Arshad Ali Khan

Arshad Ali Khan

Auto-generated profile, based on publicly available information

Arshad Ali Khan

Elm Europe, Southampton, U.K.

Research Focus & Metrics

Academic Career Length

11Years (2016-Present)

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

Representative Publications

View All Publications
23 September 2024 in IEEE Access
Verified
Journal Article
Open access
UBUmmara Bibi
Ummara Bibi
MMMahrukh Mazhar
Mahrukh Mazhar
DSDilshad Sabir
Dilshad Sabir
MBMuhammad Fasih Uddin Butt
Muhammad Fasih Uddin Butt
AHAli Hassan
Ali Hassan
WAWadood Abdul
Wadood Abdul
With ongoing advancements in natural language processing (NLP) and deep learning methods, the demand for computational and memory resources has considerably increased, which signifies the determination of efficient and compact models in resource-constrained environments. A comprehensive overview of the most recent advancements in pruning and quantization methods for deep neural networks is provided in this paper. Numerous cutting-edge techniques that harness the complementary advantages of pruning and quantization have been analyzed, highlighting their effectiveness in reducing model size, enhancing computational efficiency, and minimizing memory usage. These techniques include Quantization and Sparsity Aware Fine Tuning, Compression Learning by In-Parallel Pruning-Quantization, GroupReduce, Quantization-Pruned Attention, Structured Pruning, Normalized Linear Quantization (Prune and NLQ), Quantization and Pruning for Sentiment Analysis, an Automatic mixed-precision Quantization approach for BERT compression (AQ-BERT), Mixed Precision Quantization, Unstructured Pruning, and Quantization, and Magnitude Pruning. The datasets utilized, models employed, and outcomes achieved are taken into account within this research. The utilization of pruning and quantization techniques across diverse deep-learning tasks, NLP, and sentiment analysis are also discussed. Moreover, issues such as compatibility with hardware configurations, optimization complexities, accuracy degradation, and other constraints have been analysed. Several challenges and limitations of weight or unit pruning that are utilized for optimizing memory and quantization techniques to enhance precision are explored. The in-depth analysis of these state-of-the-art techniques and experiments provides a broad understanding. Furthermore, strategies to effectively reduce the computational and memory demands of neural networks without compromising their performance are also analysed.

1 April 2019 in Advanced Engineering Informatics
Verified
Journal Article
Closed access
YAYousaf Ali
Yousaf Ali
MAMuhammad Asees Awan
Muhammad Asees Awan
M QammarM Qammar
M Qammar
JKJehangir Khan
Jehangir Khan
Antonella PetrilloAntonella Petrillo
Antonella Petrillo
AKArshad Ali Khan
Arshad Ali Khan

24 September 2019 in The Engineering Economist
Verified
Journal Article
Closed access
YAYousaf Ali
Yousaf Ali
MSMuhammad Sabir
Muhammad Sabir
M QammarM Qammar
M Qammar
MAMehnab Ali
Mehnab Ali
AKArshad Ali Khan
Arshad Ali Khan
Pakistan Railways has faced a severe financial crisis in recent years. Pakistan has recently become a partner with China in a mega-investment project under an agreement called the China-Pakistan Economic Corridor (CPEC). Among other things, CPEC also includes a range of investments in Pakistan Railways. This particular study focuses on the analysis of US$8.2 billion investment in the upgrade and expansion of the Karachi-Peshawar railways link, which is also known as the ML-1 (Main Line 1). The study found ML-1 as economically viable with a payback period of 10 years. Furthermore, ML-1 project investment is expected to result in uplifting Pakistan Railways, mainly through an increase in freight and passenger transportation. Some risk factors may hinder the expected economic return from the CPEC investment in Pakistan Railways. These factors include consistency in the government policies, the status of the Pakistani economy in upcoming years, and law and order situations in the country. The study has a utility for the governments of both countries and larger business communities have stakes in the trade between the two countries. It is equally beneficial for the international community, businesses (both in China and Pakistan) and locals of the region associated with the CPEC infrastructure.

