In this work, I participated in the development of an Intrusion Detection System (IDS) that monitors network traffic and system activity, identifying security threats and suspicious behaviour. It protects against unauthorised access, attacks, anomalies or policy violations. The system is based on machine learning techniques.
In this work, a secure P2P chat application will be proposed in the Java programming language. The system consists of a robust, fully decentralized (P2P) and end-to-end encrypted network architecture.
This work focuses on using Vision Transformer (ViT) models to detect diabetic retinopathy in retinal images. The model is trained on a large dataset from Kaggle, using data mining and pre-processing techniques for reliable results. Traditional machine learning methods are also explored to improve detection. The goal is to create a scalable solution for timely diagnoses, improving patient outcomes in managing diabetic retinopathy.
This project aims to develop an intelligent system that combines a chatbot interface with an efficient regression model to predict house prices using machine learning techniques. In addition, the project incorporates the use of the LLMs API (gpt 3.5 turbo), in order to improve the chatbot’s conversational capabilities.
Study applies deep learning for better speech recognition in noise. CNN achieved high accuracy in clean audio but struggled with noise. SEGAN proposed as a solution for improved robustness.
This work highlights predicting discharge modes in DBD plasma systems using machine learning. Parameters affecting the discharge include voltage, gas, and electrode configuration. It explores data collection challenges and aims to improve prediction accuracy.
This study optimizes investments in NVIDIA Corporation using stock analysis and predictive modeling. Data preprocessing and analysis are performed, followed by four modeling techniques: logistic regression, PCA, KNN, and LSTM. The LSTM model captures temporal dependencies for time series prediction. Results and investment strategies for NVIDIA stock are presented.
This work presents an innovative approach to teaching the Tifinagh script using AI technology. The platform utilizes handwriting recognition with a dataset of 82500 images, achieving 99.81%-99.93% accuracy. It also includes a web-based learning platform with real-time feedback.
Introducing Tifinagh-MNIST: 82.5k grayscale images (28x28 pixels) of handwritten Tifinagh letters used in Tamazight languages. 33 classes with 2.5k images each. Training set: 66k images. Test set: 16.5k images. AI tools development, classification, and data generation applications. Publicly available to advance AI for Tamazight text processing.
A blockchain system to manage monetary transactions between the different nodes of a decentralized network.