Detecting personal identifiable information (PII), personal healthcare information(PHI) and similar sensitive data within structured datasets is imperative for various reasons. This has become an important problem for organizations to solve automated detection especially amidst high volumes of data collection with the rise of data protection laws and user privacy importance. Multiple sensitive data detection methods exist in structured and unstructured datasets. Structured data pose a greater challenge for such problems due to lack of external context and general understanding of the data. We describe the importance of the problem's existence. We list out existing solutions in the space with their limitations and direct towards the solution using generative artificial intelligence. We want to focus on how the latest advancements in LLMs can provide a truly global solution to the global problem such as sensitive data detection.

In this work, we present a framework to provide network services to fixed home users in rural areas through joint utilization of fifth-generation (5G) next-generation node B (gNB) and an unmanned aerial vehicle (UAV), wherein, users outside the service area of gNB are served by the UAV. In this regard, we propose an optimization framework to maximize the coverage radius of the gNB considering the 5G rural macro (RMa) propagation scenario. Next, we propose optimization frameworks for UAV-assisted non-orthogonal multiple access (NOMA) and orthogonal multiple access (OMA) data transmission schemes. The aforementioned frameworks for UAV-assisted data transmission schemes minimize the communication power and compute the optimal UAV hovering height to meet the desired QoS requirements of the users. Thereafter, we investigate a framework to design a wide elliptical beam to serve the users under the coverage of the UAV. Next, in this paper, we present an exhaustive mathematical framework to determine the angular velocity of the rotors, the total current consumed by the rotors and the hovering power consumption of the UAV, wherein the UAV hovers at the optimal height depending upon the NOMA or OMA data transmission scheme. We propose a graphical methodology to compute the battery life of the UAV. The simulation results demonstrate that the proposed method of UAV-assisted NOMA and OMA data transmission coupled with the wide elliptical beam significantly reduces the communication power requirements to meet the desired QoS requirements of the ground users as compared to the conventional UAV-assisted NOMA and OMA transmission schemes. Furthermore, the proposed methodology considerably enhances the battery life of the UAV, thereby resulting in a greater hovering time of the UAV compared to the conventional UAV-assisted NOMA and OMA transmission schemes.

Arm weight support increases range of motion and breaks abnormal movement patterns such as flexion synergy. Conventional tools for arm weight support are simple devices to be used at homes but continuous feedback and monitoring is not possible. Rehabilitation robots could be a potential solution but are very complex for home use. Thus, a hybrid system which is simple and provides feedback is needed to increase adherence to home therapy. The kinematic design and optimization of link lengths of one such hybrid robot-Motorized Arm Support (MARS), is proposed in this study. Two kinematic chains of MARS (4 DOF and 5 DOF MARS) were considered for supporting planar reaching movements of the arm along with shoulder flexion/ extension. The 5 DOF MARS reaches 4.4% more work area of the arm but increases the complexity of the system as it requires additional mechanical elements such as dampers to prevent impact forces at the arm. Thus, to keep the system simple, the small loss in work area is considered acceptable and the 4 DOF kinematic chain of MARS is chosen for further development.

The testing and safety cases of Assisted and Automated Driving functions require considerations for non ideal environmental conditions, such as adverse and extreme weather. In these extreme conditions, perception sensors (e.g. camera, LiDAR, RADAR), used to build the situational awareness of the vehicle, might produce noisy and degraded data, and it is therefore key to consider: (i) how to reliably and robustly measure data degradation; (ii) how to evaluate de-noising techniques. This paper focuses on de-snowing of LiDAR data, as falling snow is one of the most variable and dangerous conditions to be encounter while driving-and LiDAR can provide essential 3D information to still enable safe vehicle navigation. Using the WADS dataset, which contains segmented pointclouds including falling and deposited snow points, 4 different state-of-the-art desnowing techniques are compared using an array of adapted pointcloud quality metrics, and combined with perception based metrics. The different metrics are able to capture different aspects of the data degradation, and hereby the novel De-Snow Score (DSS) is proposed and applied to have a holistic evaluation of the de-noising techniques. Based on DSS, the most promising de-noising algorithms are identified. The proposed methodology and De-Snow Score can pave the way for a standardised approach when measuring perception sensor data degradation and de-noising.

