Create synthetic 2D dataset with gaussian centers to test clustering algorithms

synthetic2d_dataset.py

import numpy as np
import matplotlib.pyplot as plt

n=200 # samples per center
centers= [ [10,5], [-2,4], [13,-25], [11,20], [15,-30], [3,-2], ] # centers
#centers= [ [10,5], [-2,4], [11,20], [15,-34], ]
dataset=np.zeros((0,3))
sigma=2

for i in range(len(centers)):
    correlation=np.random.rand()
    center=centers[i]
    cluster=np.random.multivariate_normal(center, [[sigma, correlation],[correlation, sigma]], n)
    label=np.zeros((n,1))+i
    cluster=np.hstack([cluster,label])
    dataset=np.vstack([dataset,cluster])
print(dataset.shape)


plt.scatter(dataset[:,0],dataset[:,1])
filename="2d_complex.csv"
np.savetxt(filename,dataset,delimiter=",",fmt='%f')