-
Lightweight Clinical Decision Support System using QLoRA-Fine-Tuned LLMs and Retrieval-Augmented Generation
Authors:
Mohammad Shoaib Ansari,
Mohd Sohail Ali Khan,
Shubham Revankar,
Aditya Varma,
Anil S. Mokhade
Abstract:
This research paper investigates the application of Large Language Models (LLMs) in healthcare, specifically focusing on enhancing medical decision support through Retrieval-Augmented Generation (RAG) integrated with hospital-specific data and fine-tuning using Quantized Low-Rank Adaptation (QLoRA). The system utilizes Llama 3.2-3B-Instruct as its foundation model. By embedding and retrieving cont…
▽ More
This research paper investigates the application of Large Language Models (LLMs) in healthcare, specifically focusing on enhancing medical decision support through Retrieval-Augmented Generation (RAG) integrated with hospital-specific data and fine-tuning using Quantized Low-Rank Adaptation (QLoRA). The system utilizes Llama 3.2-3B-Instruct as its foundation model. By embedding and retrieving context-relevant healthcare information, the system significantly improves response accuracy. QLoRA facilitates notable parameter efficiency and memory optimization, preserving the integrity of medical information through specialized quantization techniques. Our research also shows that our model performs relatively well on various medical benchmarks, indicating that it can be used to make basic medical suggestions. This paper details the system's technical components, including its architecture, quantization methods, and key healthcare applications such as enhanced disease prediction from patient symptoms and medical history, treatment suggestions, and efficient summarization of complex medical reports. We touch on the ethical considerations-patient privacy, data security, and the need for rigorous clinical validation-as well as the practical challenges of integrating such systems into real-world healthcare workflows. Furthermore, the lightweight quantized weights ensure scalability and ease of deployment even in low-resource hospital environments. Finally, the paper concludes with an analysis of the broader impact of LLMs on healthcare and outlines future directions for LLMs in medical settings.
△ Less
Submitted 6 May, 2025;
originally announced May 2025.
-
FireLite: Leveraging Transfer Learning for Efficient Fire Detection in Resource-Constrained Environments
Authors:
Mahamudul Hasan,
Md Maruf Al Hossain Prince,
Mohammad Samar Ansari,
Sabrina Jahan,
Abu Saleh Musa Miah,
Jungpil Shin
Abstract:
Fire hazards are extremely dangerous, particularly in sectors such as the transportation industry, where political unrest increases the likelihood of their occurrence. By employing IP cameras to facilitate the setup of fire detection systems on transport vehicles, losses from fire events may be prevented proactively. However, the development of lightweight fire detection models is required due to…
▽ More
Fire hazards are extremely dangerous, particularly in sectors such as the transportation industry, where political unrest increases the likelihood of their occurrence. By employing IP cameras to facilitate the setup of fire detection systems on transport vehicles, losses from fire events may be prevented proactively. However, the development of lightweight fire detection models is required due to the computational constraints of the embedded systems within these cameras. We introduce FireLite, a low-parameter convolutional neural network (CNN) designed for quick fire detection in contexts with limited resources, in response to this difficulty. With an accuracy of 98.77\%, our model -- which has just 34,978 trainable parameters achieves remarkable performance numbers. It also shows a validation loss of 8.74 and peaks at 98.77 for precision, recall, and F1-score measures. Because of its precision and efficiency, FireLite is a promising solution for fire detection in resource-constrained environments.
△ Less
Submitted 30 September, 2024;
originally announced September 2024.
