Digital Commons Network^™

Transfer Entropy In Graph Convolutional Neural Networks

Defensive Publications Series

Graph Convolutional Networks (GCN) are Graph Neural Networks where the convolutions are applied over a graph. In contrast to Convolutional Neural Networks, GCN’s are designed to perform inference on graphs, where the number of nodes can vary, and the nodes are unordered. In this study, we address two important challenges related to GCNs: i) oversmoothing; and ii) the utilization of node relational properties (i.e., heterophily and homophily). Oversmoothing is the degradation of the discriminative capacity of nodes as a result of repeated aggregations. Heterophily is the tendency for nodes of different classes to connect, whereas homophily is the tendency of …

Graph Neural Networks Do Not Always Oversmooth

Defensive Publications Series

Graph neural networks (GNNs) have emerged as powerful tools for processing relational data in applications. However, GNNs suffer from the problem of oversmoothing, the property that the features of all nodes exponentially converge to the same vector over layers, prohibiting the design of deep GNNs. In this work we study oversmoothing in graph convolutional networks (GCNs) by using their Gaussian process (GP) equivalence in the limit of infinitely many hidden features. By generalizing methods from conventional deep neural networks (DNNs), we can describe the distribution of features at the output layer of deep GCNs in terms of a GP: as …

Active Power Sharing Of Thermostat Device With Wax-Actuated Zone Valves, Andrew Chan, Aaron Okano

Defensive Publications Series

This publication describes improved power stealing for thermostats coupled to wax-actuated zone valves. In some implementations, a thermostat that is coupled to a wax-actuated zone valve may further be coupled to a zone controller. In this case, a leakage current from the zone controller is employed to prevent an inrush limiter from discharging and causing a voltage spike. Alternatively, when the thermostat is not coupled to the zone controller, an Active Power Steal (APS) is disabled when the wax-actuated zone valve is in a powered state such that the inrush limiter charges smoothly without a voltage spike.

A Single Sign-On Mechanism For Pre-Boot Authentication Integrated With Bitlocker* Pre-Boot Pin Feature, Hp Inc

Defensive Publications Series

The protection of a business’s data is a primary concern and in many cases this is done with Microsoft BitLocker*. Unfortunately, there are a wide variety of physical attacks that could be performed to potentially leak the BitLocker* keys or allow exfiltration of the data. Two examples are:

• TPM Bus Snooping where they keys are read as they are transferred across the bus
• “Cold-Boot” attacks where a memory freezing technique is used to access the contents of DRAM (including but not limited to BitLocker* decryption key and other potentially sensitive data on the PC)

The standard mitigation for these types …

Dynamic Warming Adjustment Intended To Equalize Drop Weight Difference Along Printing Swath., Hp Inc

Defensive Publications Series

In thermal inkjet printing systems, drop weight depends on the temperature of the DIE, besides of other factors. The temperature of the DIE depends on the content which is fired all the time, because ejecting ink is heating the system itself. Temperature differences are causing drop weight increase , creating then a difference on drop velocity. But, when the swath is built around two staggered printheads addressing the same color, and one of these printheads is partially used, then exists a bias on temperature each other, creating then line banding while printing. Ideally the system should fire with the same …

Reformulating Pharmaceuticals In Low Gwp Propellants, Anonymous

Defensive Publications Series

Studies are being carried out in developing formulations for current and proprietary active pharmaceutical ingredients (APIs) to be delivered by a metered dose inhaler in HFC-152a or HFO-1234ze. Studies are being carried out on reformulating existing API formulations, in particular existing marketed inhaled formulations in 134a or 227ea, with HFC-152a or HFO-1234ze. The API formulations may be in any suitable form, such as in solution or suspension.

