Sitemap
A list of all the posts and pages found on the site. For you robots out there is an XML version available for digesting as well.
Pages
Posts
Chat with Your Obsidian Notes Using the Falcon Mamba 7B Model
Published:
In the age of information, it’s easy to drown in the sea of notes we’ve meticulously collected over time. Imagine having a chat application that understands your notes and fetches the exact information you need, no matter how vast your repository is. Enter Falcon Mamba 7B and Retrieval-Augmented Generation (RAG) — a perfect blend of speed, precision, and language understanding that turns your note-taking chaos into order.
Balancing Regression Datasets with KNNOR-Reg Oversampling Technique
Published:
Enhancing model performance in machine learning often begins with addressing the quality of your data. One of the most challenging issues is the imbalance in the target variable distribution, which can severely impact the accuracy of regression models.
MemGPT: Assimilating Information from Multiple PDFs
Published:
In this post, I explore the capabilities of MemGPT in handling Retrieval-Augmented Generation (RAG) tasks, particularly focusing on its ability to assimilate information from multiple PDF files. While GPT-4 has its strengths, it often struggles with tasks requiring the consolidation of information from several documents. MemGPT, on the other hand, excels in this area, as demonstrated through a series of comparisons.
Advanced Retrieval with LlamaPacks: Elevating RAG in Fewer Lines of Code!
Published:
In this post, I explore five Retrieval-Augmented Generation (RAG) methods open-sourced by LlamaIndex, which claim to simplify the RAG process to just about one line of code. Through a series of experiments, I test these methods and compare their effectiveness in terms of accuracy and efficiency.
Super Quick: Retrieval Augmented Generation Using Ollama
Published:
In this post, I delve into the capabilities of Ollama, a powerful infrastructure that simplifies local execution of open-source models and interactions with PDFs. Ollama acts like a Docker system for Large Language Models (LLMs), allowing easy setup of local LLM servers, fine-tuning, and more.
Intel OneAPI — Detecting Weed in Crops
Published:
In this post, I discuss my participation in the Intel® oneAPI Hackathon for Open Innovation, where the challenge was to detect weeds in cropland images using computer vision and deep learning techniques. The goal of this project was to develop a solution that allows for the targeted application of pesticides, minimizing environmental impact while maximizing crop yield.
Monkey Pox Detection from Images
Published:
In this article, I explore the application of deep learning techniques to detect Monkeypox from images, a timely topic given the World Health Organization’s declaration of Monkeypox as a global health emergency. The project involves collecting and processing image data of various types of pox, including Chickenpox, Measles, and Monkeypox, to train models capable of distinguishing between these conditions.
Detecting Road Damages From Image And Video
Published:
In this post, I take you through the process of building an object detection model using YOLOv5, specifically for detecting road damages from images and videos. This project demonstrates how to leverage transfer learning by training a deep neural network on your own dataset, allowing the model to detect specific objects of interest, such as abnormalities in X-Ray images or road damages in surveillance footage.
Analyzing Satellite Images for Disasters
Published:
In this post, I delve into the development of deep neural network models designed to analyze satellite images for disaster detection. This project was part of the AWS Disaster Response Hackathon, aiming to create a system capable of identifying and categorizing disaster-stricken regions based on satellite imagery.
Tipping the Scales: A Novel Augmentation Technique for Imbalanced Data
Published:
In this article, I discuss the critical issue of imbalanced data in machine learning and introduce a novel augmentation technique designed to address this challenge. Even the most sophisticated models can fail if the underlying data is imbalanced, leading to misleadingly high accuracy rates while misclassifying minority cases.
Scraping Websites with Python, Selenium, and Tor: The Big Data Heist
Published:
In this post, I tackle the common challenges faced when scraping websites, particularly the frustration of being blocked after making too many consecutive requests. I explore how to use Python, Selenium, and Tor to bypass these limitations, enabling you to continue scraping without interruptions.
portfolio
AI-powered Resource Management Platform for Healthcare Providers
AI-Powered Solution to Improve Hospital Efficiency and Reduce Patient Readmissions
Context-Based Biased LLM (CB-LLM): Case of LLM for Palestine
Leveraging AI for Social Justice and Personalized Education
publications
Multi-Cluster Jumping Particle Swarm Optimization for Fast Convergence
Published in IEEE Access, 2020
This paper introduces the Multi-Cluster Jumping Particle Swarm Optimization (PSO) algorithm, designed to address the limitations of classical PSO by improving convergence speed and avoiding local minima, particularly in high-dimensional data spaces.
Recommended citation: Rehman, A. U., Islam, A., & Belhaouari, S. B. (2020). "Multi-Cluster Jumping Particle Swarm Optimization for Fast Convergence." IEEE Access, 8, 189382--189394. https://doi.org/10.1109/ACCESS.2020.3031003
Download Paper | Download Slides
Class Aware Autoencoders for Better Feature Extraction
Published in 2021 International Conference on Electrical, Communication, and Computer Engineering (ICECCE), 2021
This paper introduces a modified operation of autoencoders that incorporates class labels into the learning process, resulting in improved feature extraction and higher classification accuracy across multiple datasets.
