Three-dimensional (3D) convolution with 3D batch normalization

    公开(公告)号:US10282663B2

    公开(公告)日:2019-05-07

    申请号:US15237575

    申请日:2016-08-15

    Abstract: The technology disclosed uses a 3D deep convolutional neural network architecture (DCNNA) equipped with so-called subnetwork modules which perform dimensionality reduction operations on 3D radiological volume before the 3D radiological volume is subjected to computationally expensive operations. Also, the subnetworks convolve 3D data at multiple scales by subjecting the 3D data to parallel processing by different 3D convolutional layer paths. Such multi-scale operations are computationally cheaper than the traditional CNNs that perform serial convolutions. In addition, performance of the subnetworks is further improved through 3D batch normalization (BN) that normalizes the 3D input fed to the subnetworks, which in turn increases learning rates of the 3D DCNNA. After several layers of 3D convolution and 3D sub-sampling with 3D across a series of subnetwork modules, a feature map with reduced vertical dimensionality is generated from the 3D radiological volume and fed into one or more fully connected layers.

    SYSTEM AND METHOD FOR PROVIDING PARKING NAVIGATION FOR A FIELD SERVICE VEHICLE

    公开(公告)号:US20190130764A1

    公开(公告)日:2019-05-02

    申请号:US15796399

    申请日:2017-10-27

    Inventor: VIJAY KARANI

    Abstract: A system and method for providing parking navigation for a field service vehicle has been developed. First, a destination and length of stay for the field service vehicle is determined. Next, a listing of available parking locations in proximity to the destination is retrieved with a parking navigation system located on-board the vehicle. A list of available parking locations for the field service vehicle is selected based on the estimated length of stay at the service destination, any physical requirements of the field service vehicle and efficient access to the destination. The selected parking location is transmitted to the driver of the field service vehicle with the parking navigation system.

    HIERARCHICAL AND INTERPRETABLE SKILL ACQUISITION IN MULTI-TASK REINFORCEMENT LEARNING

    公开(公告)号:US20190130312A1

    公开(公告)日:2019-05-02

    申请号:US15885727

    申请日:2018-01-31

    Abstract: The disclosed technology reveals a hierarchical policy network, for use by a software agent, to accomplish an objective that requires execution of multiple tasks. A terminal policy learned by training the agent on a terminal task set, serves as a base task set of the intermediate task set. An intermediate policy learned by training the agent on an intermediate task set serves as a base policy of the top policy. A top policy learned by training the agent on a top task set serves as a base task set of the top task set. The agent is configurable to accomplish the objective by traversal of the hierarchical policy network. A current task in a current task set is executed by executing a previously-learned task selected from a corresponding base task set governed by a corresponding base policy, or performing a primitive action selected from a library of primitive actions.

    Alerting system having a network of stateful transformation nodes

    公开(公告)号:US10275298B2

    公开(公告)日:2019-04-30

    申请号:US15291858

    申请日:2016-10-12

    Abstract: An alerting system is provided that includes a network of transformation nodes, and state change processors. The transformation nodes include input transformation nodes, output transformation nodes, and intermediate nodes that connect the input and output transformation nodes. Each input transformation node can receive an events stream, and is coupled to one of the output transformation nodes by one or more intermediate transformation nodes. Each transformation node (except the input transformation nodes) can receive state updates from those transformation nodes that it subscribes to. Each output transformation node can generate a check result when stored state information for each of the transformation nodes that the output transformation node subscribes to collectively indicates that the check result should be generated. Each output transformation node is coupled to one of the state change processors that can determine whether the check results should trigger an action, and if so, can then perform an action.

    Scheduling jobs for processing log files using a database system

    公开(公告)号:US10275281B2

    公开(公告)日:2019-04-30

    申请号:US15281559

    申请日:2016-09-30

    Abstract: Disclosed are some examples of systems, methods, apparatus and computer program products for scheduling jobs to process log files. In some implementations, a scheduler can be configured to identify a final state of a previously scheduled first job as a failure in execution by one or more processors. For instance, the first job may have been defined to process a first input log file saved to a database during a first timeframe of a sequence of timeframes of a designated interval. The scheduler can detect a second input log file saved to the database. The scheduler can then define a second job to process the first input log file and the second input log file and schedule the second job in association with a second timeframe of the sequence.

    Priority-driven boxcarring of action requests from component-driven cloud applications

    公开(公告)号:US10250716B2

    公开(公告)日:2019-04-02

    申请号:US15199768

    申请日:2016-06-30

    Abstract: Improved perceived load time for browser and mobile application pages is achieved by adjusting boxcarring of action requests from coupled data consuming applications on the user device, using the priority level of regions and components in component-driven cloud applications. Priority labels differentiate among display regions rendered by the data consuming application and the priority labels further differentiate among components within respective display regions. The middleware application batches the action requests into batches based at least in part on the priority labels, into boxcars segregated by priority label according to a predetermined segregation schedule, and dispatches the boxcars of batched action requests to the server. Performance is also dynamically speeded up, by adjusting inter-boxcar intervals used to dispatch batches of action requests from the user device to a production server, based on the dynamically measured network communication latency between the user device and the server.

    Matching objects using match rules and lookup key

    公开(公告)号:US10235476B2

    公开(公告)日:2019-03-19

    申请号:US14723866

    申请日:2015-05-28

    Abstract: Some embodiments of the present invention include a method for identifying match candidates in a database object and may include generating, by a database system, a match key associated with a lookup field of a database object. The method may further include activating, by the database system, a matching rule associated with the match key, and receiving, by the database system, a request to search for duplicate candidates in the database object. The request may include an input entity. The duplicate candidates in the database object may be identified by using the input entity and the matching rule.

    SYNCHRONIZE COLLABORATION ENTITY FILES
    680.
    发明申请

    公开(公告)号:US20190065521A1

    公开(公告)日:2019-02-28

    申请号:US16171179

    申请日:2018-10-25

    Abstract: A database system is arranged to support collaboration among users of a file in a feed system of a multi-user social network database driven system, wherein at least some of the users are permitted to upload a file to the feed from a remote computing device, and other users may have permission to collaborate on modifying the file. A sync client application executable on a remote user machine enables an authorize user to select “sync all files in the feed,” which triggers (a) download of the latest version of all files in the feed from the cloud to all users associated with the feed, and (b) subsequent automatic synchronization of the files by selectively uploading to the cloud database a modified local copy of the file from a user machine, and then downloading a latest version of the modified file to all of the users associated with the feed.

Patent Agency Ranking