Federated Learning (FL) and blockchain technology have emerged as transformative solutions to current challenges. FL offers a decentralized machine-learning approach, prioritizing privacy, while blockchain promises transparency, security, and decentralization for transactions and data storage. FL enables machine learning models to train across multiple devices while keeping data localized, enhancing data privacy and security. Additionally, FL allows for efficient utilization of computational resources by distributing the training process among participating devices. This not only reduces the burden on individual devices but also enables faster model training and improved accuracy through collaborative learning.Blockchain, originally designed for cryptocurrencies like Bitcoin, creates immutable, transparent, and secure records in a decentralized manner, fostering trust in peer-to-peer networks.The integration of federated learning (FL) and blockchain has the potential to revolutionize data privacy, security, and transparency. By combining FL's ability to train models across multiple devices with blockchain's immutable and decentralized record-keeping, a new paradigm of secure and trustworthy machine learning can be achieved. However, challenges and considerations need to be addressed to fully realize the potential of these technologies.This literature review explores this intersection, including its intricacies, potential, challenges, and future trajectory.

VeCS which stands for vertical conductive structure is an advanced design and process approach to improve BGA, connectors signal/power integrity (SI/PI), and routing density. The VeCS has conductive traces surrounded by a conductive shield. Because the VeCS is a semi-coaxial structure, it has similar advantages that coaxial structures have. The first advantage is well-managed characteristic impedance along with vertical. The signal always references the metal shield, thus the characteristic impedance does not vary vertically. Another advantage is low loss. The loss can be from both insertion and return losses. The VeCS has a lower return loss compared to PCB vias, which improves the loss. As scalable models for the via have been introduced, this paper also proposes a scalable model for the VeCS. The scalable model is an analytical approach to analyze the VeCS and includes cascaded RLC blocks. 3D electromagnetic (EM) simulations provided the insertion losses depending on the height of the VeCS, and optimization algorithms established the scalable model. The optimization algorithms include simulated annealing (SA) to find the equivalent circuit and linear regression to generalize the equivalent circuits depending on the height. In conclusion, this paper introduces the VeCS and proposes its scalable model to efficiently analyze and design the VeCS. The 3D full-wave simulation shows that The VeCS has better electrical performances in the time and frequency domain.

Vertical conductive structure (VeCS) technology, an advanced interconnect technology for signal transition on different printed circuit board (PCB) layers, was developed to meet needs for more intelligent PCB technologies with decreased layer count, and improved signal integrity and power integrity with respect to those of conventional transition via structures. In contrast to the conventional via structure, the VeCS consists of a plate signal trace and an oval-shaped shielding ground racket comprising a semi-coaxial structure or a vertical transmission line. The VeCS achieves a continuous characteristic impedance that improves electrical performance regarding losses, crosstalk, and electromagnetic interference. Herein, the physical structure of the VeCS is first introduced. Subsequently, the time domain reflectometer impedance of the VeCS is extracted and analyzed through full-wave simulations. The crosstalk between VeCS-based and via-based channels on the PCB is compared to clarify the advantages of VeCS technology. Finally, measurements are used to further demonstrate the advantages of VeCS technology.

The digital era has brought about a myriad of challenges in the job market, where job seekers often struggle to find positions that align with their skills and preferences, while employers face difficulties in identifying the most suitable candidates for their job openings. Existing job recommendation systems, although advanced, often lack the precision and personalization needed to address these challenges effectively. This research paper addresses these issues by proposing an progressive approach that leverages the Heterogeneous Information Network-based GraphSAGE (HINSAGE) algorithm within knowledge graphs. By harnessing the rich semantic and structural information present in heterogeneous information networks (HINs), the study aims to improve job recommendation accuracy and personalization, ultimately benefiting both job seekers and employers by facilitating better job matches and streamlining the recruitment process in the digital job market.

