Clustering algorithm implementaion and visualization from scratch with python

Four popular clustering algorithms (for d >= 2 dimensions, k >= 2 clusters):

• (1) k-means clustering
• (2) Gaussian mixture model - expectation maximization algorithm (EM-GMM)
• (3) mean-shift clustering
• (4) agglomerative clustering

Python implementations:

• KMeans.py: k-means clustering
• KMeans_Ver0.py: second version of k-means implementation as a function
• GaussianMM.py: EM-GMM
• GaussianMM_Ver0.py: second version of EM-GMM implementation with functions of AIC, BIC and predict
• MeanShift.py: mean-shift clustering
• Agglomerative: agglomerative clustering

Evaluations and tests:

• test_2d_visualization.py: tests on 2D datasets with visualization, compared with sklearn implementation
• data_2d_test folder: datasets for tests
• test_2d_visualization_results folder: output images of tests

Visualization Results

The following figures compare the clustering results of my own implementations with those of scikit-learn's implementations. Each dataset is processed using different clustering algorithms.

---

Blobs Dataset

Algorithm	My Implementation	Scikit-learn
Agglomerative
EM-GMM
K-Means
Mean-Shift

---

Moons and Stars Dataset

Algorithm	My Implementation	Scikit-learn
Agglomerative
EM-GMM
K-Means
Mean-Shift

---

Sticks Dataset

Algorithm	My Implementation	Scikit-learn
Agglomerative
EM-GMM
K-Means
Mean-Shift

---