-
ResEmoteNet: Bridging Accuracy and Loss Reduction in Facial Emotion Recognition
Authors:
Arnab Kumar Roy,
Hemant Kumar Kathania,
Adhitiya Sharma,
Abhishek Dey,
Md. Sarfaraj Alam Ansari
Abstract:
The human face is a silent communicator, expressing emotions and thoughts through its facial expressions. With the advancements in computer vision in recent years, facial emotion recognition technology has made significant strides, enabling machines to decode the intricacies of facial cues. In this work, we propose ResEmoteNet, a novel deep learning architecture for facial emotion recognition desi…
▽ More
The human face is a silent communicator, expressing emotions and thoughts through its facial expressions. With the advancements in computer vision in recent years, facial emotion recognition technology has made significant strides, enabling machines to decode the intricacies of facial cues. In this work, we propose ResEmoteNet, a novel deep learning architecture for facial emotion recognition designed with the combination of Convolutional, Squeeze-Excitation (SE) and Residual Networks. The inclusion of SE block selectively focuses on the important features of the human face, enhances the feature representation and suppresses the less relevant ones. This helps in reducing the loss and enhancing the overall model performance. We also integrate the SE block with three residual blocks that help in learning more complex representation of the data through deeper layers. We evaluated ResEmoteNet on four open-source databases: FER2013, RAF-DB, AffectNet-7 and ExpW, achieving accuracies of 79.79%, 94.76%, 72.39% and 75.67% respectively. The proposed network outperforms state-of-the-art models across all four databases. The source code for ResEmoteNet is available at https://github.com/ArnabKumarRoy02/ResEmoteNet.
△ Less
Submitted 23 November, 2024; v1 submitted 1 September, 2024;
originally announced September 2024.
-
A Cutting-Edge Deep Learning Method For Enhancing IoT Security
Authors:
Nadia Ansar,
Mohammad Sadique Ansari,
Mohammad Sharique,
Aamina Khatoon,
Md Abdul Malik,
Md Munir Siddiqui
Abstract:
There have been significant issues given the IoT, with heterogeneity of billions of devices and with a large amount of data. This paper proposed an innovative design of the Internet of Things (IoT) Environment Intrusion Detection System (or IDS) using Deep Learning-integrated Convolutional Neural Networks (CNN) and Long Short-Term Memory (LSTM) networks. Our model, based on the CICIDS2017 dataset,…
▽ More
There have been significant issues given the IoT, with heterogeneity of billions of devices and with a large amount of data. This paper proposed an innovative design of the Internet of Things (IoT) Environment Intrusion Detection System (or IDS) using Deep Learning-integrated Convolutional Neural Networks (CNN) and Long Short-Term Memory (LSTM) networks. Our model, based on the CICIDS2017 dataset, achieved an accuracy of 99.52% in classifying network traffic as either benign or malicious. The real-time processing capability, scalability, and low false alarm rate in our model surpass some traditional IDS approaches and, therefore, prove successful for application in today's IoT networks. The development and the performance of the model, with possible applications that may extend to other related fields of adaptive learning techniques and cross-domain applicability, are discussed. The research involving deep learning for IoT cybersecurity offers a potent solution for significantly improving network security.
△ Less
Submitted 18 June, 2024;
originally announced June 2024.
-
Exploration of Activation Fault Reliability in Quantized Systolic Array-Based DNN Accelerators
Authors:
Mahdi Taheri,
Natalia Cherezova,
Mohammad Saeed Ansari,
Maksim Jenihhin,
Ali Mahani,
Masoud Daneshtalab,
Jaan Raik
Abstract:
The stringent requirements for the Deep Neural Networks (DNNs) accelerator's reliability stand along with the need for reducing the computational burden on the hardware platforms, i.e. reducing the energy consumption and execution time as well as increasing the efficiency of DNN accelerators. Moreover, the growing demand for specialized DNN accelerators with tailored requirements, particularly for…
▽ More
The stringent requirements for the Deep Neural Networks (DNNs) accelerator's reliability stand along with the need for reducing the computational burden on the hardware platforms, i.e. reducing the energy consumption and execution time as well as increasing the efficiency of DNN accelerators. Moreover, the growing demand for specialized DNN accelerators with tailored requirements, particularly for safety-critical applications, necessitates a comprehensive design space exploration to enable the development of efficient and robust accelerators that meet those requirements. Therefore, the trade-off between hardware performance, i.e. area and delay, and the reliability of the DNN accelerator implementation becomes critical and requires tools for analysis. This paper presents a comprehensive methodology for exploring and enabling a holistic assessment of the trilateral impact of quantization on model accuracy, activation fault reliability, and hardware efficiency. A fully automated framework is introduced that is capable of applying various quantization-aware techniques, fault injection, and hardware implementation, thus enabling the measurement of hardware parameters. Moreover, this paper proposes a novel lightweight protection technique integrated within the framework to ensure the dependable deployment of the final systolic-array-based FPGA implementation. The experiments on established benchmarks demonstrate the analysis flow and the profound implications of quantization on reliability, hardware performance, and network accuracy, particularly concerning the transient faults in the network's activations.