Recommended citation: Islam, A., & Belhaouari, S. B. (2021). "Class Aware Autoencoders for Better Feature Extraction." In 2021 International Conference on Electrical, Communication, and Computer Engineering (ICECCE) (pp. 1--5). IEEE. https://doi.org/10.1109/ICECCE52056.2021.9514202
Download Paper
Cyber-Physical System Demonstration of an Automated Shuttle-Conveyor-Belt Operation for Inventory Control of Multiple Stockpiles: A Proof of Concept
Published in IEEE Access, 2022
This paper presents a cyber-physical system (CPS) demonstration in the mining industry, showcasing an automated shuttle-conveyor-belt operation for managing multiple stockpiles. The CPS autonomously controls inventory using mixed-integer optimization and a deep neural network, providing a proof of concept for smart manufacturing in Industry 4.0.
Recommended citation: Yaqot, M., Franzoi, R. E., Islam, A., & Menezes, B. C. (2022). "Cyber-Physical System Demonstration of an Automated Shuttle-Conveyor-Belt Operation for Inventory Control of Multiple Stockpiles: A Proof of Concept." IEEE Access, 10, 127636--127653. https://doi.org/10.1109/ACCESS.2022.3226942
Download Paper | Download Slides
KNNOR: An Oversampling Technique for Imbalanced Datasets
Published in Applied Soft Computing, 2022
This paper introduces KNNOR, a novel oversampling technique that addresses class imbalance in datasets. The method focuses on enhancing the predictive performance of ML models by ensuring a more reliable augmentation of the minority class, overcoming issues such as within-class imbalance and the small disjunct problem.
Recommended citation: Islam, A., Belhaouari, S. B., Rehman, A. U., & Bensmail, H. (2022). "KNNOR: An Oversampling Technique for Imbalanced Datasets." Applied Soft Computing, 115, 108288. https://doi.org/10.1016/j.asoc.2021.108288
Download Paper | Download Slides
K Nearest Neighbor OveRsampling Approach: An Open Source Python Package for Data Augmentation
Published in Software Impacts, 2022
This paper introduces the K Nearest Neighbor OveRsampling (KNNOR) algorithm, a novel data augmentation technique implemented as an open-source Python package. The algorithm addresses the challenges of imbalanced datasets by generating artificial data points that enhance classifier accuracy without adding noise.
Recommended citation: Islam, A., Belhaouari, S. B., Rehman, A. U., & Bensmail, H. (2022). "K Nearest Neighbor OveRsampling Approach: An Open Source Python Package for Data Augmentation." Software Impacts, 12, 100272. https://doi.org/10.1016/j.simpa.2022.100272
Download Paper | Download Slides
Smart Pruning of Deep Neural Networks Using Curve Fitting and Evolution of Weights
Published in International Conference on Machine Learning, Optimization, and Data Science, 2022
This paper presents innovative methods for pruning deep neural networks, making them lighter and faster while maintaining accuracy, which is crucial for their deployment on embedded and edge devices.
Recommended citation: Islam, A., & Belhaouari, S. B. (2022). "Smart Pruning of Deep Neural Networks Using Curve Fitting and Evolution of Weights." In International Conference on Machine Learning, Optimization, and Data Science (pp. 62--76). Springer.
Download Paper
Analysing Keystroke Dynamics Using Wavelet Transforms
Published in 2022 IEEE International Carnahan Conference on Security Technology (ICCST), 2022
This paper presents a novel method for analyzing keystroke dynamics using wavelet transforms, enhancing the robustness of classifiers in detecting unauthorized smartphone access and offering a potential alternative to costly authentication methods.
Recommended citation: Islam, A., & Belhaouari, S. B. (2022). "Analysing Keystroke Dynamics Using Wavelet Transforms." In 2022 IEEE International Carnahan Conference on Security Technology (ICCST) (pp. 1--5). IEEE. https://doi.org/10.1109/ICCST52959.2022.9896483
Download Paper
IJCB 2022 Mobile Behavioral Biometrics Competition (MobileB2C)
Published in 2022 IEEE International Joint Conference on Biometrics (IJCB), 2022
This paper outlines the framework and findings of the IJCB 2022 Mobile Behavioral Biometrics Competition (MobileB2C), which benchmarks mobile user authentication systems based on behavioral biometric traits.
Recommended citation: Stragapede, G., Vera-Rodriguez, R., Tolosana, R., Morales, A., Fierrez, J., Ortega-Garcia, J., Rasnayaka, S., Seneviratne, S., Dissanayake, V., Liebers, J., et al. (2022). "IJCB 2022 Mobile Behavioral Biometrics Competition (MobileB2C)." In 2022 IEEE International Joint Conference on Biometrics (IJCB) (pp. 1--7). IEEE. https://doi.org/10.1109/IJCB54206.2022.10007985
Download Paper
Framework Design for Similar Video Detection: A Graph-Based Video Clustering Approach
Published in 2022 International Symposium on Multidisciplinary Studies and Innovative Technologies (ISMSIT), 2022
This paper presents a graph-based video clustering approach for detecting similar videos, offering a novel strategy to identify and manage duplicated content efficiently.