Detachable Clip-On Horn For Wearable Device Audio Amplification, Tim Gladwin, Jesse Lippert, Rajiv Gomes

Defensive Publications Series

Audio output from speakers of compact wearable devices such as a smartwatch, is often of poor quality due to various reasons such as small form factor, impedance mismatch, etc. This disclosure describes a detachable clip-on horn that can be coupled to a wearable device such as a smartphone to improve audio output. The horn can be magnetically coupled utilizing pre-existing magnets on the wearable device, or by other suitable mechanisms. The utilization of a tapered horn coupled to the device improves impedance matching between the device speaker and open air, providing a smooth transition for sound waves from the small …

Operating System With Automatic End-To-End Encryption (E2ee) For User Content, Dave Kleidermacher

Defensive Publications Series

While end-to-end encryption (E2EE) can prevent access to user data by unauthorized parties, E2EE is supported by a limited set of applications. Retrofitting existing applications and services to support E2EE is expensive. This disclosure describes techniques that provide E2EE for user personal data across applications and services at the level of the operating system of a user device by automatic detection and/or manual specification of such content. With user permission, personal data is detected automatically by the operating system, e.g., using on-device machine learning or heuristics. When such data is detected, prior to sharing the data with another user, the …

Systems And Methods For Improved Image Stabilization, Ajit Ninan

Defensive Publications Series

A method for improved image stabilization may include shifting image data from a memory device integrated with a display device, where the image data represents motion of the display device and corresponds with a time of capture of the image data; analyzing the image data via a processor to determine a stabilization correction to apply to the image data, analyzing including, determining the stabilization correction to correct for a displacement of at least a portion of the image data; applying the determined stabilization correction to the image data stored in the memory device to achieve stabilized image data; and outputting …

The Composite Light Capturing Configuration That Captures A Wide Area Of The User’S Field Of View At The Same Time Which Partially Occludes The User’S Vision., Punarjeewa Abeysekera

Defensive Publications Series

This paper describes a composite light capturing configuration that captures a wide area of the user’s field of view at the same time which partially occludes the user’s vision. The composite light capturing configuration consists of a main light capturing configuration allocated for the right eye and a main light capturing configuration allocated for the left eye. The main light capturing configuration allocated for the right eye will contain 3 sub light capturing configurations. And the main light capturing configuration allocated for the left eye will contain 3 sub light capturing configurations. The 2 main light capturing configurations collectively will …

Play Neuroball ! : Unleashing The Power Of Spherical Spiking Neural Networks, Renee Gagnon

Defensive Publications Series

This document presents a novel concept in artificial neural networks: the Spherical Spiking Neural Network (Spherical SNN). This approach combines the biological realism of spiking neural networks with a three-dimensional, spherical architecture that more closely mimics the structure of biological brains.

Defensive Publication: Single-Layer 256 Neuron Spike Neural Network With Standing Wave Activation, Renee Gagnon

Defensive Publications Series

This defensive publication details the innovative design and implementation of a single-layer 256 neuron Spike Neural Network (Spike NN) employing shapelets and a standing wave method for neural activation. Spike Neural Networks, inspired by biological neural systems, utilize discrete spikes for information processing, enhancing energy and computational efficiency. In this approach, shapelets—small, local patterns within time series data—are integrated to provide a new dimension of pattern recognition and neural activation. The standing wave method introduces a novel mechanism where interference patterns between outward and inward waves create areas of sustained neural activation. This model not only mimics the complex dynamics …

Patterning Methods For 3d Liquid Crystal Polarization Hologram Optical Elements, Kellie Shaw

Defensive Publications Series

A method of forming a 3D liquid crystal polarization hologram optical element includes irradiating a liquid crystal composition with linearly polarized light, polymerizing the liquid crystal composition, and aligning liquid crystal polymers within the liquid crystal composition along predetermined orientations, where the aligning includes a process selected from point-by-point patterning, slit-by-slit patterning, layer-by-layer patterning, and volumetric patterning.

Reformulating Pharmaceuticals In Low Gwp Propellants

Defensive Publications Series

Studies are being carried out in developing formulations for current and proprietary active pharmaceutical ingredients (APIs) to be delivered by a metered dose inhaler in HFC-152a or HFO-1234ze. Studies are being carried out on reformulating existing API formulations, in particular existing marketed inhaled formulations, with HFC-152a or HFO-1234ze. The API formulations may be in any suitable form, such as in solution or suspension.