△ Less
Submitted 17 January, 2024;
originally announced January 2024.
-
A Novel Falling-Ball Algorithm for Image Segmentation
Authors:
Asra Aslam,
Ekram Khan,
Mohammad Samar Ansari,
M. M. Sufyan Beg
Abstract:
Image segmentation refers to the separation of objects from the background, and has been one of the most challenging aspects of digital image processing. Practically it is impossible to design a segmentation algorithm which has 100% accuracy, and therefore numerous segmentation techniques have been proposed in the literature, each with certain limitations. In this paper, a novel Falling-Ball algor…
▽ More
Image segmentation refers to the separation of objects from the background, and has been one of the most challenging aspects of digital image processing. Practically it is impossible to design a segmentation algorithm which has 100% accuracy, and therefore numerous segmentation techniques have been proposed in the literature, each with certain limitations. In this paper, a novel Falling-Ball algorithm is presented, which is a region-based segmentation algorithm, and an alternative to watershed transform (based on waterfall model). The proposed algorithm detects the catchment basins by assuming that a ball falling from hilly terrains will stop in a catchment basin. Once catchment basins are identified, the association of each pixel with one of the catchment basin is obtained using multi-criterion fuzzy logic. Edges are constructed by dividing image into different catchment basins with the help of a membership function. Finally closed contour algorithm is applied to find closed regions and objects within closed regions are segmented using intensity information. The performance of the proposed algorithm is evaluated both objectively as well as subjectively. Simulation results show that the proposed algorithms gives superior performance over conventional Sobel edge detection methods and the watershed segmentation algorithm. For comparative analysis, various comparison methods are used for demonstrating the superiority of proposed methods over existing segmentation methods.
△ Less
Submitted 12 May, 2021; v1 submitted 6 May, 2021;
originally announced May 2021.
-
A Single-Layer Asymmetric RNN: Potential Low Hardware Complexity Linear Equation Solver
Authors:
Mohammad Samar Ansari
Abstract:
A single layer neural network for the solution of linear equations is presented. The proposed circuit is based on the standard Hopfield model albeit with the added flexibility that the interconnection weight matrix need not be symmetric. This results in an asymmetric Hopfield neural network capable of solving linear equations. PSPICE simulation results are given which verify the theoretical predic…
▽ More
A single layer neural network for the solution of linear equations is presented. The proposed circuit is based on the standard Hopfield model albeit with the added flexibility that the interconnection weight matrix need not be symmetric. This results in an asymmetric Hopfield neural network capable of solving linear equations. PSPICE simulation results are given which verify the theoretical predictions. Experimental results for circuits set up to solve small problems further confirm the operation of the proposed circuit.
△ Less
Submitted 4 May, 2021; v1 submitted 1 May, 2021;
originally announced May 2021.
-
A Survey of Modern Deep Learning based Object Detection Models
Authors:
Syed Sahil Abbas Zaidi,
Mohammad Samar Ansari,
Asra Aslam,
Nadia Kanwal,
Mamoona Asghar,
Brian Lee
Abstract:
Object Detection is the task of classification and localization of objects in an image or video. It has gained prominence in recent years due to its widespread applications. This article surveys recent developments in deep learning based object detectors. Concise overview of benchmark datasets and evaluation metrics used in detection is also provided along with some of the prominent backbone archi…
▽ More
Object Detection is the task of classification and localization of objects in an image or video. It has gained prominence in recent years due to its widespread applications. This article surveys recent developments in deep learning based object detectors. Concise overview of benchmark datasets and evaluation metrics used in detection is also provided along with some of the prominent backbone architectures used in recognition tasks. It also covers contemporary lightweight classification models used on edge devices. Lastly, we compare the performances of these architectures on multiple metrics.
