This repo focuses on the topological machine learning research performed within the Topological ML Group, advancing graphical deep learning methods through persistent homology pipelines and vectorization techniques applied to complex datasets.
The project builds and experiments with advanced topological data analysis (TDA) methods, specifically persistent homology, to extract shape and connectivity features from graphs and datasets. These methods are integrated with graphical representation learning on temporal graph data, with practical applications involving node and edge filtrations such as sublevel and superlevel filtrations. Cutting-edge graph descriptors including degree, centrality, Forman Ricci curvature, and Heat Kernel Signatures are utilized to enrich feature extraction.
In this repository, my research contributions include:
- Developing and refining persistent homology pipelines on classified datasets
- Implementing Betti vectorization techniques for topological feature representation
- Applying temporal graph machine learning models, including XGBoost and MLP Classifiers, on filtration-based features
- Focusing on temporal graph branches and graphical representation learning methods
- Deploying sophisticated topological descriptors to enhance machine learning model input through structural and geometric insights
All research is conducted under the Topological-ML Lab framework: https://sites.google.com/view/topo-ml
This research is carried out under the guidance of Professor Baris Coskunuzer and the Topological-ML Lab, with a focus on advancing topological machine learning methodologies.
This is an active research project with ongoing development of pipelines, data integration, and novel vectorization approaches. The work is designed to support the lab’s experimental machine learning efforts on graph-structured data.
For more details about the research group and projects, please visit:
Topological ML Group Website