd Seide

Miltiadis Vafiadis

Xie Chen

Benjamin Eric Ross

The use of a pipelined algorithm that performs parallelized computations to train deep neural networks (DNNs) for performing data analysis may reduce training time. The DNNs may be one of context-independent DNNs or context-dependent DNNs. The training may include partitioning training data into sample batches of a specific batch size. The partitioning may be performed based on rates of data transfers between processors that execute the pipelined algorithm, considerations of accuracy and convergence, and the execution speed of each processor. Other techniques for training may include grouping layers of the DNNs for processing on a single processor, distributing a layer of the DNNs to multiple processors for processing, or modifying an execution order of steps in the pipelined algorithm.

Deep neural networks training for speech and pattern recognition

Techniques to manage a whiteboard for multimedia conference events are described. An apparatus may comprise a whiteboard manager component operative to manage whiteboard recording, image processing and reproduction operations for a multimedia conference event. The whiteboard manager component may comprise an image quantizer module operative to receive an image of a writing surface with pen strokes, and quantize each pixel of the image into a predetermined number of colors, an attribute extractor module communicatively coupled to the image quantizer module, the image quantizer module operative to extract stroke attribute information for the pen strokes from the quantized pixels, and a whiteboard interface module communicatively coupled to the attribute extractor module, the whiteboard interface module operative to send the stroke attribute information to a meeting console for rendering as digital pen strokes on a digital writing surface. Other embodiments are described and claimed.

Techniques to manage a whiteboard for multimedia conference events

Techniques to manage a whiteboard for multimedia conference events are described. An apparatus may comprise a whiteboard manager component operative to manage whiteboard recording, image processing and reproduction operations for a multimedia conference event. The whiteboard manager component may comprise an image quantizer module operative to receive an image of a writing surface with pen strokes, and quantize each pixel of the image into a predetermined number of colors, an attribute extractor module communicatively coupled to the image quantizer module, the attribute extractor module operative to extract stroke attribute information for the pen strokes from the quantized pixels, and a whiteboard interface module communicatively coupled to the attribute extractor module, the whiteboard interface module operative to send the stroke attribute information to a meeting console for rendering as digital pen strokes on a digital writing surface. Other embodiments are described and claimed.

The present invention is generally directed to assisting users to position a pointer on a target. Embodiments of the present invention provide a system, method, and computer-readable medium that guide a pointer toward a target. After the pointer is located on a target, pointer movement continues to be adjusted so that the user may easily keep the pointer located on the target.

System and method for acquiring a target with intelligent pointer movement

A system and related techniques generate a unified representation of graphical representations, such as documents, graphic images or others, for displaying on a display screen, printing on a printer, faxing via a facsimile machine or outputting in other ways. In embodiments a media integration layer, which may reside in the operating system level, may generate renderings from graphics primitives, and send that rendering via a mediation engine to a device driver, for instance a print engine located in a printer. The device engine may return a preview of the output representation to the media integration layer for adjustments or to accept and print the output, all without a necessity for or reliance upon a device-specific driver in the operating system. Since both the display screen and the output device operate off of the same base representation, neither the display nor output device needs to attempt to emulate the output characteristics of the other, and screen-to-print uniformity may be enhanced.

System and method for generating unified image output

A system and related techniques generate printing and other output operations via a resolution-independent visual tree object. According to embodiments the operating system or other resource may receive the visual tree object and related data from applications or other sources, and perform logic to scale or process the visual output for resolution, effects or other purposes directly from that object. The operating system or other resource itself may be configured to receive the componentized visual tree object and condition or render that object for output to a printer, facsimile machine, copy machine or other output device. The operating system may serialize or spool that object from the source object in native form, relieving applications of the need to prepare documents or other output objects to generate specified resolutions, effects or to adapt the output to specific output drivers.

System and method for generating resolution-independent output via visual tree object

A system and related techniques detect and manage the set of rights associated with a componentized electronic document, including to separately protect and manage individual text, graphic, audio, video or other parts of a multimedia presentation. Unlike prior digital asset management platforms which could enforce encryption or other protection simply around the entire object, the invention permits individualized control over separate media components, to permit distribution for example of some parts, such as text passages, while preventing unauthorized access or copying of audio, photographic, video or other files or content. The electronic document may maintain a list of media components and the set of rights associated with each such component, and a content server or other source may apply key encryption or other safeguards to guide the delivery of the overall media object or its parts, for instance to view in a browser, print on a printer or otherwise access, manipulate or output the content.

System and method for management of a componentized electronic document retrievable over a network

The present invention provides a system, method, and computer-readable medium that adjusts pointer movement when an input device causes the pointer to intersect one or more alignment areas. Pointer movement is adjusted in the alignment areas to allow users to easily align GUI objects. The adjusted pointer movement is not proportional to input received from an input device. Instead alignment areas have 'friction' in that the pointer is less responsive to input compared to other areas of the computer display.

System and method for aligning objects using non-linear pointer movement

An algorithm is provided for optimizing the layout of labels associated with a visual data object such as a chart. The labels are first placed into a chart definition file that defines the labels as a shape. An initial layout of the labels is created and scored. The shapes are then manipulated iteratively until an optimal layout of the labels is obtained that corresponds to the layout having the score closest to an optimal score. The optimal layout is then used in rendering the visual data object on a screen.

System and method for automatic label placement on charts

A system and related techniques automatically exercise the output capabilities of a printer or other output device, and return a log or record of those capabilities to the print manager of a host operating system or other destination. According to embodiments, the host, client or other machine may decide to query a printer, facsimile machine, copy or multifunction machine or other device to determine its range of output capabilities such as available fonts, graphical or rending routines, firmware compatibility or other functions or capabilities. The host may transmit a composite electronic document, containing various components including intermediate language objects which may be interpreted to execute various output processing, such as exercising a set of fonts, scaling, rotating, or other rendering or output functions. The results may be captured in the printer or other output device, for instance by dumping an output buffer or returning execution codes, and returned to the host machine where the available capabilities may be displayed, including in embodiments a rendering of the physical output or simulated physical output on the user interface. The invention offers enhanced profiling and verification for printers and other output systems, compared to manual test pages and other test procedures.

System and method for automatic testing of output device capabilities via electronic document

The present invention can provide an extensible file format compatible with previous, current, and future versions of an application program. The present invention can comprise including file version watermarks in an original data file of an application program. The file version watermarks can indicate various properties of the original data file. The file version watermarks can include a high version watermark, a last version watermark, a low version watermark, a creation version watermark, and an object version watermark. Each file version watermark can indicate changes, additions, or deletions made to the file. The present invention can use the file version watermarks to determine whether the file, or particular information in the file, corresponds to the previous, active, or future versions of an application program. Accordingly, the present invention can then load or save the original data file based on a result of the determination.

Growing community of inventors

Redmond, WA, United States of America

Adam Clyde Eversole