| |
 |
| Paper: |
Detection of Rare Objects in Massive Astronomical Datasets Using Innovative Knowledge Discovery Technology |
| Volume: |
347, Astronomical Data Analysis Software and Systems XIV |
| Page: |
66 |
| Authors: |
Soto, A.; Cansado, A.; Zavala, F. |
| Abstract: |
Our work presents an application of knowledge discovery technology aimed to help scientists in the detection of rare types of astrophysical objects. Our main idea is that while computer have the power to search huge amounts of data, an expert has the domain knowledge to efficiently lead this search. Our system builds upon two main components: a probabilistic model able to scale to large datasets and a set of modules to interact with the scientist. Here, we focus on the probabilistic model used to represent the joint uncertainty among the attributes of the objects registered in a sky survey catalog. This model consists of a combination of a Bayesian network and a set of Gaussian mixture models (GMMs) trained with an accelerated version of the expectation maximization (EM) algorithm. The model is currently being tested using data from the release 1 of the Sloan Digital Sky Survey. The results indicate that the system is able to accurately detect a set of simulated rare objects, but it also provides a large number of false positives. |
|
|
|
|
|
