Abrar S. Alhazmi

Dr. Abrar S. Alhazmi

orcid

Research Focus & Metrics

Academic Career Length

2Years (2025-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
10 June 2026 in IEEE Open Journal of Vehicular Technology
Verified
Journal Article
Open access
AAAubida A. Al-Hameed
Aubida A. Al-Hameed
AAAhmad N. Abdulfattah
Ahmad N. Abdulfattah
AAAbrar S. Alhazmi
Abrar S. Alhazmi
ASAhmad Showail
Ahmad Showail
SYSafwan Hafeedh Younus
Safwan Hafeedh Younus
MAMohamad A. Ahmed
Mohamad A. Ahmed
Cellular-connected unmanned aerial vehicles (UAVs) experience frequent handovers at altitude because line-of-sight propagation exposes each device to many cells with comparable signal strength, causing failures that threaten safety-critical missions. Machine learning can anticipate handovers proactively, but most predictors are opaque, and existing interpretable alternatives ignore domain knowledge such as the monotonic relationships defined by 3GPP A3 events. This paper introduces the Hierarchical Attribution Neural Additive Model (HA-NAM). HA-NAM decomposes each prediction into per-feature shape functions in three groups (spatial radio quality, UAV mobility, and handover history), enforces hard monotonicity on 12 of 15 features, and adds lightweight cross-group interaction modules. We evaluate HA-NAM on a 500 ms look-ahead A3 prediction task in a 57-cell urban macro deployment with L3-filtered RSRP, spatially consistent LoS/NLoS states, and trajectory-level data splitting. HA-NAM achieves AUC $0.952 \pm 0.003$ and the lowest missed-handover rate among all intrinsically interpretable models. LSTM and GRU baselines yield only a $+0.8$ pp AUC gain at the cost of per-feature transparency. On real 5G UAV measurements from the AERPAW testbed, HA-NAM achieves AUC 0.934 to 0.961 across flight-level splits, and simulation-trained shape functions show statistically significant correlation with real handover labels. Edge benchmarks confirm sub-10 ms inference on a Raspberry Pi 4 and sub-6 ms on an NVIDIA Jetson AGX Orin. Physics-constrained additive models can therefore match black-box accuracy while providing auditable, per-feature explanations that transfer from simulation to real aerial networks.

1 January 2026 in IEEE Access
Verified
Journal Article
Open access
SYSafwan Hafeedh Younus
Safwan Hafeedh Younus
AAAbrar S. Alhazmi
Abrar S. Alhazmi
AAAubida A. Al-Hameed
Aubida A. Al-Hameed
ASAhmad Showail
Ahmad Showail
MAMohamad A. Ahmed
Mohamad A. Ahmed
OYOthman Younus
Othman Younus

29 May 2026 in IEEE Open Journal of the Communications Society
Verified
Journal Article
Open access
AAAubida A. Al-Hameed
Aubida A. Al-Hameed
AAAbrar S. Alhazmi
Abrar S. Alhazmi
SYSafwan Hafeedh Younus
Safwan Hafeedh Younus
ASAhmad Showail
Ahmad Showail
MAMohamad A. Ahmed
Mohamad A. Ahmed
AI-driven xApps for Open Radio Access Network (O-RAN) energy saving achieve substantial power reductions, but they operate as black boxes. Operators cannot audit their decisions, decompose them for cross-xApp conflict negotiation, or constrain them against physically unsafe actions under distribution shift. This paper proposes the Energy-Saving Neural Additive Model (ES-NAM), an inherently interpretable architecture for a near-real-time RAN Intelligent Controller xApp that performs multi-level Advanced Sleep Mode (ASM) control. Unlike black-box and post-hoc explainability approaches, ES-NAM decomposes each sleep decision into independent, auditable per-feature shape functions, organized into three physically grounded groups: traffic load, radio coverage, and operational history. Hard monotonicity constraints, enforced through softplus weight reparameterization, guarantee that physically unsafe actions such as deep-sleeping a heavily loaded cell cannot occur. The guarantee holds independently of the training data. A cumulative link ordinal head preserves the natural ordering of sleep levels, and group-level attribution scores enable a tunable protocol for evidence-based conflict resolution with co-deployed xApps. On a 57-cell O-RAN-compliant simulation, ES-NAM achieves comparable energy-saving performance of black-box baselines and achieves zero monotonicity violations on out-of-distribution data. On a 10,000-scenario conflict-resolution evaluation with held-out validation and test splits, the attribution-based protocol achieves 99.5% override precision at the data-driven operating point. Cross-dataset validation confirms that these structural guarantees transfer to unseen network conditions.

