ML-Algorithms

18 core machine-learning algorithms — explained, derived, and built from scratch. One notebook per topic: the intuition, the maths, and a fully worked example on a real dataset, side-by-side with the scikit-learn version. Run any of them in your browser with one click — zero setup.

Why this repo?

• 🧠 Theory and practice — every algorithm covers the intuition, the underlying mathematics (cost functions, optimisation, key formulas), and a real worked example.
• 🛠️ From scratch + scikit-learn — see the algorithm implemented from first principles, then how you'd actually use it in practice.
• 📊 Real datasets & proper evaluation — the right metrics and plots for each problem, not just .fit() / .score().
• ▶️ Zero setup — one-click Open in Colab on every notebook, or pip install -r requirements.txt to run locally.
• 🎯 Great for self-taught learners, students, and interview prep — a single, consistent path from linear regression to anomaly detection.

Curriculum

Work through the topics in order — each builds on the previous ones. Click Colab to run a notebook in the browser with zero setup.

Supervised Learning

#	Topic	Notebook	Run	Status
1	Linear Regression	01_linear_regression.ipynb		✅ Done
2	Ridge, Lasso & ElasticNet	02_ridge_lasso_elasticnet.ipynb		✅ Done
3	Logistic Regression	03_logistic_regression.ipynb		✅ Done
4	Support Vector Machines (SVM)	04_svm.ipynb		✅ Done
5	Naive Bayes	05_naive_bayes.ipynb		✅ Done
6	K-Nearest Neighbor (KNN)	06_knn.ipynb		✅ Done
7	Decision Tree	07_decision_tree.ipynb		✅ Done

Ensemble Methods

#	Topic	Notebook	Run	Status
8	Random Forest	08_random_forest.ipynb		✅ Done
9	AdaBoost	09_adaboost.ipynb		✅ Done
10	Gradient Boosting	10_gradient_boosting.ipynb		✅ Done
11	XGBoost	11_xgboost.ipynb		✅ Done

Unsupervised Learning

#	Topic	Notebook	Run	Status
12	Unsupervised ML (overview)	12_unsupervised_overview.ipynb		✅ Done
13	Principal Component Analysis (PCA)	13_pca.ipynb		✅ Done
14	K-Means Clustering	14_kmeans.ipynb		✅ Done
15	Hierarchical Clustering	15_hierarchical_clustering.ipynb		✅ Done
16	DBSCAN Clustering	16_dbscan.ipynb		✅ Done
17	Silhouette Analysis	17_silhouette_analysis.ipynb		✅ Done
18	Anomaly Detection	18_anomaly_detection.ipynb		✅ Done

Each notebook follows the same structure

1. Concept & intuition — what the algorithm does and why
2. The mathematics — cost functions, optimisation, key formulas
3. Assumptions / requirements
4. A real dataset — chosen to showcase the algorithm
5. EDA — understanding the data first
6. Training with scikit-learn
7. Evaluation with the right metrics and plots
8. From scratch (where it deepens understanding)
9. Pros, cons & when to use
10. Summary

Getting started

# Install dependencies
pip install -r requirements.txt

# Launch Jupyter and open a notebook
jupyter notebook

Datasets are a mix of scikit-learn / seaborn built-ins (zero setup) and real-world datasets where they make a more compelling example. Local data files live in data/.

Project structure

ML-Algorithms/
├── 01-linear-regression/
│   └── 01_linear_regression.ipynb
├── 02-ridge-lasso-elasticnet/
├── 03-logistic-regression/
├── ...                          # one folder per topic, 01 … 18
├── 18-anomaly-detection/
├── data/                        # local datasets (most come from sklearn/seaborn)
├── assets/                      # shared images + generate_gallery.py (README banner)
├── requirements.txt
├── CONTRIBUTING.md
├── LICENSE
└── README.md

Roadmap

Done: all 18 classical algorithms above. Planned / ideas (contributions welcome — see below):

• Time-series forecasting (ARIMA, Prophet)
• Neural networks from scratch (MLP, backprop)
• Model selection & cross-validation deep-dive
• Feature engineering & pipelines
• A shared "datasets" cheatsheet

⭐ Star the repo to follow along as new topics land.

Contributing

Found a typo, a clearer explanation, or want to add a topic from the roadmap? PRs and issues are very welcome — see CONTRIBUTING.md.

License

Released under the MIT License.