Recommended citation: Al-Thani, N. F., Islam, A., Belhaouari, S. B., & Faramarzinia, S. (2022). "Framework Design for Similar Video Detection: A Graph-Based Video Clustering Approach." In 2022 International Symposium on Multidisciplinary Studies and Innovative Technologies (ISMSIT) (pp. 571--576). IEEE. https://doi.org/10.1109/ISMSIT56059.2022.9932834
Download Paper
Fast and Efficient Image Generation Using Variational Autoencoders and K-Nearest Neighbor OveRsampling Approach
Published in IEEE Access, 2023
This paper presents a novel method for generating artificial images by combining Variational Autoencoders (VAEs) with the K-Nearest Neighbor OveRsampling (KNNOR) approach. The technique addresses common issues in Generative Adversarial Networks (GANs), such as mode collapse and distortion, producing more realistic and efficient image generation.
Recommended citation: Islam, A., & Belhaouari, S. B. (2023). "Fast and Efficient Image Generation Using Variational Autoencoders and K-Nearest Neighbor OveRsampling Approach." IEEE Access, 11, 28416--28426. https://doi.org/10.1109/ACCESS.2023.3259236
Download Paper | Download Slides
Neural Network Optimization with Weight Evolution
Published in ICML 2023 Workshop Neural Compression: From Information Theory to Applications, 2023
This paper presents a novel approach to neural network optimization through weight evolution, achieving higher compression with minimal loss of accuracy compared to traditional magnitude pruning methods.
Recommended citation: Belhaouari, S. B., & Islam, A. (2023). "Neural Network Optimization with Weight Evolution." In ICML 2023 Workshop Neural Compression: From Information Theory to Applications.
Download Paper
Pushing Boundaries: Exploring Zero-Shot Object Classification with Large Multimodal Models
Published in 2023 Tenth International Conference on Social Networks Analysis, Management and Security (SNAMS), 2023
This paper investigates the efficacy of Large Multimodal Models (LMMs) in zero-shot object classification tasks, demonstrating their potential in achieving high accuracy across diverse datasets without fine-tuning.
Recommended citation: Islam, A., Biswas, M. R., Zaghouani, W., Belhaouari, S. B., & Shah, Z. (2023). "Pushing Boundaries: Exploring Zero-Shot Object Classification with Large Multimodal Models." In 2023 Tenth International Conference on Social Networks Analysis, Management and Security (SNAMS) (pp. 1--5). IEEE. https://doi.org/10.1109/SNAMS60348.2023.10375440
Download Paper
Oversampling Techniques for Imbalanced Data in Regression
Published in Expert Systems with Applications, 2024
This paper discusses novel oversampling techniques tailored specifically for handling imbalanced data in regression tasks. These methods improve predictive accuracy by addressing issues unique to regression models, such as skewed distributions and variance within minority data points.
Recommended citation: Belhaouari, S. B., Islam, A., Kassoul, K., Al-Fuqaha, A., & Bouzerdoum, A. (2024). "Oversampling Techniques for Imbalanced Data in Regression." Expert Systems with Applications, 252, 124118. https://doi.org/10.1016/j.eswa.2024.124118
Download Paper | Download Slides
Introducing Novel Radon-Based Transform for Disease Detection From Chest X-Ray Images
Published in 2024 6th International Conference on Pattern Analysis and Intelligent Systems (PAIS), 2024
This paper presents a novel Radon-based transform, named RadEx, for enhancing feature extraction in chest X-ray images, significantly improving the accuracy of lung disease detection models.
Recommended citation: Islam, A., Mohsen, F., Shah, Z., & Belhaouari, S. B. (2024). "Introducing Novel Radon-Based Transform for Disease Detection From Chest X-Ray Images." In 2024 6th International Conference on Pattern Analysis and Intelligent Systems (PAIS) (pp. 1--5). IEEE. https://doi.org/10.1109/PAIS62114.2024.10541204
Download Paper
talks
Pruning and Optimizing Large Language Models in an Era of GPU Scarcity
Published:
This talk, presented at the ICAI’24 as part of the 2024 World Congress in Computer Science, Computer Engineering, & Applied Computing (CSCE’24), focuses on the critical issue of optimizing large language models (LLMs) during a time of GPU scarcity. The talk discusses novel pruning techniques, including “evolution of weights” and “smart pruning,” aimed at reducing the computational and environmental costs associated with training and deploying these models.
teaching
Teaching Assistant
Graduate course, Hamad Bin Khalifa University, College Of Science & Engineering, 2022
Machine Learning Course Support
Deep Fake Detection Using CNN
Personal Course, Remote, 2024
Week 1: Introduction to Deep Learning and Deepfakes
Topics:
- Overview of deep learning, neural networks, and their applications
- Introduction to deepfakes: history, types, and impact on society
- How AI generates deepfakes using GANs (Generative Adversarial Networks)
Homework/Practice:
- Watch introductory videos on deep learning
- Read articles on GANs and deepfake ethics