This repository hosts various projects focusing on applying machine learning techniques to biomedical data. Each sub-directory contains a different project, and below, you will find a brief explanation for each.
In the electrocardiogram directory, we explore machine learning techniques for processing time series data, with a particular focus on electrocardiograms. The project encompasses the application of different machine learning models to analyze, classify, and understand ECG signals, which are crucial for detecting various cardiac conditions.
The molecules directory introduces the use of the RDKit library for handling and visualizing molecular structures and their different representations. This section also features projects that leverage Graph Neural Networks (GNNs) using PyTorch Geometric for molecule classification and analysis.
The genomic directory hosts a project that entails the development of a binary classifier for DNA sequences using 1D convolutional neural networks. This classifier is designed to analyze and predict specific features or characteristics present in genomic sequences.
In the federated-learning section, we delve into constructing models using federated learning systems. Federated learning is a machine learning approach that trains an algorithm across multiple decentralized devices or servers holding local data samples, without exchanging them.
The EEG directory delves into the intricate realm of electroencephalogram (EEG) signals and their relationship with machine learning. EEG is an esteemed diagnostic tool that captures the brain's electrical activity non-invasively via electrodes placed on the scalp. Its clinical significance spans the diagnosis of epilepsy, sleep disorders, the depth of anesthesia, and a myriad of other neural disorders.
In this project, we traverse the spectrum of deep learning methodologies tailored for EEG data. Our approach encapsulates feature extraction and robust classification regimes to discern a predisposition to alcoholism. The dataset, originating from a study on genetic predispositions to alcoholism, brings forth the challenges and intricacies of real-world EEG data.
A standout feature of our exploration is the harmonization of EEG data with avant-garde neural network architectures. Specifically, the integration of Convolutional Neural Networks (CNNs) showcases promising results, especially when dealing with raw EEG data or image-based inputs. a predisposition to alcoholism
- Name: Charles-André Arsenec
- Email: [email protected]