26 July 2016 in Communications in Computer and Information Science
Verified
Book Chapter
Closed access
PDF file
AKArshad Khan
Arshad Khan
TTThanassis Tiropanis
Thanassis Tiropanis
DMDavid Martin
David Martin

2 July 2025 in IEEE Access
Verified
Journal Article
Open access
SASyed Salman Alam
Syed Salman Alam
NSNajam Us Saqib
Najam Us Saqib
AKArshad Khan
Arshad Khan
WBWaqas Bukhari
Waqas Bukhari
RSRiad Souissi
Riad Souissi
Maintaining compliance with light intensity standards is critical to ensuring proper illumination, particularly during nighttime, and for preserving healthy ecosystems in urban and peri-urban areas. Traditional methods, relying on manual inspections with handheld sensors, are labour-intensive and inefficient. This paper presents a novel vehicle-mounted light intensity measurement system, the “Innovative Light Evaluation Rig” (InLer), designed to significantly improve streetlight monitoring compared to conventional methods. InLer operates as an autonomous IoT device with on-edge processing, leveraging strategically placed sensors to capture light intensity in both vertical and horizontal planes. Its ability to measure light intensity at large scale continuous monitoring in three directions simultaneously within a single sampling cycle significantly enhances operational efficiency. The system features an aerodynamic and mechanically robust design, incorporating three independent lux sensors, wireless connectivity, waterproof components, and magnetic mounts for easy installation. Additionally, InLer integrates GNSS/GPS technology for precise geolocation, enabling detailed spatial analysis when combined with light-intensity data. A comprehensive calibration analysis presented in this paper validates the system’s accuracy and consistency, establishing InLer as a state-of-the-art solution with wide-ranging applications across multiple sectors.

25 May 2026 in IEEE Access
Verified
Journal Article
Open access
SGSyeda Adan Fatima Gilani
Syeda Adan Fatima Gilani
MBMuhammad Fasih Uddin Butt
Muhammad Fasih Uddin Butt
ZKZirwa Khan
Zirwa Khan
AKArshad Ali Khan
Arshad Ali Khan
WAWadood Abdul
Wadood Abdul
Palmprint (PPR) and palmvein recognition (PVR) have emerged as reliable biometric modalities due to their rich structural information and resistance to impersonation. In this study, we design a dual-modality recognition pipeline that evaluates both palm-based traits using a combination of conventional machine learning (ML) classifiers and recent deep feature extractors. Classical models, including Support Vector Machines (SVM), K-Nearest Neighbors (KNN), Random Forests (RF), and the lightweight MobileNetV2 network, serve as strong baselines, producing accuracies of 98.8%, 98.8%, and 99.4% for PPR; and 95.5%, 97.5%, and 96.8% for PVR. To explore more expressive feature representations, we further employ EfficientNet-B3, Swin Transformer, and Vision Transformer (ViT). EfficientNet-B3 achieves a training accuracy of 94.8% and a validation accuracy 92.3%. The Swin Transformer further improves performance reaching 98.5% validation accuracy with a precision of 98.7%, an AUC-ROC of 0.996, and an inference time of 12 ms per image, indicating efficient inference performance. ViT also demonstrates stable validation performance under the adopted protocol, achieving 99.44% accuracy during training and 98.11% on validation data. To leverage the strengths of both CNN and Transformers models, a hybrid design is developed in which EfficientNet’s spatial feature maps are combined with Swin Transformer’s hierarchical attention features, resulting in more stable and discriminative representations. Finally, the complete recognition system is deployed through a Streamlit-based interface for real-time palm verification, demonstrating the feasibility of proposed approach under the adopted evaluation setting. Overall, the results indicate that merging traditional ML pipelines with contemporary CNN and Transformer-based approaches leads to strong closed-set identification performance of both PPR and PVR systems.