△ Less
Submitted 12 May, 2021; v1 submitted 24 April, 2021;
originally announced April 2021.
-
DMR-based Technique for Fault Tolerant AES S-box Architecture
Authors:
Mahdi Taheri,
Saeideh Sheikhpour,
Mohammad Saeed Ansari,
Ali Mahani
Abstract:
This paper presents a high-throughput fault-resilient hardware implementation of AES S-box, called HFS-box. If a transient natural or even malicious fault in each pipeline stage is detected, the corresponding error signal becomes high and as a result, the control unit holds the output of our proposed DMR voter till the fault effect disappears. The proposed low-cost HFS-box provides a high capabili…
▽ More
This paper presents a high-throughput fault-resilient hardware implementation of AES S-box, called HFS-box. If a transient natural or even malicious fault in each pipeline stage is detected, the corresponding error signal becomes high and as a result, the control unit holds the output of our proposed DMR voter till the fault effect disappears. The proposed low-cost HFS-box provides a high capability of fault-tolerant against transient faults with any duration by putting low area overhead, i.e. 137%, and low throughput degradation, i.e. 11.3%, on the original implementation.
△ Less
Submitted 11 September, 2020;
originally announced September 2020.
-
Tethered Balloon Technology for Green Communication in Smart Cities and Healthy Environment
Authors:
S. H. Alsamhi,
Mohd Samar Ansari,
Liang Zhao,
Sau Nguyen Van,
S. K. Gupta,
Amr A. Alammari,
Akram Hatem Saber,
Mohammed Y. A. M. Hebah,
Marwan Ahmed Abdullah Alasali,
Hasan Mohsen Aljabali,
Mohd Najim,
Ashutosh Srivastava
Abstract:
The development and adopting of advanced communication technologies provide mobile users more convenience to connect any wireless network anytime and anywhere. Therefore, a large number of base stations (BS) are demanded keeping users connectivity, enhancing network capacity, and guarantee a sustained users Quality of Experiences (QoS). However, increasing the number of BS leads to an increase in…
▽ More
The development and adopting of advanced communication technologies provide mobile users more convenience to connect any wireless network anytime and anywhere. Therefore, a large number of base stations (BS) are demanded keeping users connectivity, enhancing network capacity, and guarantee a sustained users Quality of Experiences (QoS). However, increasing the number of BS leads to an increase in the ecological ad radiation hazards. In order to green communication, many factors should be taken into consideration, i.e., saving energy, guarantee QoS, and reducing pollution hazards. Therefore, we propose tethered balloon technology that can replace a large number of BS and reduce ecological and radiation hazards due to its high altitude and feasible green and healthy broadband communication. The main contribution of this paper is to deploy tethered balloon technology at different altitude and measure the power density. Furthermore, we evaluate the measurement of power density from different height of tethered balloon comparison with traditional wireless communication technologies. The simulation results showed that tethered balloon technology can deliver green communication effectively and efficiently without any hazardous impacts.
△ Less
Submitted 24 December, 2019;
originally announced December 2019.
-
FireNet: A Specialized Lightweight Fire & Smoke Detection Model for Real-Time IoT Applications
Authors:
Arpit Jadon,
Mohd. Omama,
Akshay Varshney,
Mohammad Samar Ansari,
Rishabh Sharma
Abstract:
Fire disasters typically result in lot of loss to life and property. It is therefore imperative that precise, fast, and possibly portable solutions to detect fire be made readily available to the masses at reasonable prices. There have been several research attempts to design effective and appropriately priced fire detection systems with varying degrees of success. However, most of them demonstrat…
▽ More
Fire disasters typically result in lot of loss to life and property. It is therefore imperative that precise, fast, and possibly portable solutions to detect fire be made readily available to the masses at reasonable prices. There have been several research attempts to design effective and appropriately priced fire detection systems with varying degrees of success. However, most of them demonstrate a trade-off between performance and model size (which decides the model's ability to be installed on portable devices). The work presented in this paper is an attempt to deal with both the performance and model size issues in one design. Toward that end, a `designed-from-scratch' neural network, named FireNet, is proposed which is worthy on both the counts: (i) it has better performance than existing counterparts, and (ii) it is lightweight enough to be deploy-able on embedded platforms like Raspberry Pi. Performance evaluations on a standard dataset, as well as our own newly introduced custom-compiled fire dataset, are extremely encouraging.
