arXiv subjects

Abdullah Ahmed

Publications and source records attributed to Abdullah Ahmed.

6 recordsLinked to original sources

arXiv2026

Effects of Heterogeneity in Two-Cell Feedforward Networks

As the need for higher performance from biological and electronic sensors continues to outpace current technologies, new strategies for designing, developing, and implementing novel sensor systems are emerging. A recently introduced feedforward network-based approach can simultaneously enhance a signal while steering a radiating beam in radio frequency communication systems. Furthermore, the approach is also model-independent, thus making it suitable for other applications. In this work, we aim to understand the effects of inhomogeneities in feedforward arrays, which are inevitable in real-world implementations. We investigate a collection of two-cell feedforward networks composed of pitch-fork cells and Stuart-Landau oscillators and quantify the effects of parameter inhomogeneities using system reduction, analytical and computational bifurcation analyses, and a singularity theory approach. Contrary to common intuition, inhomogeneity in the excitation parameter can be exploited to enhance the network output growth rate. While frequency inhomogeneity in Stuart-Landau networks primarily has an adverse effect on signal amplification, phase locking persists over a surprisingly broad range of inhomogeneity.

math.DS↗

arXiv2025

Latent Sensor Fusion: Multimedia Learning of Physiological Signals for Resource-Constrained Devices

Latent spaces offer an efficient and effective means of summarizing data while implicitly preserving meta-information through relational encoding. We leverage these meta-embeddings to develop a modality-agnostic, unified encoder. Our method employs sensor-latent fusion to analyze and correlate multimodal physiological signals. Using a compressed sensing approach with autoencoder-based latent space fusion, we address the computational challenges of biosignal analysis on resource-constrained devices. Experimental results show that our unified encoder is significantly faster, lighter, and more scalable than modality-specific alternatives, without compromising representational accuracy.

eess.SP↗

arXiv2025

Advancing Conversational Diagnostic AI with Multimodal Reasoning

Large Language Models (LLMs) have demonstrated great potential for conducting diagnostic conversations but evaluation has been largely limited to language-only interactions, deviating from the real-world requirements of remote care delivery. Instant messaging platforms permit clinicians and patients to upload and discuss multimodal medical artifacts seamlessly in medical consultation, but the ability of LLMs to reason over such data while preserving other attributes of competent diagnostic conversation remains unknown. Here we advance the conversational diagnosis and management performance of the Articulate Medical Intelligence Explorer (AMIE) through a new capability to gather and interpret multimodal data, and reason about this precisely during consultations. Leveraging Gemini 2.0 Flash, our system implements a state-aware dialogue framework, where conversation flow is dynamically controlled by intermediate model outputs reflecting patient states and evolving diagnoses. Follow-up questions are strategically directed by uncertainty in such patient states, leading to a more structured multimodal history-taking process that emulates experienced clinicians. We compared AMIE to primary care physicians (PCPs) in a randomized, blinded, OSCE-style study of chat-based consultations with patient actors. We constructed 105 evaluation scenarios using artifacts like smartphone skin photos, ECGs, and PDFs of clinical documents across diverse conditions and demographics. Our rubric assessed multimodal capabilities and other clinically meaningful axes like history-taking, diagnostic accuracy, management reasoning, communication, and empathy. Specialist evaluation showed AMIE to be superior to PCPs on 7/9 multimodal and 29/32 non-multimodal axes (including diagnostic accuracy). The results show clear progress in multimodal conversational diagnostic AI, but real-world translation needs further research.

cs.CL↗

arXiv2025

Bandwidth Reservation for Time-Critical Vehicular Applications: A Multi-Operator Environment

Onsite bandwidth reservation requests often face challenges such as price fluctuations and fairness issues due to unpredictable bandwidth availability and stringent latency requirements. Requesting bandwidth in advance can mitigate the impact of these fluctuations and ensure timely access to critical resources. In a multi-Mobile Network Operator (MNO) environment, vehicles need to select cost-effective and reliable resources for their safety-critical applications. This research aims to minimize resource costs by finding the best price among multiple MNOs. It formulates multi-operator scenarios as a Markov Decision Process (MDP), utilizing a Deep Reinforcement Learning (DRL) algorithm, specifically Dueling Deep Q-Learning. For efficient and stable learning, we propose a novel area-wise approach and an adaptive MDP synthetic close to the real environment. The Temporal Fusion Transformer (TFT) is used to handle time-dependent data and model training. Furthermore, the research leverages Amazon spot price data and adopts a multi-phase training approach, involving initial training on synthetic data, followed by real-world data. These phases enable the DRL agent to make informed decisions using insights from historical data and real-time observations. The results show that our model leads to significant cost reductions, up to 40%, compared to scenarios without a policy model in such a complex environment.

cs.LG↗

arXiv2024

Unveiling the Stealthy Threat: Analyzing Slow Drift GPS Spoofing Attacks for Autonomous Vehicles in Urban Environments and Enabling the Resilience

Autonomous vehicles (AVs) rely on the Global Positioning System (GPS) or Global Navigation Satellite Systems (GNSS) for precise (Positioning, Navigation, and Timing) PNT solutions. However, the vulnerability of GPS signals to intentional and unintended threats due to their lack of encryption and weak signal strength poses serious risks, thereby reducing the reliability of AVs. GPS spoofing is a complex and damaging attack that deceives AVs by altering GPS receivers to calculate false position and tracking information leading to misdirection. This study explores a stealthy slow drift GPS spoofing attack, replicating the victim AV's satellite reception pattern while changing pseudo ranges to deceive the AV, particularly during turns. The attack is designed to gradually deviate from the correct route, making real-time detection challenging and jeopardizing user safety. We present a system and study methodology for constructing covert spoofing attacks on AVs, investigating the correlation between original and spoofed pseudo ranges to create effective defenses. By closely following the victim vehicle and using the same satellite signals, the attacker executes the attack precisely. Changing the pseudo ranges confuses the AV, leading it to incorrect destinations while remaining oblivious to the manipulation. The gradual deviation from the actual route further conceals the attack, hindering its swift identification. The experiments showcase a robust correlation between the original and spoofed pseudo ranges, with R square values varying between 0.99 and 1. This strong correlation facilitates effective evaluation and mitigation of spoofing signals.

cs.CR↗

arXiv2021

Benchmarking Modern Named Entity Recognition Techniques for Free-text Health Record De-identification

Electronic Health Records (EHRs) have become the primary form of medical data-keeping across the United States. Federal law restricts the sharing of any EHR data that contains protected health information (PHI). De-identification, the process of identifying and removing all PHI, is crucial for making EHR data publicly available for scientific research. This project explores several deep learning-based named entity recognition (NER) methods to determine which method(s) perform better on the de-identification task. We trained and tested our models on the i2b2 training dataset, and qualitatively assessed their performance using EHR data collected from a local hospital. We found that 1) BiLSTM-CRF represents the best-performing encoder/decoder combination, 2) character-embeddings and CRFs tend to improve precision at the price of recall, and 3) transformers alone under-perform as context encoders. Future work focused on structuring medical text may improve the extraction of semantic and syntactic information for the purposes of EHR de-identification.

cs.CL↗