14 October 2026 in Mathematics
Verified
Journal Article
Open access
PDF file
AAAbrar S. Alhazmi
Abrar S. Alhazmi
MAMohammed M Alenazi
Mohammed M Alenazi
Integrated sensing and communication (ISAC), semantic communication, and mobile edge computing (MEC) jointly create a high-dimensional, non-convex, and time-coupled resource allocation problem in sixth-generation (6G) networks that existing heuristics and deep reinforcement learning (DRL) methods solve either slowly or without stability guarantees. This study proposes DiffLyap-MEC, a framework that couples generative diffusion models with Lyapunov drift-plus-penalty optimization for online decision-making in ISAC-enabled multi-user, multi-server MEC systems. A mixed-integer non-linear program jointly optimizes task offloading ratios, semantic compression levels, ISAC beamforming, and edge CPU and bandwidth allocation under long-term energy-queue stability and sensing-accuracy constraints. Lyapunov theory decouples the long-term problem into per-slot subproblems with a provable O(1/V) optimality versus backlog trade-off, and a conditional denoising diffusion policy generates near-optimal offloading and beamforming decisions in a fixed number of reverse steps, with a closed-form water-filling solution embedded as a differentiable layer. Bounds on queue stability, policy convergence, and the distortion versus latency trade-off are derived. Simulations with 3GPP-compliant channel models show that DiffLyap-MEC reduces average end-to-end latency by 27.3% to 33.9% and energy consumption by 22.2% to 27.1% relative to proximal policy optimization (PPO), soft actor–critic (SAC), and Lyapunov-only baselines while maintaining 97.3% sensing accuracy at 0.74 ms inference time per slot on an NVIDIA RTX A6000 GPU with a batch size of one.

4 November 2025 in Research Square
Verified
Preprint
Open access
PDF file
MAMohammed M Alenazi
Mohammed M Alenazi
AAAbrar S. Alhazmi
Abrar S. Alhazmi
The proliferation of Internet of Things (IoT) devices in financial markets creates unprecedented opportunities for distributed intelligence in cryptocurrency trading systems. This paper presents EdgeFL-Crypto, a novel federated split learning architecture that leverages IoT sensor networks and edge-cloud computing paradigms for real-time cryptocurrency volatility forecasting. Our approach addresses the unique challenges of deploying transformer-based models across resource-constrained IoT devices by implementing a hierarchical federated learning protocol that splits model computation between edge and cloud layers. The framework integrates mobile edge computing nodes as intermediate aggregators, enabling efficient model training while preserving data locality in IoT ecosystems. We introduce an adaptive split point selection mechanism that dynamically partitions transformer layers based on device capabilities and network conditions, optimizing the trade-off between edge computation and cloud processing. The system employs blockchain-secured aggregation for trustworthy model updates across distributed IoT networks, ensuring integrity in multi-stakeholder environments. Experimental evaluation on real-world cryptocurrency data from IoT-enabled trading terminals demonstrates 23.7% improvement in prediction accuracy while reducing communication overhead by 41.2% compared to traditional cloud-centric approaches. The framework achieves sub-200ms inference latency suitable for real-time IoT applications, with differential privacy guarantees (ε=1.0) protecting sensitive trading data at the edge. Our results establish EdgeFL-Crypto as a practical solution for deploying sophisticated AI models in IoT-driven financial systems, bridging the gap between edge intelligence and cloud computing for next-generation fintech applications.

7 April 2026 in Mathematics
Verified
Journal Article
Open access
PDF file
AAAbrar S. Alhazmi
Abrar S. Alhazmi
Non-convex stochastic optimization presents fundamental mathematical challenges across machine learning, wireless networks, data center resource allocation, and optical wireless communication systems, where complex loss landscapes with multiple local minima and saddle points impede classical variational inference methods. This paper introduces the Quantum-Inspired Variational Inference (QIVI) framework, which systematically integrates quantum mechanical principles (superposition, entanglement, and measurement operators) into classical variational inference through rigorous mathematical formulations grounded in Hilbert space theory and operator algebras. We develop a unified optimization framework that encodes classical parameters as quantum-inspired states within finite-dimensional complex Hilbert spaces, employing unitary evolution operators and adaptive basis selection governed by gradient covariance eigendecomposition. The core mathematical contribution establishes that QIVI achieves a convergence rate of O(log2T/T1/2) for σ-strongly non-convex functions, provably improving upon the classical O(T−1/4) rate, yielding a theoretical speedup factor of 1.85–1.96×. Comprehensive experiments across synthetic benchmarks, Bayesian neural networks, and real-world applications in network optimization and financial portfolio management demonstrate 23–47% faster convergence, 15–35% superior objective values, and 28–46% improved uncertainty calibration. The principal contributions include: (i) a rigorous Hilbert space-based mathematical framework for quantum-inspired variational inference grounded in operator algebras, (ii) a novel hybrid quantum–classical algorithm (QIVI) with adaptive basis selection via gradient covariance eigendecomposition, (iii) formal convergence proofs establishing provable improvement over classical methods, (iv) comprehensive empirical validation across diverse problem domains relevant to machine learning and network optimization, and (v) demonstration of the framework’s applicability to optimization problems arising in wireless networks, data center resource allocation, and network system design. Statistical validation using the Friedman test (χ2=847.3, p<0.001) and post hoc Wilcoxon signed-rank tests with Holm–Bonferroni correction confirm that QIVI’s improvements over all baseline methods are statistically significant at the α=0.05 level across all benchmark categories. The framework discovers 18.1 out of 20 true modes in multimodal distributions versus 9.1 for classical methods, demonstrating the potential of quantum-inspired optimization approaches for challenging stochastic problems arising in machine learning, wireless communication, and network optimization.