Digital Commons Network^™

Architecture To Increase Parallelism In Multisite Testing By Reducing I/O Scan Pins, Achin Grover, Deepak Mittal, Atul Chittora, Mayank Parasrampuria

Defensive Publications Series

Performing quality control on any system-on-a-chip (SoC) requires testing wafers and packages at multiple sites to save testing time and costs. Currently, the parallelism feasible with multisite testing is limited by the number of Input/Outputs (IOs) required because typical testers provide only a limited number of IO channels. This disclosure describes a circuit architecture to reduce the number of scan IN and scan OUT pins for multisite SoC testing by providing scan INs and scan OUTS from the same input port. The scan OUT comparison can then be performed internally within the SoC via one status bit. The status bit …

Foldable Display Including Transparent Layer Mounted On Carrier Film, Howard Hou, Brad Chen

Defensive Publications Series

This publication describes improved foldable display structures that incorporate a cover glass carrier film (“carrier film”) to separate a flexible transparent layer from lower layers of the foldable display structure. A foldable display structure may include, for example, a transparent layer that is formed of a layer of ultra-foldable glass (UFG) that may include a recess filled with an optical-index matching polymer (OIMP) or segments of ultra-thin glass (UTG) joined by a section of OIMP. Conventionally, the transparent layer may be joined to lower layers of the foldable display structure, such as a panel layer, with an optically-clear adhesive (OCA). …

Floating Network Interface Operator, Sebastien Michel

Defensive Publications Series

A Floating Network Interface Operator for Kubernetes® is proposed herein that automates the provisioning, management, and assignment of network interfaces in a Kubernetes® environment, ensuring seamless integration with cloud provider constructs and Kubernetes® resources. By leveraging custom resource definitions and advanced reconciliation loops, the Floating Network Interface Operator enhances network management, supports multi-availability zone setups, and provides robust network configurations tailored to complex application requirements.

Personalizing Ai Models Using Low-Rank Adaptation And Direct Preference Optimization, Saravanan Ganesh, Yi Li, Yunfei Hu, Krystal Kallarackal, Kelvin Nguyen, Chu-Cheng Lin

Defensive Publications Series

Aligning a large language model (LLM) to individual preferences is difficult to perform at scale. This disclosure describes techniques that leverage direct preference optimization (DPO) and low-rank adaptation (LoRA) to enable scalable alignment of artificial intelligence (AI) models. With user permission, user’s edits to suggestions from the model are obtained. The context of the user’s written data or interaction with the LLM is obtained. The user’s edits serve as training data to contextually fine-tune the model using LoRA and DPO. Training data is created as a side product of the user's tasks assisted by the LLM. The training data is …

Remote Cryptographic Device Verification For Theft Deterrence, Siddarth Pandit, Max Bires

Defensive Publications Series

This disclosure describes techniques that leverage remotely provisioned attestation certificates (RPAC) to actively deter device theft. Devices that are reported as lost are identified, and a robust access restriction mechanism is activated that denies both device-local and backend services to the device. To deny service, a bipartite mechanism is deployed that includes enabling a user to declare that their device is stolen and conveying to various services that the device is stolen.

Enhanced Deixis In Video Conferencing With Floating Video And Synthesized Gestures, Xun Qian, David Kim, Ruofei Du

Defensive Publications Series

In traditional video conferencing with screensharing, presenters struggle to reference specific on-screen elements effectively due to the limitations of small video feeds and basic pointer tools. This disclosure describes video conferencing (VC) techniques for enhancing the ability to point (deixis) in a videoconference using synthesized hand gestures. The video feed of the presenter is integrated into the shared screen, while synthesized hand gestures are overlaid to align with speech and pointer movements. With user permission, a multimodal machine learning pipeline accepts as input shared screen content, pointer/controller data, and ongoing speech to generate gestures and to determine optimal video placement. …

