Abstract: Obtain, from an existing machine learning classifier, original probabilistic scores classifying samples taken from two or more groups into two or more classes via supervised machine learning. Associate the original probabilistic scores with a plurality of original Lagrange multipliers. Adjust values of the plurality of original Lagrange multipliers via low-dimensional convex optimization to obtain updated Lagrange multipliers that satisfy fairness constraints as compared to the original Lagrange multipliers. Based on the updated Lagrange multipliers, closed-form transform the original probabilistic scores into transformed probabilistic scores that satisfy the fairness constraints while minimizing loss in utility. The fairness constraints are with respect to the two or more groups.

Abstract: A technique for computationally-efficient privacy-preserving homomorphic inferencing against a decision tree. Inferencing is carried out by a server against encrypted data points provided by a client. Fully homomorphic computation is enabled with respect to the decision tree by intelligently configuring the tree and the real number-valued features that are applied to the tree. To that end, and to the extent the decision tree is unbalanced, the server first balances the tree. A cryptographic packing scheme is then applied to the balanced decision tree and, in particular, to one or more entries in at least one of: an encrypted feature set, and a threshold data set, that are to be used during the decision tree evaluation process. Upon receipt of an encrypted data point, homomorphic inferencing on the configured decision tree is performed using a highly-accurate approximation comparator, which implements a “soft” membership recursive computation on real numbers, all in an oblivious manner.

Abstract: Aspects of the disclosure relate to using machine learning techniques for dynamic occupancy prediction. A computing platform may receive first data that is non-personalized and is associated with a first specific physical space, and may receive second data associated with a second specific physical space. The computing platform may normalize, for a common subspace of the first specific physical space and the second specific physical space, the first data and the second data. Using the normalized data, the computing platform may generate a predicted occupancy value for the common subspace. The computing platform may send commands directing a local HVAC control system, deployed at the common subspace, to perform a resource control action for the common subspace, which may cause the local HVAC control system to perform the resource control action for the common subspace.

Abstract: Embodiments of the present invention relate to apparatuses, systems, methods and computer program products for dynamic network database integration. Specifically, the system typically provides dynamic correlation, transformation and combination of data stored at a plurality of database locations. In some aspects, the system receives a retrieval request for retrieving one or more records. The system then typically retrieves, transforms and dynamically integrates unstructured data stored at disparate database locations to construct one or more structured records. The system may also construct the one or more structured records such that the one or more structured records are compatible with the user interface of the user device application API. The system may also be configured to facilitate dynamic modification of the retrieval request, in run-time, using custom HTTP headers.

Abstract: A system for scanning a gas mixture using a sensor is disclosed. The system includes a matrix multiplication module configured to pre-multiply a B matrix with a diagonal matrix that is created with a vector of nominal concentration of the set of gases to obtain an adjusted B matrix (Ba) and a mass to charge ratio extraction module that is configured to select a set of mass to charge ratios for the set of gases to scan in a time budget based on the adjusted B matrix (Ba). The B matrix is a multiplication of P, T, C and R matrices, wherein P is a convolution matrix representing peak shapes, T is transmission efficiencies at each integral mass to charge ratio, R is relative ionization potentials for each gas and C is a reference spectrum representing idealized responses for each gas at the integral mass to charge ratio.

Abstract: A process and a system for creating a visual guide for developing training data for a classification of image, where the training data includes images tagged with labels for the classification of the images. A processor may prompt a user to define a framework for the classification. For an initial set of images within the training data, qualified human classifiers are prompted to locate the images within the framework and to tag the images with labels. The processor determines whether the tagged images have consistent labels, and, if so, the processor adds images to the training data. The processor may add the images by providing a visual guide, the visual guide including tagged images arranged according to their locations within the framework their labels, and prompting human classifiers to tag the additional images with labels for the classification, according to the visual guide.

Abstract: Disclosed are various examples for facilitating enrollment of a client device into more than one management framework. A client device can be enrolled with a management service as a fully managed device. The client device can also be enrolled with the management service as a personal or bring-your-own-device (BYOD), which causes a workspace to be created on the device that is segregated from the rest of the client device. Both enrollments can be managed by a remotely executed management service.

Abstract: An overvoltage protection circuit includes an input terminal, an output terminal, a clamp transistor, and a selector circuit. The clamp transistor is configured to control current flow between the input terminal and the output terminal. The clamp transistor includes a first terminal coupled to the input terminal, a second terminal coupled to the output terminal. The selector circuit is configured to control a resistance of the clamp transistor based on a voltage at the input terminal. The selector circuit includes a first terminal coupled to the first terminal of the clamp transistor, a second terminal coupled to the second terminal of the clamp transistor, and a third terminal coupled to a third terminal of the clamp transistor.

Abstract: A method and structure for operating a combustion system of a gas turbine engine to mitigate low frequency combustion acoustics is generally provided. The method includes flowing an oxidizer through a fuel nozzle passage defining an inner wall and an outer wall, in which each of the inner wall and the outer wall are contoured from a first radius to a second radius smaller than the first radius; flowing the oxidizer at a higher axial velocity at the inner wall relative to the outer wall upstream of a fuel injection port; flowing a fuel through the fuel injection port to the fuel nozzle passage to mix with the flow of oxidizer to produce a fuel-oxidizer mixture; and igniting the fuel-oxidizer mixture downstream of the fuel injection port.

