Abstract: The disclosure relates to method and system for on-device inference in a deep neural network (DNN). The method comprises: determining whether one or more layers of the DNN satisfy one of a first, a second and a third condition, the one or more layers including one or more convolution layers and one or more resampling layers; performing the on-device inference based on the determination, wherein performing the on-device inference comprises at least one of: optimizing the one or more convolution layers in the one or more parallel branches based on the one or more layers of the DNN satisfying the first condition, optimizing the at least one of the resampling layers based on the one or more layers of the DNN satisfying the second condition, and modifying operation of the at least one of the resampling layers based on the one or more layers of the DNN satisfying the third condition.

Abstract: A hyperspectral stereoscopic CubeSat with computer vision and artificial intelligence capabilities consists of a device and a data processing methodology. The device comprises a number of VIS-NIR-TIR hyperspectral sensors, a central processor with memory, a supervisor system running independently of the imager system, radios, a solar panel and battery system, and an active attitude control system. The device is launched into low earth orbit to capture, process, and transmit stereoscopic hyperspectral imagery in the visible and infrared portions of the electromagnetic spectrum. The processing methodology therein comprises computer vision and convolutional neural network algorithms to perform spectral feature identification and data transformations.

Abstract: An example operation may include one or more of identifying application instances in a cluster group, receiving a request to perform a consensus decision regarding ownership of the cluster group via peer nodes of a blockchain, identifying individual consensus decision votes from the peer nodes, and performing the consensus decision based on the individual consensus decision votes.

Abstract: A first backend housekeeping operation of a backend storage system is detected by a storage virtualization system. The first backend housekeeping operation is related to a first backend storage portion of the backend storage system. The storage virtualization system transfers data to the backend storage system for storage and retrieval. The storage virtualization system issues commands to the backend storage system. A first virtualized operation related to a first virtualized storage portion of the storage virtualization system is identified. A storage overhead condition is determined based on the first backend storage portion and based on the first virtualized operation. A storage corrective action is performed based on the storage overhead condition. The storage corrective action prevents the potential performance of one or more additional backend housekeeping operations.

Abstract: The present disclosure refers to methods and apparatuses for processing of high-resolution video content. In an embodiment, a method includes generating a first group of video frames from the video content. The first group of video frames has a first resolution lower than a resolution of the video content and a first rate-distortion score. The method further includes generating a second group of video frames from the video content. The second group of video frames has a second resolution lower than the resolution of the video content and a second rate-distortion score. The method further includes selecting an optimal group of video frames from the first and second groups of video frames based on a comparison between the first and second rate-distortion scores. The optimal group of video frames has a rate-distortion score lower than the first and the second rate-distortion scores.

Abstract: A method includes: receiving, by a computing device, a digital voice stream; receiving, by the computing device, converted text that represents the digital voice stream; identifying, by the computing device, an erroneously converted portion of the converted text; selecting, by the computing device, the erroneously converted portion for explainability processing; parsing, by the computing device, the erroneously converted portion into parts based on a predetermined parsing level; collecting, by the computing device, supplementary input data related to the erroneously converted portion; and determining, by the computing device and based on the supplemental input data, a reason why the erroneously converted portion was erroneously converted.

Abstract: Embodiments of the present invention provide methods, computer program products, and systems. Embodiments of the present invention can allocate a dedicated traffic channel to a plurality of connected devices. Embodiments of the present invention can dynamically determine an appropriate visualization for one or more connected devices of the plurality of connected devices based on received information. Embodiments of the present invention render the dynamically determined appropriate visualization for the one or more connected devices of the plurality of connected devices using the dedicated traffic channel.

Abstract: Disclosed is a computer implemented method to manage queue overlap in storage systems, the method comprising, identifying, by a storage system, a plurality of queues including a first queue and a second queue. The storage system includes a plurality of cores, including a first core and a second core, and wherein the first queue is associated with a first host and the second queue is associated with a second host. The method also comprises, determining the first queue and the second queue are being processed by the first core. The method further comprises, monitoring the workload of each cores and identifying a load imbalance, wherein the loam imbalance a difference between a first workload associated with the first core, and a second workload associated with the second core. The method also comprises, notifying the second host that the load imbalance is present.

Abstract: Disclosed is an IoT enabled, portable, multiple stimuli reinforced non-nutritive Sucking Training System, Method and Device that implements a function that measures suction pressure of a non-nutritive sucking event produced by a pre-mature infant; then converts the measured pressure value using transducer to digital form; then transmits the pressure value in digital form using IoT (Internet-of-Things) controller to a server and/or to an app and/or a software installed on a smartphone device. The app and/or software installed on the smartphone device which receives the pressure measurement uses an algorithm to produce a stimulus such as music and/or haptic feedback for the pre-mature infant, to train the premature infant to suck optimally; thereby improving self-feeding capability of the premature infant.

Abstract: An approach is provided in which the approach detects a set of personal information data corresponding to a set of users in a set of training data. The approach transforms the set of training data into a set of semantically equivalent training data by replacing the set of personal information with a set of semantic equivalent data. The approach then trains a machine learning model using the set of semantically equivalent training data.