Dualprompt Gnn: Tackling Graph Heterogeneity Via Bi- Perspective Prompting

Defensive Publications Series

The challenges tied to unstructured graph data are manifold, primarily falling into node, edge, and graph-level problem categories. Graph Neural Networks (GNNs) serve as effective tools to tackle these issues. However, individual tasks often demand distinct model architectures, and training these models typically requires abundant labeled data, a luxury often unavailable in practical settings. Recently, various "prompt tuning" methodologies have emerged to empower GNNs to adapt to multitask learning with limited labels. The crux of these methods lies in bridging the gap between pretraining tasks and downstream objectives. Nonetheless, a prevalent oversight in existing studies is the homophily-centric nature of …

Heterophily-Aware Personalized Graph Masked Autoencoder

Defensive Publications Series

Graph Masked AutoEncoder (GMAE) has recently attracted vast interest in handling graph-related tasks by adopting the 'masking-reconstruction' learning paradigm. Most existing GMAE-based methods adhere to the homophily assumption, i.e., connected nodes share the same attributes or labels. However, this assumption is not always right because most graphs from real-world applications are mixed by both homophilic and heterophilic edges. Therefore, it is necessary to distinguish them to improve the representative ability of GMAE. In this paper, we propose a heterophily-aware personalized graph masked autoencoder (HAP-GMAE). Specifically, we design a teacher-guided edge discriminator that distinguishes homophilic and heterophilic edges by leveraging the …

Adaptive Node-Subgraph Contrastive Learning For Heterophilic Graph Fraud Detection

Defensive Publications Series

Fraud detection that aims to discern frauds from the majority of benigns has become an increasingly prominent research field. Recently, Graph Neural Networks (GNNs) have been widely applied in graph-based fraud detection due to their outstanding data analysis and mining capabilities. However, owing to the inherent homophily-heterophily mixture and class imbalance of fraud graphs, most GNNs with homophily assumption inevitably suffer from local abnormal signal loss during information propagation, posing significant challenges in situations where frauds are rare and valuable. To address the aforementioned issues, we present a novel adaptive node-subgraph contrastive learning approach for graph-based fraud detection, dubbed ANS-GFD. …

Consistent And Homophily-Aware Representation Learning For Multiplex Graphs

Defensive Publications Series

Although unsupervised multiplex graph representation learning (UMGRL) has been a hot research topic, existing UMGRL methods still has limitations to be addressed. For example, previous works either preserve structural information by ignoring the impact of heterophily in the graph structure or only focus on node-level consistency by ignoring class-level consistency. To address these issues, in this paper, we propose a new UMGRL method, CH-MGRL (Consistency and Homophily-Aware Multiplex Graph Representation Learning), to explore both homophily and consistency in the multiplex graph. Specifically, we propose to restructure the multi-order relationships of every graph between every node and its multi-order neighbors to …

Federated Ai Learning In Healthcare: Comprehensive Framework For Managing Urgent Care Denials, Kush Sharma

Defensive Publications Series

Federated learning (FL) represents a paradigm shift in artificial intelligence (AI) by enabling collaborative model training across decentralized entities—such as hospitals, insurance providers, and clinics—without requiring the exchange of raw patient data. This approach is particularly transformative in healthcare, where data privacy regulations like the Health Insurance Portability and Accountability Act (HIPAA) in the U.S. and impose strict limitations on data sharing. In the context of urgent care denials—such as rejected pre-authorizations, triage prioritization errors, or resource allocation decisions—FL offers a way to improve AI-driven decision-making while preserving patient confidentiality.