Multi-modal knowledge graph completion (MMKGC) aims to automatically discover the unobserved factual knowledge from a given multi-modal knowledge graph by collaboratively modeling the triple structure and multi-modal information from entities. However, real-world MMKGs present challenges due to their diverse and imbalanced nature, which means that the modality information can span various types (e.g., image, text, numeric, audio, video) but its distribution among entities is uneven, leading to missing modalities for certain entities. Existing works usually focus on common modalities like image and text while neglecting the imbalanced distribution phenomenon of modal information. To address these issues, we propose a comprehensive framework NativE to achieve MMKGC in the wild. NativE proposes a relation-guided dual adaptive fusion module that enables adaptive fusion for any modalities and employs a collaborative modality adversarial training framework to augment the imbalanced modality information. We construct a new benchmark called WildKGC with five datasets to evaluate our method. The empirical results compared with 21 recent baselines confirm the superiority of our method, consistently achieving state-of-the-art performance across different datasets and various scenarios while keeping efficient and generalizable. Our code and data are released at https://github.com/zjukg/NATIVE.

Deep learning (DL) methods are currently being explored to recover images from sparse-view, limited-data, and undersampled acquisitions in medical applications. Although DL-based solutions may appear visually appealing based on likability/subjective criteria (such as less noise, smooth features), they may also suffer from imperceptible fakes. This issue is further exacerbated by a lack of easy-to-use techniques and robust metrics for the identification of fakes in DL-based outputs. In this work, we propose performing Fourier Ring Correlation (FRC)-based analysis over small patches and concomitantly (s)canning across DL-based outputs and their reference counterparts to identify fakes. We term the metrics as sFRC. We describe the rationale behind sFRC and provide its mathematical framework. The thresholds required for the sFRC can be set using predefined fake features or imaging theory-based fake maps. We use sFRC to identify fakes for two undersampled medical imaging problems (CT super-resolution and MRI subsampled recovery). We demonstrate the effectiveness of sFRC in finding fake features for the two imaging problems and its agreement with a different imaging theory-based method on fake feature maps. Finally, we quantify the incidences of fakes from DL-based methods relative to indistribution versus out-of-distribution data and the increment in subsampling rate.

Deep brain stimulation of the subthalamic nucleus (STN-DBS) is an established treatment for motor impairment due to Parkinson's disease (PD) progression. While treated subjects mostly experience significant amelioration of symptoms, some still report adverse effects. In particular, changes in gait patterns due to the electrical stimulation have shown mixed results across studies, with overall gait velocity improvement described as the core positive outcome. This retrospective study investigates changes in the gait parameters of 50 PD patients before and 6 months after STN-DBS, by exploiting a purely data-driven approach. First, unsupervised learning identifies clusters of subjects with similar variations in the gait parameters after STN-DBS. This analysis highlights two dominant clusters (Silhouette score: 0.45, Dunn index: 0.18), with one of them associated to a worsening in walking. Then, supervised machine learning models (i.e., Support Vector Machine and Ensemble Boosting models) are trained using pre-surgery gait parameters, clinical scores, and demographic information to predict the two gait change clusters. In a Leave-One-Subject-Out validation, the best model achieves balanced accuracy over 82%, denoting moderate predictability of both clusters. Moreover, feature importance analysis reveals the variability in the step width and in the step length asymmetry during the preoperative gait test as promising biomarkers to predict gait response to STN-DBS.

In contemporary heavy-load distribution networks, preceding feasibility assessment is imperative before incorporating additional energy units. However, the feasibility examination for massive combined operational scenarios of relevant units is computationally intensive with repetitive power flow calculations. To this end, this paper proposes a rapid assessment framework, the kernel of which is to learn from formerly examined scenarios, thus forming expansive feasible/infeasible regions to geometrically rule in/out subsequent scenarios. By sidestepping the power flow computation in most scenarios, we accelerate the assessment process. Furthermore, enlightened by heuristic hypersurface search, such prechecking efficiency can be boosted by resorting the scenario sequence. In a risk-averse manner, this framework can be conceptualized using the exact grid model, though it is initially designed for the convex grid model. Evidenced by testing on two different-scale distribution grids, the framework shows a significant assessment efficiency improvement and strict accuracy guarantee, where we observe at least 76.13% assessment time reduction and zero accuracy loss. We anticipate this work to be a starting point for more sophisticated geometry-accelerating feasibility assessment methods.