Abstract: The invention relates to Quench box assembly comprising quench pipe and quench box, to mix quench gas and vapor-liquid effluent from previous catalyst bed to achieve equilibrium temperature before entering the next bed. The quench pipe is in the form of ring having aperture while quench box consists of swirling section and a mixing chamber. The swirling section consists of inclined baffles to provide swirling action to incoming stream and the turbulence created by the swirling action increases the heat transfer rate thus requiring the smaller reactor volume to attain equilibrium temperature. The perforated plate being open from all the sides allowing the liquid to flow uniformly from all directions thus providing uniform distribution on the distributor tray. Hence, eliminates the requirement of rough liquid distributor before the distribution tray.

Abstract: A network device may receive a first data packet. The network device may determine that a level of available computing resources satisfies a threshold level. The network device may perform a secure socket layer (SSL) proxy function based on the level of available computing resources satisfying the threshold level. The network device may receive a second data packet. The network device may determine that the level of available computing resources fails to satisfy the threshold level. The network device may determine a security characteristic associated with the second data packet. The network device may determine a security rating associated with the second data packet based on the security characteristic. The network device may selectively perform the SSL proxy function based on the security rating.

Abstract: A processor may identify that a user is interacting with a device, where the interacting is identified from the user touching the device. The processor may receive tactile data associated with the user from one or more tactile sensors. The processor may extract, utilizing an AI model, one or more features of the tactile data. The processor may classify, utilizing the AI model, the tactile data as having a tactile data characteristic. The processor may output the classification to an interaction management module.

Abstract: A storage system includes a controller and block-based storage device(s) that include LSA blocks arranged as a log structured array (LSA). The controller creates a first and second LUN, each including LBA locations. The controller assigns the first LUN to a first host and the second LUN to a second host, accumulates first host data associated with a first LBA location of the first LUN, writes a block-size worth of such data to a first LSA block, and maps the first LBA location to the first LSA block. In response to receiving an ODX Copy Offload instruction, the controller determines the first host data should be migrated to a target LBA location in the second LUN, determines the first and second LUN are exclusively mappable to the same LSA, maps the target LBA location to the first LSA block, and unmaps the first LBA location from the first LSA block.

Abstract: A device may receive work item data identifying work items associated with requirements from different tools of a project and may perform data cleansing to remove and/or modify particular words from the work item data and to generate cleansed work item data. The device may perform natural language processing on the cleansed work item data to identify synonyms for words in the cleansed work item data and may replace abbreviations in the cleansed work item data with full form text to generate final work item data. The device may identify keywords in the final work item data and may process the final work item data, the synonyms, and the keywords, with a machine learning model, to identify mappings between work items of the final work item data and to determine a confidence score for the mappings. The device may perform actions based on the mappings and the confidence score.

Abstract: Embodiments of the present invention provide a computer system, a computer program product, and a method that comprises collecting data capable of being replicated from a computing device; detecting risks of the computing device, wherein detecting risks comprises detecting the computing device's surroundings, location, speed, and condition; initiating data replication on the computing device once the risks are determined to reach a predetermined threshold; and storing the replicated data within a cloud storage system using a 5G network.

Abstract: A computer-implemented method is used for each of two or cameras on a device, where a first camera of the two or cameras is located under a display of the device. The method further includes defining, for each camera, a best effect region on the display including a first best effect region for the first camera. The method also includes identifying, for each of one or more feeds, a feed display area including a first feed display area, where each feed display location uses a portion of the display. The method includes activating the first camera for a first feed, wherein the first feed has a first feed display area that overlaps the first best effect region.

Abstract: A computer-implemented method, system and computer program product for performing real-time monitoring of machine learning models. Real-time model state data and metadata (e.g., operating dataset) of the machine learning models located within an orchestration plane of a network are collected by agents located within the machine learning models. The portion of the collected real-time model state data and metadata that is to be provided to the user by the service orchestrator (configured to monitor the machine learning models in the service orchestration plane via the use of agents in the machine learning models) is selected and marked. The marked collected real-time model state data and metadata are then provided to the user by the service orchestrator. In this manner, real-time monitoring of the machine learning models in the orchestration plane, such as the service orchestration plane, of a broadband cellular network (e.g., fifth generation broadband cellular network) is achieved.

Abstract: Call propensity source data may be received that include a first percentage of call propensity source data that correspond to presence of post-visit phone calls to a customer service of an entity after some customer visits to a web site of an entity and a second percentage of call propensity source data that correspond to absence of post-visit phone calls to the customer service after other customer visits to the website. A machine-learning model is trained based on a plurality of features in at least a portion of the call propensity source data to generate a trained machine-learning model. The trained machine-learning model is applied to multiple features included in at least one of corresponding website activity data and corresponding activity error data of a customer to generate a probability score that measures a likelihood of the customer calling the customer service regarding an issue that is unresolved via the website.

Abstract: A computer-implemented system and method for a transportation system comprises, using a processor associated with a service, initializing information collectors and response monitors by loading configurations, user settings, and data structures to capture device information from a plurality of devices that each run a virtual agent (VA). Fifth generation (5G) communication links are established between 5G server orchestration service instances (SOSIs) and respective VAs on the devices. Captured live status information by the SOSIs is received from various devices. A 5G dynamic ad-hoc network (DAHN) connects a user device of a user having visual impairment and a vehicle stop device, the DAHN creation being triggered by the user device entering a stop device boundary. An SOSI receives user-vehicle stop information after the connecting to the DAHN. Control information is transmitted to the vehicle device VA related to the user when the user device is located within the stop boundary.