△ Less
Submitted 4 September, 2019; v1 submitted 28 May, 2019;
originally announced May 2019.
-
Depth-Map Generation using Pixel Matching in Stereoscopic Pair of Images
Authors:
Asra Aslam,
Mohd. Samar Ansari
Abstract:
Modern day multimedia content generation and dissemination is moving towards the presentation of more and more `realistic' scenarios. The switch from 2-dimensional (2D) to 3-dimensional (3D) has been a major driving force in that direction. Over the recent past, a large number of approaches have been proposed for creating 3D images/videos most of which are based on the generation of depth-maps. Th…
▽ More
Modern day multimedia content generation and dissemination is moving towards the presentation of more and more `realistic' scenarios. The switch from 2-dimensional (2D) to 3-dimensional (3D) has been a major driving force in that direction. Over the recent past, a large number of approaches have been proposed for creating 3D images/videos most of which are based on the generation of depth-maps. This paper presents a new algorithm for obtaining depth information pertaining to a depicted scene from a set of available pair of stereoscopic images. The proposed algorithm performs a pixel-to-pixel matching of the two images in the stereo pair for estimation of depth. It is shown that the obtained depth-maps show improvements over the reported counterparts.
△ Less
Submitted 15 May, 2019; v1 submitted 9 February, 2019;
originally announced February 2019.
-
Predictive Estimation of the Optimal Signal Strength from Unmanned Aerial Vehicle over Internet of Things Using ANN
Authors:
S. H. Alsamhi,
Ou Ma,
M. S. Ansari
Abstract:
This paper proposes an intelligent technique for maximizing the network connectivity and provisioning desired quality of service (QoS) of integration of internet of things (IoT) and unmanned aerial vehicle (UAV). Prediction of the signal strength and fading channel conditions enable adaptive data transmission which turn enhances the QoS for the end users/ devices with reducing the power consumptio…
▽ More
This paper proposes an intelligent technique for maximizing the network connectivity and provisioning desired quality of service (QoS) of integration of internet of things (IoT) and unmanned aerial vehicle (UAV). Prediction of the signal strength and fading channel conditions enable adaptive data transmission which turn enhances the QoS for the end users/ devices with reducing the power consumption for data transmissions. UAV is data gathering robot from the difficult or impossible area for humans to reach. Hence, Atmospheric dynamics and environment influence the signal strength during traveling in space among UAV, IoT devices, and humankind. Therefore, Signal moving from the smart UAV is sensitive to the effects of attenuation, reflection, diffraction, scattering, and shadowing. We analysis the ability ANN to predictively estimate the signal strength and channel propagation from the drone and physical medium parameters, using ANN. Moreover, the results show that the distortion of the signal can be reduced and enhanced significantly.
△ Less
Submitted 19 May, 2018;
originally announced May 2018.
-
Greening Internet of Things for Smart Everythings with A Green-Environment Life: A Survey and Future Prospects
Authors:
S. H. Alsamhi,
Ou Ma,
M. Samar Ansari,
Qingliang Meng
Abstract:
Tremendous technology development in the field of Internet of Things (IoT) has changed the way we work and live. Although the numerous advantages of IoT are enriching our society, it should be reminded that the IoT also consumes energy, embraces toxic pollution and E-waste. These place new stress on the environments and smart world. In order to increase the benefits and reduce the harm of IoT, the…
▽ More
Tremendous technology development in the field of Internet of Things (IoT) has changed the way we work and live. Although the numerous advantages of IoT are enriching our society, it should be reminded that the IoT also consumes energy, embraces toxic pollution and E-waste. These place new stress on the environments and smart world. In order to increase the benefits and reduce the harm of IoT, there is an increasing desire to move toward green IoT. Green IoT is seen as the future of IoT that is environmentally friendly. To achieve that, it is necessary to put a lot of measures to reduce carbon footprint, conserve fewer resources, and promote efficient techniques for energy usage. It is the reason for moving towards green IoT, where the machines, communications, sensors, clouds, and internet are alongside energy efficiency and reducing carbon emission. This paper presents a thorough survey of the current on-going research work and potential technologies of green IoT with an intention to provide some clues for future green IoT research.