This elaborates on a federated AI framework designed to address …

Aggn: Adaptive Granularity Graph Networks For Heterophilic Environments

Defensive Publications Series

Graph neural networks (GNNs) have shown significant success in learning graph representations. However, recent studies reveal that GNNs often fail to outperform simple MLPs on heterophilous graph tasks, where connected nodes may differ in features or labels, challenging the homophily assumption. Existing methods addressing this issue often overlook the importance of information granularity and rarely consider implicit relationships between distant nodes. To overcome these limitations, we propose the Adaptive Granularity Graph Network (AGGN), a novel GNN model specifically designed for heterophilous graphs. AGGN enhances node embeddings by aggregating multi-view information at various granularity levels and incorporating implicit data from distant, …

Halo: Heterophily-Aware Label-Free Ordering For Unsupervised Graph Fraud Detection

Defensive Publications Series

Graph fraud detection (GFD) has rapidly advanced in protecting online services by identifying malicious fraudsters. Recent supervised GFD research highlights that heterophilic connections between fraudsters and users can greatly impact detection performance, since fraudsters tend to camouflage themselves by building more connections to benign users. Despite the promising performance of supervised GFD methods, the reliance on labels limits their applications to unsupervised scenarios; Additionally, accurately capturing complex and diverse heterophily patterns without labels poses a further challenge. To fill the gap, we propose a Heterophily-guided Unsupervised Graph fraud dEtection approach (HUGE) for unsupervised GFD, which contains two essential components: a …

Learning Resilient Graph Structures In Heterophilic Settings

Defensive Publications Series

Graphs provide a fundamental way to model relationships between entities and are central to numerous machine learning tasks. Standard graph-based methods often assume the provided graph structure is both accurate and complete. However, real-world graphs frequently suffer from noise and sparsity, negatively impacting downstream tasks like node classification and clustering. While graph representation learning has advanced significantly, many methods implicitly assume graph homophily (connections predominantly between nodes of the same class), struggling when faced with heterophily (connections predominantly between different classes). This paper introduces a novel method, Resilient Graph Learning for Heterophily (RGLH), designed to learn high-quality graph structures directly …

Robust Graph Learning Through Spatial-Spectral Synergy Against Structural Attacks

Defensive Publications Series

Graph Convolutional Networks (GCNs) are powerful tools for learning from graph data but exhibit significant vulnerability to adversarial structural attacks that manipulate node connections. While various defense strategies focusing independently on the spatial or spectral domains exist, they often fail to leverage the complementary strengths of both perspectives. This paper introduces the Spatial- Spectral Graph Convolutional Network (S²-GCN), a novel framework designed to enhance GCN robustness against structural attacks by synergistically combining spatial and spectral defense mechanisms. S²-GCN comprises two core GCN-based modules operating in parallel. The spectral module utilizes a graph structure derived from learnable low-frequency spectral components, adaptively …

Balancedgraphformer: Enhancing Graph Transformers With Localized Training Against Over-Globalization

Defensive Publications Series

As Transformers gain traction in graph machine learning, the issue of "over-globalization" has emerged, where their global attention mechanisms excessively emphasize distant vertices, potentially diluting vital local information. This is particularly detrimental in graphs where local neighborhoods hold significant predictive value. Existing methods often lack flexibility in local processing or fail to effectively integrate local and global contexts. This paper introduces BalancedGraphFormer, a novel framework designed to localize graph transformer training. It integrates a dedicated local module with a complementary global module. The local module captures fine-grained neighborhood patterns, while the global module integrates broader context without overshadowing local details. …

Quantum-Coherent Graphene Via Entangled Plasma Deposition (Coherium™) Room Temp Superconductor, Peter Branton

Defensive Publications Series

Twisted bilayer graphene (TBG) at a magic angle (~1.1°) has demonstrated superconductivity below ~1.7 K due to emergent flat-band behaviour. However, its mechanical assembly limitations, thermal fragility, and scalability issues hinder real-world applications. The Coherium™ process addresses these limitations by synthesizing entangled carbon plasma into a structurally coherent graphene lattice with integrated AI stabilization, targeting room-temperature Superconducting properties.