A Wireless Charging Coil Stack-Up For Thermal Mitigation And Qi Compliance, Yanchao Li, Liang Jia

Defensive Publications Series

This publication describes a wireless-charging receiver module for wireless charging of an electronic device that may reduce thermal issues, reduce electromagnetic interference, and/or improve receiver coil symmetry.

Reducing Internal Electronic Noise In Electronic Devices, Shengyin Ding, Mike Derer

Defensive Publications Series

This publication describes systems and techniques to reduce internal electronic noise in electronic devices. In aspects, an electronic device includes one or more microphones that are mounted to a surface of an internal structure and a fastener assembly having a screw, an isolation bracket, and a washer. The washer can be utilized at an interface between at least two hardware components in the electronic device, such as the screw, the isolation bracket, and/or the structure on which the one or more microphones are mounted. The washer can reduce electronic noises (“e-noise”) and vibrations commonly found in electronic devices and can …

Abacus Memory Manager For Model State Snapshots, Renee Gagnon

Defensive Publications Series

The Abacus Memory Manager introduces a novel methodology for managing model state snapshots. Inspired by the concept of beads on a wire, this methodology allows for the creation, storage, and retrieval of model states, enabling reproducibility and consistency in model outputs across various types of neural networks, such as Spiking Neural Networks (SNN) and Deep Q-Networks (DQN).

Defensive Publication: Multi-Modal Modular Neural Network For Game Ai, Renee Gagnon

Defensive Publications Series

This publication describes a DQN-based Modular Neural Network architecture designed for Game AI applications. The system combines a Deep Q-Network (DQN) with a Convolutional Neural Network (CNN), Long Short-Term Memory Network (LSTM), and Spiking Neural Network (SNN) in a modular, sequential structure to process game state information, learn optimal strategies, and make decisions.

Integration Of Dual Neural Networks In Game Engines Using Python Api And Socket Communication, Renee Gagnon

Defensive Publications Series

This defensive publication describes a method for integrating dual neural networks, specifically a Spiking Neural Network (SNN) and a Long Short-Term Memory (LSTM) model, into game engines using Python API and socket communication. This approach enhances AI capabilities in first-person shooter (FPS) games by enabling real-time, dynamic behaviors in non-player characters (NPCs). The document outlines the architecture, implementation, and potential applications of this method, providing valuable insights for researchers and developers in the field of game AI.

A Location-Readiness Planner That Can Provide A Heatmap And Insights For A Wi-Fi Network, Nikodin Ristanovic, Taha Hajar, Salvatore Valenza, Samer M Salam, Jim Florwick

Defensive Publications Series

Wi-Fi location services typically involve requirements that span beyond traditional Received Signal Strength Indicator (RSSI) coverage information. For example, inspecting RSSI coverage alone typically does not provide a good indication of whether a wireless network can support location services in a given area. In order to address such issues, a location-readiness planner is proposed herein in which the planner may encompass a tool that is capable of computing location-readiness in all points of a three-dimensional (3D) area utilizing a predefined location-readiness metric. The planner can recommend access point (AP) placements using the predefined location-readiness metric and an efficient genetic algorithm. …

Providing Property Search And Visualization Based On Traditional Architecture, Saravanan Theckyam Santhamoorthy, Hardeepsingh Jethra, Jital Patel

Defensive Publications Series

Current real-estate search platforms do not enable users to match against traditional architectural guidelines (TAG) such as Vaastu or Feng Shui. This disclosure describes techniques that assess the TAG compliance of a particular property. A trained machine learning (ML) model accepts as input the type of TAG compliance, a floor plan of the property, a map of its surroundings, and, with user permission, personal preferences of the user. The model outputs a TAG compliance report, which can be in the form of a three-dimensional visualization of the property with virtual walkthrough features. As a user navigates the 3D visualization, they …

Process For Purification Of Abemaciclib, Anonymous

Defensive Publications Series

The present application relates to a process for purification of Abemaciclib.