Drug design is undergoing a transformation as we challenge conventional methods by integrating state-of-the-art artificial intelligence with the intricate domain of molecular biology. At the heart of our endeavor lies a significant challenge: the scarcity of datasets containing active compounds for emerging target proteins. To confront this obstacle, we're pioneering an innovative approach. We're merging the advanced Generative Pre-trained Transformer (GPT) architecture with the nuanced capabilities of Long Short-Term Memory (LSTM) networks, with the aim of generating Simplified Molecular Input Line Entry System (SMILES) strings to unveil novel therapeutic pathways. Additionally, we're employing a Bidirectional Encoder Representations from Transformers (BERT) pretraining strategy to enrich our model with comprehensive molecular data, including amino acid sequences and molecular SMILES datasets. Through meticulous fine-tuning on a meticulously curated protein-ligand complex dataset, we're achieving precise conditional generation via autoregressive supervised learning. Our research introduces a groundbreaking method to assess molecular affinity, validated against established proteins, showcasing superior binding affinities compared to certain FDA-approved drugs in docking experiments. By pushing the boundaries of generative algorithms and establishing a robust framework for evaluating molecular affinity, we're driving forward the field of de novo drug design, offering promising therapeutic avenues and enabling deeper exploration of the chemical landscape.

A document by Guzman Llambias . Click on the document to view its contents.

The stress dependence of the magnetic properties of non-oriented Fe-Si steel sheets has been investigated by measurement and analysis of hysteresis loop, magnetization curve, and energy losses taken at different peak polarization values Jp (0.5 T – 1.5 T) between DC and f = 400 Hz. The salient feature of the material response to the stress lies in the monotonic deterioration of the soft magnetic properties, across the whole (Jp - f) domain, on passing from the maximum tensile stress (σ = +30 MPa) to the maximum compression (σ = -30 MPa). This is understood in terms of stress-induced redistribution of the domains between easy axes, making magnetic hardening by compression directly related to unfavorably directed domains and 90° domain-wallmediated magnetization transitions. The loss decomposition is carried out across the whole investigated frequency range, taking into account the skin effect at the highest frequencies. Quasi-static and dynamic losses follow a same trend with σ, both monotonically increasing on passing from the tensile to the compressive stress limits, according to the theoretically expected relationship existing between the hysteresis and the excess loss components. The latter is shown to identify the correlation regions where the magnetization is reversed of size comparable with the average grain size and loosely following the dependence of the loss figure on the applied stress.

The need for processing at the edge the increasing amount of data that is being produced by multitudes of sensors has led to the demand for mode power efficient computational systems, by exploring alternative computing paradigms and technologies. Neuromorphic engineering is a promising approach that can address this need by developing electronic systems that faithfully emulate the computational properties of animal brains. In particular, the hippocampus stands out as one of the most relevant brain region for implementing auto associative memories capable of learning large amounts of information quickly and recalling it efficiently. In this work, we present a computational spike-based memory model inspired by the hippocampus that takes advantage of the features of analog electronic circuits: energy efficiency, compactness, and real-time operation. This model can learn memories, recall them from a partial fragment and forget. It has been implemented as a Spiking Neural Networks directly on a mixed-signal neuromorphic chip. We describe the details of the hardware implementation and demonstrate its operation via a series of benchmark experiments, showing how this research prototype paves the way for the development of future robust and low-power mixed-signal neuromorphic processing systems.

The advent of the Internet of Things (IoT) has ushered in transformative possibilities for smart cities, with the potential to revolutionize urban living through enhanced connectivity and data-driven decision-making. However, the effective realization of IoT in smart cities hinges upon the seamless interoperability of diverse devices and systems. To address this critical need, the oneM2M standards initiative has emerged as a foundational framework for IoT interoperability. In this research paper, we perform an exploratory analysis of three prominent open-source oneM2M based interoperability systems-Mobius, OM2M, and ACME. We leverage an existing large-scale system provided by our Smart City Living Lab deployed at IIIT Hyderabad, sprawling a 66-acre campus featuring over 370 nodes across eight verticals. We investigate the architectural characteristics of each solution, considering their strengths and limitations in facilitating IoT interoperability. Through this analysis, our paper aims to provide valuable insights for stakeholders seeking to implement IoT interoperability solutions in the context of smart cities. By evaluating the strengths and limitations of Mobius, OM2M, and ACME, we seek to offer guidance for selecting the most suitable solution. Our analysis reveals that the optimal framework choice depends on specific quality constraints: Mobius excels in performance, while ACME offers advantages in ease of setup for smaller-scale implementations.

A document by Christoph Funda . Click on the document to view its contents.