Efficient Partitioning For Devices With Multiple User Modes, Harshad Dhabu, Abhijit Adsule

Defensive Publications Series

This document describes techniques that enable a computing device (e.g., a smart watch, ring, glasses, etc.) to more efficiently transition between multiple user modes, such as an adult mode and a child mode. Using a single system image including applications installed on the computing device for every user mode may lead to system image bloat, diminished performance and power inefficiency. Alternatively, creating separate stock keeping units (SKUs) for each user mode may introduce manufacturing and inventory complexities. Instead, the device may divide the device's storage and the applications corresponding to each user mode into three partitions. The first partition may …

Machine Learning System And Method For Predicting Risk Of Drilling Component Failure, Baker Hughes Company

Defensive Publications Series

The invention addresses the problem of drilling tool failure risk assessment in the oil and gas industry. Tool failure in drilling operations can lead to significant financial losses, downtime, and safety risks. The invention aims to provide a solution for predicting and preventing these failures, which is crucial for maintaining safe and efficient drilling operations.

The invention proposes a solution by combining data-driven techniques with expert knowledge. It begins with data collection, including historical data on tool performance, and involves data cleaning and pre-processing. Relevant features are identified, and a predictive model is developed using machine learning techniques. The model …

Stop Encoding And Transferring Remote Imaging Data When The Client Window Is Minimized Or Covered By Other Windows, Hp Inc

Defensive Publications Series

The document discusses a system designed to pause the encoding and transferring of remote imaging data when the client window is minimized or covered by other windows. This system aims to reduce CPU/GPU and network bandwidth consumption, thereby conserving resources and minimizing costs for users. When the client window is minimized or covered, the host stops encoding imaging data, which saves network bandwidth and enhances security by preventing remote pixels from being captured by other applications. The system resumes encoding and transferring imaging data when the client window is restored or becomes visible.

Security Approach To Prevent Data Leak, Hp Inc

Defensive Publications Series

The system is capable to detect QR code-like pattern and further identify it is static or changing. If it's changing, we block it by applying mask image.

Method Of Energy-Based Low-Parametric Inversion In Lwd-Resistivity Geosteering, Baker Hughes Company

Defensive Publications Series

This submission outlines an innovative method, Energy-Based Low-Parametric Inversion, applied to logging while drilling (LWD) resistivity geosteering problems.

The submission points out a direct commercial relevance. The new capabilities of the method are linked to quality of LWD-resistivity service, its throughput, performance and scalability. Also, this method can provide the new service features and be closely synergetic with other logging methods.

Segmented Shield And Antenna For A Lwd Downhole Tool, Baker Hughes Company

Defensive Publications Series

One of the most promising concepts for a half shell LWD antenna is a concept with removable parts, especially an easy removable mechanical shield, coil and ferrites.

Quantum Smart Entangled Fusion Reactor, Peter Branton

Defensive Publications Series

The invention relates to a compact, high-efficiency energy generation system employing

quantum-entangled hydrogen isotope fusion within nanoscale confinement structures. It

comprises a multiplicity of individually controlled fusion traps formed at nanometer scale, each

configured to receive pre-entangled pairs of hydrogen isotopes such as protium, deuterium, or

tritium. Each trap is stimulated by attosecond-timed electromagnetic resonance pulses to induce

fusion via quantum tunneling. Fusion events are monitored and optimised through an integrated

artificial intelligence (AI) control layer that ensures only coherence-validated, phase-aligned

pairs are activated. The system includes capacitive and inductive energy reclamation

components, and operates in inert gas environments without …

Virtual Styling Assistance Using Artificial Intelligence, Shailesh Maheshwari, Tyler Christian Gore, Ke Dong, Boon-Lock Yeo

Defensive Publications Series

This document describes techniques to integrate computing devices with personal style. While colored themes and user personalization are ubiquitous in mobile computing devices (e.g., smartphones, smart watches, fitness trackers, etc.), these options often do not consider real world styling such as the user’s attire and accessories. Mobile devices may be connected to a companion device with a camera, microphone, etc. By leveraging the device’s camera and artificial intelligence, the device can suggest device border colors, display themes, and other recommendations to integrate with the user’s style and enhance the user’s appearance. In some examples, the device may use a microphone …