Abstract: One embodiment provides for a non-transitory machine readable medium storing instructions which, when executed by one or more processors, cause the one or more processors to perform operations comprising providing an interface to define a neural network using machine-learning domain specific terminology, wherein the interface enables selection of a neural network topology and abstracts low-level communication details of distributed training of the neural network.

Abstract: Examples described herein include systems and methods for contextually providing automated device enrollment into a management system. A management application on a user device can receive network settings for connecting to a local server. The network settings can be preconfigured by an administrator. The management application can cause the user device to send an enrollment request and a device identifier to the local server. The device identifier can be used to validate the device and provide a security token to the management application. The management application can use the security token to complete enrollment of the user device.

Abstract: Method and system for generating a coded image format representing surface data of a 3-dimensional object is disclosed. The method includes superposing an imaginary spherical surface over the object based on a reference point within the object. The surface points of the front portion and back portion of the object are mapped to a first circular region and a second circular region. The distance information of the surface points of the front portion and the back portion from the reference point are encoded as greyscale values to obtain shape information. The color information of the surface points of the front portion and back portion of object are encoded as color values to obtain color information of the front portion and the back portion of the object. The encoded values are stored in pixels of raster graphics image. A coded image format is generated based on the raster graphics image.

Abstract: A threat protection system provides for detecting links in a document and analyzing whether one of the detected links is a malicious link that may direct a user of the document to a malicious universal resource locator (URL). In one implementation of the described technology, when a user selects a link in a document, a link activation module calls a threat protection client module that performs a reputation check for the link. If the selected link is malicious, the threat protection client module sends a URL of a warning page to the link activation module.

Abstract: Systems and methods for localizing and tracking mobile objects are provided. A method includes determining an initial location of a node in a multi-hop network based on multi-lateration from an unmanned aerial vehicle. The method also includes applying an adaptive aperture to address a non-uniform velocity of the node based on the turn and a velocity vector. A determination whether localization for the node can be implemented using first hop nodes in the multi-hop network is made. In response to a determination that localization cannot be implemented using the first hop nodes, inertial sensor measurements associated with the node are accessed and the inertial sensor measurements are integrated with the adaptive aperture to improve localization accuracy.

Abstract: A catheter system may include a luer adapter, which may include an outer surface having threading or a recess. The catheter system may also include a flow control plug, which may include a proximal end and a distal end. The proximal end of the flow control plug may include a filter element permeable to air and not to blood. The distal end of the flow control plug may include a cylinder and a taper-shaped luer tip spaced apart from the cylinder. An inner surface of the cylinder may include a protrusion engaged in a snap-fit with the recess or corresponding threading mated with the threading.

Abstract: A computer-implemented method, system, and non-transitory computer-readable storage medium for enhancing performance of a first model. The first model is trained with a training data set. A second model receives the training data set associated with the first model. The second model provides the first model with a hardness value associated with prediction of each data point of the training data set. The first model determines a confidence value regarding predicting each data point based on the training data set, and determines a ratio of the hardness value of a prediction of each data point by the second model with respect to the confidence value of the first model. The first model is retrained with a re-weighted training data set when the determined ratio is lower than a value of ?.

Abstract: A method, apparatus and computer program product for homomorphic inference on a decision tree (DT) model. In lieu of HE-based inferencing on the decision tree, the inferencing instead is performed on a neural network (NN), which acts as a surrogate. To this end, the neural network is trained to learn DT decision boundaries, preferably without using the original DT model data training points. During training, a random data set is applied to the DT, and expected outputs are recorded. This random data set and the expected outputs are then used to train the neural network such that the outputs of the neural network match the outputs expected from applying the original data set to the DT. Preferably, the neural network has low depth, just a few layers. HE-based inferencing on the decision tree is done using HE inferencing on the shallow neural network. The latter is computationally-efficient and is carried without the need for bootstrapping.

Abstract: Apparatuses, methods, and systems are disclosed for feedback request determination. One method (1100) includes determining (1102), at a first sidelink device, whether to request feedback corresponding to communications with a second sidelink device. The method (1100) includes, in response to determining to request feedback corresponding to the communications with the second sidelink device, transmitting (1104) first information to the second sidelink device indicating a request for feedback. The method (1100) includes transmitting (1106) second information to the second sidelink device indicating resources for communicating the feedback.

Abstract: An information handling system stylus performs touch inputs at a touchscreen display with an active writing tip that receives a capacitive signal from a processing resource, such as an ASIC. A tip position controller selectively retracts and extends the writing tip at an opening formed in a tip cap disposed at the end of the stylus. A power interface cuts off power to the processing resource when the writing tip retracts and applies power to the processing resource when the writing tip extends.

Abstract: Systems and methods are disclosed for a smart mouse that may include transmitting, by the mechanical switch of the smart mouse, a signal to a microcontroller unit of the smart mouse in response to a movement of the smart mouse caused by a user, the movement causing the mechanical switch to be in a closed position, the signal indicating that the mechanical switch is in the closed position; receiving, by the microcontroller unit, the signal indicating that the mechanical switch is in the closed position; and in response to receiving the signal: causing, by a power switch of the smart mouse, at least one battery of the smart mouse to provide power to an optical unit of the smart mouse, the provided power causing the optical unit to be in an on state, the on state of the optical unit causing the smart mouse to be electrically enabled for use.