Abstract: Embodiments are disclosed for expanding a seed scene using proposals from a generative model of scene graphs. The method may include clustering subgraphs according to respective one or more maximal connected subgraphs of a scene graph. The scene graph includes a plurality of nodes and edges. The method also includes generating a scene sequence for the scene graph based on the clustered subgraphs. A first machine learning model determines a predicted node in response to receiving the scene sequence. A second machine learning model determines a predicted edge in response to receiving the scene sequence and the predicted node. A scene graph is output according to the predicted node and the predicted edge.

Abstract: Social network postings, including text, images or other media, may provide valuable information regarding a user of the social network with which the postings may be associated. With the authorization of the user, and upon authentication by the social network, an online marketplace may access the social network postings and extract data therefrom, and market one or more recommended items to the user based on the extracted data, which may include color pallets or texture pallets derived from photographs included in the postings.

Abstract: A method for mitigating interference across analog signal lines includes receiving a digital data stream including a plurality of discrete signal patterns configured to drive a plurality of different analog signal lines. An edge buffer for each analog signal line is populated with edge data representing pulse edges of upcoming signal patterns set to drive the analog signal line. A target buffer for a target signal line is populated with target data representing a target signal pattern. Based at least in part on determining that edge buffers corresponding to one or more potentially interfering analog signal lines include edge data corresponding to post-target pulse edges, one or more potentially interfering signal patterns are identified. A selected set of the potentially interfering signal patterns are used to modify the target signal pattern to perform preemptive interference mitigation.

Abstract: A method for mitigating interference across analog signal lines includes receiving a digital data stream including a plurality of discrete signal patterns configured to drive a plurality of different analog signal lines. An edge buffer for each analog signal line is populated with edge data representing pulse edges of upcoming signal patterns set to drive the analog signal line. A target buffer for a target signal line is populated with target data representing a target signal pattern. Edge buffers corresponding to potentially interfering analog signal lines are searched to identify potentially interfering pulse edges. A set of potentially interfering pulse edges are selected for interference mitigation, and the target signal pattern is modified to perform preemptive interference mitigation based at least in part on the selected pulse edges.

Abstract: A method for mitigating interference across analog signal lines includes receiving a digital data stream including a plurality of discrete signal patterns configured to drive a plurality of different analog signal lines. An edge buffer for each analog signal line is populated with edge data representing pulse edges of upcoming signal patterns set to drive the analog signal line. A target buffer for a target signal line is populated with target data representing a target signal pattern. Edge buffers corresponding to potentially interfering analog signal lines are searched to identify potentially interfering pulse edges. A set of potentially interfering pulse edges are selected for interference mitigation, and the target signal pattern is modified to perform preemptive interference mitigation based at least in part on the selected pulse edges.

Abstract: A method for mitigating interference across analog signal lines includes receiving a digital data stream including a plurality of discrete signal patterns configured to drive a plurality of different analog signal lines. An edge buffer for each analog signal line is populated with edge data representing pulse edges of upcoming signal patterns set to drive the analog signal line. A target buffer for a target signal line is populated with target data representing a target signal pattern. Based at least in part on determining that edge buffers corresponding to one or more potentially interfering analog signal lines include edge data corresponding to post-target pulse edges, one or more potentially interfering signal patterns are identified. A selected set of the potentially interfering signal patterns are used to modify the target signal pattern to perform preemptive interference mitigation.

Abstract: Systems and methods are disclosed for triggering an update to a machine-learning model upon detecting that a distribution of particular (e.g., recently collected) input data set is sufficiently different from a distribution training input data set used to train the model. The distributions may be determined to be sufficiently different when a classifier can identify to which distribution individual data elements belong (e.g., to at least a predetermined degree). An update to the machine-learning model can include morphing weights used by the model and/or retraining the model.

Abstract: Systems and methods are disclosed for triggering an update to a machine-learning model upon detecting that a distribution of particular (e.g., recently collected) input data set is sufficiently different from a distribution training input data set used to train the model. The distributions may be determined to be sufficiently different when a classifier can identify to which distribution individual data elements belong (e.g., to at least a predetermined degree). An update to the machine-learning model can include morphing weights used by the model and/or retraining the model.

Abstract: A system, device, and method for managing connections in an industrial installation are described. The system includes one or more field devices, one or more automation devices, and a self-configurable device. The self-configurable device is adapted to dynamically configure, based on type of the one or more field devices and the one or more automation devices, such that the self-configurable device manages a connection between the one or more field devices and the one or more automation devices. The self-configurable device is adapted to calibrate one or more field devices and manage automation functions in the industrial installation.

Abstract: Systems and methods are disclosed for triggering an update to a machine-learning model upon detecting that a distribution of particular (e.g., recently collected) input data set is sufficiently different from a distribution training input data set used to train the model. The distributions may be determined to be sufficiently different when a classifier can identify to which distribution individual data elements belong (e.g., to at least a predetermined degree). An update to the machine-learning model can include morphing weights used by the model and/or retraining the model.