Relightable Hair Appearance Modeling, Yang Zheng, Leonidas John Guibas, Delio Aleardo Vicini, Dominik Thabo Beeler, Menglei Chai, Yuxiao Zhou

Defensive Publications Series

Systems and methods are proposed for providing relightable hair appearance modeling that generates photorealistic hair. Accurate representation of human hair is significantly important in generating images that digitize humans. However, the manner in which light affects human hair makes accurately capturing human hair very challenging. Methods and systems proposed herein address this challenge by using a two-stage architecture that first captures light parameters and then utilizes a light-aware model that reconstructs the remaining details of hair. The system generates photorealistic hair and can be used in many applications including virtual reality, digital fashion, entertainment industry, and more.

Ts-Ratgnn: Two-Stage Spectral Graph Neural Network With Explicitly Optimized Rational Filters

Defensive Publications Series

Approximation-based spectral graph neural networks, which construct graph filters via function approximation, have demonstrated significant success in graph learning tasks. While existing methods predominantly rely on polynomial approximations for filter construction, rational approximation, a potentially superior alternative, remains largely underexplored. Previous attempts to utilize rational approximations have faced challenges, including high computational costs or reliance on polynomial intermediaries, limiting their full potential. This paper introduces TS-RatGNN, a novel spectral GNN featuring an explicitly-optimized rational filter. TS-RatGNN employs a distinctive two-stage framework: it sequentially applies numerator and denominator filters to the input signal. This streamlines the model architecture, facilitates efficient implementation, …

Leveraging Spectral Graph Properties For Unsupervised Cross-Domain Graph Classification

Defensive Publications Series

Unsupervised graph domain adaptation (UGDA) aims to transfer knowledge learned from labeled source graph datasets to unlabeled target graph datasets originating from a different distribution. Existing approaches often rely on spatial message-passing mechanisms, potentially overlooking the valuable information encoded in the spectral domain specific to UGDA tasks. This work begins with an empirical investigation revealing that low-frequency components of the graph spectrum tend to capture shared features across domains, whereas high-frequency components often represent domain-specific characteristics. Addressing the challenge of effectively utilizing these spectral insights, we introduce the Spectral Synergy Network (SSN), a novel framework for UGDA. SSN explicitly disentangles …

Navigating Extended Reality Maps Using Large Language Models, Tony (Tuấn) Nguyễn, Chunpo Wang, Hiro Tsujino, Stiven Morvan

Defensive Publications Series

Map applications in extended reality (XR map-apps) immersively display map data and features within extended reality (XR) environments. Currently, users navigate within an XR map-app using hand tracking and/or gaze tracking, which can be inconvenient. This disclosure describes techniques that leverage large language models (LLMs) to enable users to query an XR map-app using voice commands. With user permission, the user’s XR environment, including the user’s real environment (under XR passthrough) and the user’s virtual environment (e.g., immersive screen, audio, etc.) is shared with the LLM. A question from the user can be answered by the LLM in the context …

Robust User Gaze Calibration In Head-Mounted Devices, Jake Popham, Ivana Tosic Rodgers

Defensive Publications Series

For optimal performance, gaze tracking systems are calibrated to match user characteristics. However, it can be difficult to collect accurate calibration data because human errors tend to dominate calibration data. This disclosure describes techniques of robust user gaze calibration for head-mounted devices (HMDs) using iteratively reweighted least squares (IRLS). A series of weighted least-squares fits of the data are performed, with the weight of a sample in an iteration being adjusted as a function of its residual error on the previous iteration. By re-weighting samples across iterations, IRLS enables the influence of a sample to vary continuously. The Cauchy loss …