Short description: Jupyter Notebook tutorial for building, training, and visualizing Self‑Organizing Maps (SOMs).
- About
- Features
- Repository contents
- Requirements
- Installation
- Usage
- Notebook walkthrough
- Algorithm overview
- Hyperparameters & tips
- Visualizations included
- Evaluation metrics
- Extensions & next steps
- References and citation
- Contributing
- License
- Suggested GitHub topics
This project demonstrates how to implement and use Self‑Organizing Maps (SOMs) for exploratory data analysis. The included Jupyter Notebook(s) walk through data preparation, a from‑scratch SOM implementation (NumPy), an example using a compact SOM library, training, visualization (U‑matrix, hits, component planes), and interpretability methods for clusters and prototypes.
SOMs are useful for:
- Visualizing high‑dimensional data on a 2D grid
- Discovering cluster structure without labels
- Producing a topology preserving mapping (neighbors in input space map to neighbors on the grid)
som, self-organizing-map, kohonen, unsupervised-learning, clustering, dimensionality-reduction, data-visualization, machine-learning, python, jupyter-notebook, minisom, neural-networks, tutorial, reproducible-research
(If you'd like, I can also add these topics to the repository About box and the repo metadata directly — however, I can only modify files in the repo via this interface. To update the repository About box and topics in the repository settings, either grant me permission to update repo settings or paste the topics into the GitHub settings manually.)