|
Title:
|
FITTING MULTIVARIATE GAUSSIAN DISTRIBUTIONS WITH OPTIMUM-PATH FOREST AND ITS APPLICATION FOR ANOMALY DETECTION |
|
Author(s):
|
Leandro Aparecido Passos Júnior, Kelton Augusto Pontara da Costa, João Paulo Papa |
|
ISBN:
|
978-989-8533-45-6 |
|
Editors:
|
Hans Weghorn |
|
Year:
|
2015 |
|
Edition:
|
Single |
|
Keywords:
|
Optimum-Path Forest, Multivariate Gaussian Distribution, Anomaly Detection. |
|
Type:
|
Full Paper |
|
First Page:
|
136 |
|
Last Page:
|
142 |
|
Language:
|
English |
|
Cover:
|
|
|
Full Contents:
|
click to dowload 
|
|
Paper Abstract:
|
In this paper, we deal with the problem of estimating Gaussian parameters for anomaly detection by means of unsupervised Optimum-Path Forest (OPF), which can find out the number of clusters on-the-fly. Since the most widely techniques for such task require the number of Gaussians as a prior knowledge, they might not work well in real situations, in which the number of clusters may be different than the number of classes. Although such information may be estimated by cross-validation, it might be prohibitive for large datasets. We have shown situations in which OPF can outperform some well-known techniques, such as Mean Shift, k-means and Expectation-Maximization. |
|
|
|
|
|
|
