/DISC-landmark-selection

DISC landmark selection algorithm https://ieeexplore.ieee.org/stamp/stamp.jsp?arnumber=9449889

Primary LanguageJupyter Notebook

implementation of paper Kernel Matrix Approximation on Class-Imbalanced Data With an Application to Scientific Simulation authored by PARISA HAJIBABAEE, FARHAD POURKAMALI-ANARAKI and MOHAMMAD AMIN HARIRI-ARDEBILI

Toy Dataset

dataset = make_blobs(n_samples=[int(2000*0.02), int(2000*0.88), int(2000*0.1)], n_features=2, cluster_std=0.6,
                     centers=[[-10, 8], [-7, 5], [-5, 1]], random_state=2)

disc = DISCLandmarkSelection(num_landmarks=6, mixing_coef=0.5, compression_ratio=0.1, random_state=1)
disc.fit(dataset[0], dataset[1])
disc.plot()

toy dataset figure

Real World Datasets

Ozone

Reconstruction Error

Ozone reconstruction error plot

Recall Score

Ozone recall score plot

Mammography

Reconstruction Error

Mammography reconstruction error plot

Recall Score

Mammography recall score plot

Wine

Reconstruction Error

Wine reconstruction error plot

Recall Score

Wine recall score plot