Abstract: Methods, systems, and computer-readable media for offline evaluation of machine learning models with noise reduction are disclosed. A trigger computation system generates a plurality of experimental requests. The experimental requests do not represent unmodified requests received from clients in a production environment. At least one parameter value varies for individual ones of the experimental requests. The trigger computation system provides the experimental requests to a first machine learning model and a second machine learning model. The first machine learning model and the second machine learning model produce a set of results based at least in part on the experimental requests. The trigger computation system determines a reduced set of results for which the first machine learning model and the second machine learning model differ. An evaluation of the first machine learning model or the second machine learning model is performed using the reduced set of results.

Abstract: This disclosure relates generally to real-time path planning. Planning amidst obstacles in a cluttered indoor environment is a difficult task for a robotic agent. The disclosed method provides semidefinite programming induced free-space based path planning. Free-space is generated by an efficient environment grid resolution independent seeding technique. In the proposed resolution independent seeding technique, initial position of the robotic agent is considered as the first seed. For subsequent seeding, information of the expanded earlier seeds are employed intelligently. This process is followed unto a finite sequence, which naturally results in a contiguous navigable convex free-space. This contiguous navigable convex free-space is employed to create an undirected graph, which is then used for path planning. Path planning is done locally by evaluating the subgoal with respect to a final goal. Local planning cumulatively assists the planner to attain the final goal.

Abstract: Embodiments of the present disclosure provide systems and methods to eliminate (or filter) drift for dynamics model based localization of multirotors. The dynamics equations require drag modelling, which is dependent on velocity, to generate vehicles' acceleration along the body axis. The present disclosure considers the drag contribution, at velocity level, as a low frequency component. Incorrect or nonmodelling of this low frequency component leads to drift at velocity level. This drift can then be removed through a high pass filter to obtain drift free velocity data for pose estimation and better localization thereof.

Abstract: State of art systems in the domain of dynamism detection fail to estimate noise if depth images are being collected as input for the dynamism detection, as the noise in depth images depend on the scene being captured. Disclosed herein are method and system for determining dynamism by processing depth image of a scene. The system models depth sensor noise as ergodic stochastic process by determining that distribution estimated at each reference pixel from a plurality of neighborhood pixels in a reference image being processed is statistically same as a distribution estimated from evolution of the reference pixel over the time. After modeling the depth sensor noise in this manner, the same is eliminated/removed from the reference image, which is then processed to estimate divergence at each pixel based on temporal and spatial distribution built at pixel level in the reference image, and in turn determines dynamism in the scene.

Abstract: This disclosure relates generally to real-time path planning. Planning amidst obstacles in a cluttered indoor environment is a difficult task for a robotic agent. The disclosed method provides semidefinite programming induced free-space based path planning. Free-space is generated by an efficient environment grid resolution independent seeding technique. In the proposed resolution independent seeding technique, initial position of the robotic agent is considered as the first seed. For subsequent seeding, information of the expanded earlier seeds are employed intelligently. This process is followed unto a finite sequence, which naturally results in a contiguous navigable convex free-space. This contiguous navigable convex free-space is employed to create an undirected graph, which is then used for path planning. Path planning is done locally by evaluating the subgoal with respect to a final goal. Local planning cumulatively assists the planner to attain the final goal.

Abstract: State of art systems in the domain of dynamism detection fail to estimate noise if depth images are being collected as input for the dynamism detection, as the noise in depth images depend on the scene being captured. Disclosed herein are method and system for determining dynamism by processing depth image of a scene. The system models depth sensor noise as ergodic stochastic process by determining that distribution estimated at each reference pixel from a plurality of neighborhood pixels in a reference image being processed is statistically same as a distribution estimated from evolution of the reference pixel over the time. After modeling the depth sensor noise in this manner, the same is eliminated/removed from the reference image, which is then processed to estimate divergence at each pixel based on temporal and spatial distribution built at pixel level in the reference image, and in turn determines dynamism in the scene.

Abstract: Robotic vision-based framework wherein an on-board camera device is used for scale estimation. Unlike conventional scale estimation methods that require inputs from more than one or more sensors, implementations include a system and method to estimate scale online solely, without any other sensor, for monocular SLAM by using multirotor dynamics model in an extended Kalman filter framework. This approach improves over convention scale estimation methods which require information from some other sensors or knowledge of physical dimension of an object within the camera view. An arbitrary scaled position and an Euler angle of a multirotor are estimated from vision SLAM (simultaneous localization and mapping) technique. Further, dynamically integrating, computed acceleration to estimate a metric position. A scale factor and a parameter associated with the multirotor dynamics model is obtained by comparing the estimated metric position with the estimated arbitrary position.

