evclust.wmvec#
This module contains the main function for Adaptive Weighted Multi-View Evidential Clustering (WMVEC).
Zhe Liu, Haojian Huang, Sukumar Letchmunan, Muhammet Deveci, Adaptive weighted multi-view evidential clustering with feature preference, Knowledge-Based Systems, Volume 294, 2024, 111770, ISSN 0950-7051
Module Contents#
- evclust.wmvec.wmvec(X, c, alpha=2, delta=5, maxit=20, epsi=0.001, beta=1.1, lmbda=403.4288, type='simple', disp=True)[source]#
Weighted Multi-View Evidential Clustering (WMVEC) Algorithm. WMVEC can be viewed as a multi-view version of conventional evidential c-means clustering. Specifically, the view weight can measure the contribution of each view in clustering. WMVEC is based on objets row-data.
Parameters:#
- X (list of np.ndarray):
List of datasets from different views.
- c (int):
Number of clusters.
- alpha (float):
Parameter for distance weighting.
- beta (float):
Exponent for mass function calculation.
- lmbda (float):
Parameter for R update.
- delta (list):
List of penalties for empty sets for each view.
- epsi (float):
Convergence threshold.
- maxit (int):
Maximum number of iterations.
- type (str):
Type of focal set matrix to generate (‘simple’, ‘full’, ‘pairs’).
- disp (bool):
If True (default), intermediate results are displayed.
Returns:#
The credal partition (an object of class “credpart”).
Example:#
from evclust.wmvec import wmvec from evclust.datasets import load_prop df = load_prop() clus = wmvec(X=df, c=4, alpha=2, delta=5, maxit=20, epsi=1e-3, beta=1.1, lmbda=403.4288, type="simple", disp=True) # View weight clus['param]['R']
References:#
Zhe Liu, Haojian Huang, Sukumar Letchmunan, Muhammet Deveci, Adaptive weighted multi-view evidential clustering with feature preference, Knowledge-Based Systems, Volume 294, 2024, 111770, ISSN 0950-7051
See also
extractMass(),makeF(),Centroids_Initialization(),get_distance_wmvec(),update_Aj_wmvec(),update_M_wmvec(),update_R_wmvec(),update_V_wmvec(),update_jaccard_wmvec()Note
Keywords : Evidential clustering, Multi-view learning, Theory of belief functions, Credal partition WMVEC can be viewed as a multi-view version of conventional evidential c-means clustering. The objec-tive function of WMVEC integrating the learning of view weightsand credal partition into a unified framework, and design an optimiza-tion scheme to obtain the optimal results of WMVEC. Specifically, the view weight can measure the contribution of each view in clustering. Thecredal partition can provide a deeper understanding of the data structureby allowing samples to belong not only to singleton clusters, but also toa union of different singleton clusters, called meta-cluster.