△ Less
Submitted 2 May, 2018;
originally announced May 2018.
-
Artificial Intelligence-Based Techniques for Emerging Robotics Communication: A Survey and Future Perspectives
Authors:
S. H. Alsamhi,
Ou Ma,
M. S. Ansari
Abstract:
This paper reviews the current development of artificial intelligence (AI) techniques for the application area of robot communication. The study of the control and operation of multiple robots collaboratively toward a common goal is fast growing. Communication among members of a robot team and even including humans is becoming essential in many real-world applications. The survey focuses on the AI…
▽ More
This paper reviews the current development of artificial intelligence (AI) techniques for the application area of robot communication. The study of the control and operation of multiple robots collaboratively toward a common goal is fast growing. Communication among members of a robot team and even including humans is becoming essential in many real-world applications. The survey focuses on the AI techniques for robot communication to enhance the communication capability of the multi-robot team, making more complex activities, taking an appreciated decision, taking coordinated action, and performing their tasks efficiently.
△ Less
Submitted 28 April, 2018; v1 submitted 25 April, 2018;
originally announced April 2018.
-
Origin of Periodic Modulations in the Transient Reflectivity Signal at Cryogenic Temperatures
Authors:
Salahuddin Khan,
Rama Chari,
J. Jayabalan,
Suparna Pal,
T. K. Sharma,
A. K. Sagar,
M. S. Ansari,
P. K. Kush
Abstract:
Periodic modulations that appear in the low-temperature transient reflectivity signal of a GaAsP/AlGaAs single quantum well is studied. Similar anomalous oscillations are also observed in layered manganite [K. Kouyama et.al. J. Phys. Soc. Jpn. 76:123702(1-3), 2007]. We show that such periodic modulations are caused by changes in the linear reflectivity of the sample during transient reflectivity m…
▽ More
Periodic modulations that appear in the low-temperature transient reflectivity signal of a GaAsP/AlGaAs single quantum well is studied. Similar anomalous oscillations are also observed in layered manganite [K. Kouyama et.al. J. Phys. Soc. Jpn. 76:123702(1-3), 2007]. We show that such periodic modulations are caused by changes in the linear reflectivity of the sample during transient reflectivity measurements. Studied carried out on reflectivity of different materials under identical conditions shows that these modulations on the true transient reflectivity signal are caused by condensation of residual gases on the surface of quantum well. Methods to obtain reliable transient reflectivity data are also described.
△ Less
Submitted 28 August, 2012;
originally announced August 2012.
-
Extracting Work From A Single Heat Bath
Authors:
Arbab A. Khan,
M. S. Ansari
Abstract:
We present here a machine that is capable of extracting work from a single heat bath. Although no significant temperature gradient is involved in the operation of the machine, yet the Carnot efficiency as high as one is achievable. Working of the machine is explained on the basis of a demon suggested by Maxwell. Utilizing the kinetic energy spectrum of the molecules in solution, the demon can se…
▽ More
We present here a machine that is capable of extracting work from a single heat bath. Although no significant temperature gradient is involved in the operation of the machine, yet the Carnot efficiency as high as one is achievable. Working of the machine is explained on the basis of a demon suggested by Maxwell. Utilizing the kinetic energy spectrum of the molecules in solution, the demon can send "hotter" molecules to a higher gravitational potential at the expense of their own energies. Difference in chemical potentials due to concentration gradients and use of semi-permeable membranes ensure the continuing cyclic process.
△ Less
Submitted 2 December, 2004;
originally announced December 2004.