Abstract: Robotic vision-based framework wherein an on-board camera device is used for scale estimation. Unlike conventional scale estimation methods that require inputs from more than one or more sensors, implementations include a system and method to estimate scale online solely, without any other sensor, for monocular SLAM by using multirotor dynamics model in an extended Kalman filter framework. This approach improves over convention scale estimation methods which require information from some other sensors or knowledge of physical dimension of an object within the camera view. An arbitrary scaled position and an Euler angle of a multirotor are estimated from vision SLAM (simultaneous localization and mapping) technique. Further, dynamically integrating, computed acceleration to estimate a metric position. A scale factor and a parameter associated with the multirotor dynamics model is obtained by comparing the estimated metric position with the estimated arbitrary position.

Abstract: Conventional techniques involve fusion of Inertial Measurement Units (IMU) sensor based method and vision based localization technique for localization of rotor systems. However vision based localization technique may be prone to errors due to motion blur, drastic lighting change, sudden rotation at UAV, and the like, while the drift in IMU based localization severely impact overall solution. Embodiments of the present disclosure provide systems and methods to eliminate (or filter) drift for dynamics model based localization of multirotors. The dynamics equations require drag modelling, which is dependent on velocity, to generate vehicles' acceleration along the body axis. The present disclosure considers the drag contribution, at velocity level, as a low frequency component. Incorrect or nonmodelling of this low frequency component leads to drift at velocity level. This drift can then be removed through a high pass filter to obtain drift free velocity data for pose estimation and better localization thereof.

Abstract: Reductions in churn for assigning identifiers to entities in a knowledge graph enables several improvements to the functionality of the computing devices maintaining or accessing knowledge graphs. As the aliases or other terms used to identify a given entity change in response to updates to the knowledge graph, the identifiers assigned to various entities may change. For example, two individual entities conflated as one node may be split into two or two nodes merged into one in response to an update, and existing identifiers should be used to reduce churn. To select the existing identifiers to assign to a given updated entity, the aliases are clustered with the updated entities and the unique modal prior identifier is assigned as the identifier for the updated entity. Higher orders of modality are used to ensure as many existing identifiers are used before creating new identifiers.

Abstract: Reductions in churn for assigning identifiers to entities in a knowledge graph enables several improvements to the functionality of the computing devices maintaining or accessing knowledge graphs. As the aliases or other terms used to identify a given entity change in response to updates to the knowledge graph, the identifiers assigned to various entities may change. For example, two individual entities conflated as one node may be split into two or two nodes merged into one in response to an update, and existing identifiers should be used to reduce churn. To select the existing identifiers to assign to a given updated entity, the aliases are clustered with the updated entities and the unique modal prior identifier is assigned as the identifier for the updated entity. Higher orders of modality are used to ensure as many existing identifiers are used before creating new identifiers.

Abstract: Web pages that are known to be associated with entities, such as authors, are selected. Documents or other publications that are linked to or referenced by each web page are determined. Based on the authors of each determined document, the authors associated with each web page, and other information such as institutions or venues identified in each document, the various authors associated with the web pages are conflated or disambiguated to determine which authors, while having the same or similar names, should be treated as separate entities, and which authors, while having different names, should be treated as the same entities. Once the entity names have been conflated and disambiguated, they can be linked to social networking data or grant data associated with entities.

Abstract: The present subject matter discloses systems and methods for patentability determination of an inventive idea in an organization. In one implementation, the method for patentability determination in the organization comprises receiving high-level functionalities and granular-level functionalities of the inventive idea at a patentability determination system from an end user. Next, the end user provides importance-level inputs for each of the granular-level functionalities at the patentability determination system. The method further comprises receiving matching-level inputs for each of the granular-level functionalities against prior art documents at the patentability determination system from a plurality of evaluators. Based on the importance-level inputs and the matching-level inputs, an overall patentability index value of the granular-level functionalities is calculated.

Abstract: An XML template having one or more nodes is received. Mapping information indicating an association of data and nodes of the uploaded XML template is obtained. Once the mapping is received, the structure of the XML template is determined. Based on the determined structure and the mapping provided, an XML based SQL query is generated. The generated SQL query can be executed to provide the XML document.

Abstract: An XML template having one or more nodes is received. Mapping information indicating an association of data and nodes of the uploaded XML template is obtained. Once the mapping is received, the structure of the XML template is determined. Based on the determined structure and the mapping provided, an XML based SQL query is generated. The generated SQL query can be executed to provide the XML document.

Abstract: A resonator system is disclosed for use in illuminating an input to a diffractive element. The system includes a source of an electromagnetic field having a wavelength of &lgr;, and first and second optical elements, each of which is at least partially reflecting. The first and second optical elements are separated from one another such that the optical path between the optical elements has a distance ( 2 ⁢ m + 1 ) ⁢ λ 4 , wherein m is an integer.

Abstract: A resonator system is disclosed for use in illuminating an input to a diffractive element. The system includes a source of an electromagnetic field having a wavelength of &lgr;, and first and second optical elements, each of which is at least partially reflecting.

Abstract: A resonator system is disclosed for use in illuminating a diffractive element. The system includes a source of an electromagnetic field having a wavelength of &lgr;, and first and second optical elements, each of which is at least partially reflecting.