Process For The Preparation Of Amorphous Lenacapavir Sodium, Anonymous

Defensive Publications Series

The present disclosure relates to preparation of amorphous Lenacapavir sodium by spray drying technique.

Artificial Intelligence Based Inter Image Design Recommendation, Prajna Shetty, Alok Roy, Satyam Raj, Kandikatla Varshini

Defensive Publications Series

The present disclosure discloses a system and method for providing design recommendations to a user. The method includes parsing the one or more design images of SVG format and analyzing the text and shape parameters associated with the one or more design images. Further the method includes generating python data frames with text identifiers, shape identifiers and connecting component identifiers and generating, by the connecting shape generator, component graph from the trained design images. Further the method includes loading trained component graph from the connecting shape generator and calculating shortest paths from node to be predicted to all other nodes …

Selection Of 5g Akma Authentication Policy And Application Function Based On A Tenant Identifier Obtained From A User Equipment Esim, Ram Mohan R, Rajesh I V

Defensive Publications Series

AKMA (Authentication and Key Management for Applications) is an authentication mechanism introduced in Third Generation Partnership Project (3GPP) Technical Specification (TS) 33.535, Release 16 (and further enhanced in later releases) that can be used to leverage an operator authentication infrastructure to secure communications between a user equipment (UE) and an Application Function (AF). Proposed herein are techniques to facilitate UE authentication for multi-tenant cloud environments. Broadly, techniques proposed herein provide for using a tenant identifier (ID) associated with the eSIM (electronic or embedded Subscriber Identity Module) of a UE in order to determine/obtain the correct policy that corresponds to a …

Method, System, And Computer Program Product For Object-Oriented Reporting, Gregory Paul Loomis Visa

Defensive Publications Series

A method, system, and computer program product are provided for object oriented reporting. The method includes receiving transaction data associated with a plurality of transactions for at least one account, the transaction data comprising a plurality of fields, generating a data library comprising a plurality of modular and self-contained certified data objects, each modular data and self-contained certified object of the plurality of modular and self-contained certified data objects corresponding to at least one field of the plurality of fields, each modular and self-contained certified data object comprising a tag, a length, and a value, generating a string comprising the …

“Dynamic Schema Evolution”, Harshwardhan Tripathi

Defensive Publications Series

The present disclosure provides a solution for seamless integration of data from diverse sources by dynamically managing schema changes. The solution comprises establishing a base schema during data processing and comparing it to incoming data layouts. Any detected changes, such as new tags or fields, are automatically managed within a staging layer, preventing disruptions to the main dataset. These changes are reviewed by Business and Engineering teams for relevance and compatibility. Only after thorough validation, approved changes are processed into the final dataset. This approach ensures that only validated and reliable data is integrated, maintaining data integrity and minimizing disruptions. …

Large Language Model (Llm) Business Helper For Cloud Databases, Ahmet Salih Aksakal

Defensive Publications Series

Cloud database customers can inspect the data they store across multiple databases (potentially spanning across multiple database products) and find details such as schema and nature of stored data from individual database products. This requires technical knowledge of database querying and programming, which complicates business analytics and planning. This disclosure describes techniques to leverage data cataloging services and large language model (LLM) tools to analyze database schema information and to generate contextual information about the structures of the databases owned by a customer. By combining the power of an LLM chatbot with the schema gathering abilities of data cataloging services, …

Process For Purification Of Abemaciclib, Anonymous

Defensive Publications Series

The present application relates to a process for purification of Abemaciclib.

Delivering Personalized Professional Experiences Using Large Language Models, Victor Carbune, Stefan Munteanu, Alexandru Damian

Defensive Publications Series

This disclosure describes the use of large language models (LLMs) to enable professional service providers to account for user preferences together with their own offerings and constraints to provide personalized services. The LLM accepts as input the domain (and sub-domain) expertise and resources of a service provider and the preferences and restrictions of a client. Based on the inputs, the LLM generates a menu of offerings compatible with the specifications of service provider and client. The service provider is provided with LLM-generated suggestions and fulfillment steps that match their resources and skills. Clients are provided with LLM-generated suggestions that match …