Abstract: Embodiments are disclosed for a method of providing a user interface on a computing device. The method includes presenting a virtual keyboard on a display of the computing device, detecting input to the virtual keyboard. The method further includes, for each detected input, determining whether the input selects any of one or more delimiter keys, displaying a placeholder for the input responsive to the input not selecting any of the one or more delimiter keys, and receiving suggested candidate text from a word-level recognizer and replacing all currently displayed placeholders with the suggested candidate text responsive to the input selecting any of the one or more delimiter keys.

Abstract: A video segmentation system can be utilized to automate segmentation of digital video content. Features corresponding to visual, audio, and/or textual content of the video can be extracted from frames of the video. The extracted features of adjacent frames are compared according to a similarity measure to determine boundaries of a first set of shots or video segments distinguished by abrupt transitions. The first set of shots is analyzed according to certain heuristics to recognize a second set of shots distinguished by gradual transitions. Key frames can be extracted from the first and second set of shots, and the key frames can be used by the video segmentation system to group the first and second set of shots by scene. Additional processing can be performed to associate metadata, such as names of actors or titles of songs, with the detected scenes.

Abstract: Techniques are described for providing functionality to allow a viewer of a television show to watch a “previously on” segment of an episode of the television show and be able to watch the scenes from prior episodes referenced in the “previously on” segment.

Abstract: Embodiments are disclosed for a method of providing a user interface on a computing device. The method includes presenting a virtual keyboard on a display of the computing device, detecting input to the virtual keyboard. The method further includes, for each detected input, determining whether the input selects any of one or more delimiter keys, displaying a placeholder for the input responsive to the input not selecting any of the one or more delimiter keys, and receiving suggested candidate text from a word-level recognizer and replacing all currently displayed placeholders with the suggested candidate text responsive to the input selecting any of the one or more delimiter keys.

Abstract: Social network postings, including text, images or other media, may provide valuable information regarding a user of the social network with which the postings may be associated. With the authorization of the user, and upon authentication by the social network, an online marketplace may access the social network postings and extract data therefrom, and market one or more recommended items to the user based on the extracted data, which may include color pallets or texture pallets derived from photographs included in the postings.

Abstract: Embodiments are disclosed for a method of providing a user interface on a computing device. The method includes presenting a virtual keyboard on a display of the computing device, detecting input to the virtual keyboard. The method further includes, for each detected input, determining whether the input selects any of one or more delimiter keys, displaying a placeholder for the input responsive to the input not selecting any of the one or more delimiter keys, and receiving suggested candidate text from a word-level recognizer and replacing all currently displayed placeholders with the suggested candidate text responsive to the input selecting any of the one or more delimiter keys.

Abstract: In one embodiment, a method includes by one or more computing machines: receiving a search query from a user, generating a normalized query based on the search query, identifying multiple objects matching the search query, and calculating an engagement score for each object. The engagement score is based on real-time counters and batch counters. Each counter includes: a key listing the object, normalized query, and one of multiple types of user interactions; and a value indicating a number of user interactions with the object performed in response to search queries normalizing to the normalized query. The value for real-time and batch counters indicates user interactions during first and second threshold windows of time, respectively. The method continues with sending, to the user, a search-results interface including results corresponding to the identified objects above a threshold engagement score, and updating the real-time counters based on user interactions with the search results.

Abstract: Disclosed are various embodiments for relating items to objects detected in media. An item recognition application obtains media information associated with a broadcast being consumed form the client device. The item recognition application identifies objects in the media based on user input of the location of the object on the client device display and/or heuristic detection of the object. The user may indicate the object's location through an input device. The object may be heuristically detected based on an algorithm that differentiates the object from the background. The item recognition application identifies items related to the object. The item recognition application then performs an action with the item.

Abstract: An electronic device includes a display including a touch sensor, a contextual sensor configured to provide data indicative of an operating condition of the electronic device, one or more memories in which touch input instructions, touch event mapping instructions, and a touch event mapping model are stored, and a processor coupled to the memory. The processor is configured through execution of the touch input instructions to obtain data indicative of an initial assessment of touch event position via the touch sensor. The touch event mapping model maps the initial assessment of touch event position to an adjusted touch event position as a function of the data indicative of the operating condition of the electronic device.

Abstract: A video segmentation system can be utilized to automate segmentation of digital video content. Features corresponding to visual, audio, and/or textual content of the video can be extracted from frames of the video. The extracted features of adjacent frames are compared according to a similarity measure to determine boundaries of a first set of shots or video segments distinguished by abrupt transitions. The first set of shots is analyzed according to certain heuristics to recognize a second set of shots distinguished by gradual transitions. Key frames can be extracted from the first and second set of shots, and the key frames can be used by the video segmentation system to group the first and second set of shots by scene. Additional processing can be performed to associate metadata, such as names of actors or titles of songs, with the detected scenes.