Mathematics – Logic
Scientific paper
Jan 2010
adsabs.harvard.edu/cgi-bin/nph-data_query?bibcode=2010aas...21523005p&link_type=abstract
American Astronomical Society, AAS Meeting #215, #230.05; Bulletin of the American Astronomical Society, Vol. 42, p.579
Mathematics
Logic
Scientific paper
I present an automatic classification system built at the Time Series Center at Harvard. We use machine-learning techniques (SVM and k-NN) to first learn and then classify unclassified astronomical objects. We demonstrated the accuracy using known periodic variables (99% on OGLE variables). We applied the method to legacy catalogs such as MACHO, and discovered twice as many new variables in the MACHO catalog than previously found using traditional non-automatic techniques.
Using the newly classified objects we created an on-line morphological classification search engine. Through a web interface, a user can upload or draw a time series and find similar examples. From the matched similar objects a classification is attributed. The back engine is implemented with CUDA on a GPU using a simple indexing structure. We are able to search through millions of time series in less than a second. In case that in the results there are no known classified objects, we furthermore mine the publication abstracts of the closest matched sources. Using weighted keywords we provide yet another classification attribute.
No associations
LandOfFree
Classification of Astronomical Sources Through Conversation With Data does not yet have a rating. At this time, there are no reviews or comments for this scientific paper.
If you have personal experience with Classification of Astronomical Sources Through Conversation With Data, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Classification of Astronomical Sources Through Conversation With Data will most certainly appreciate the feedback.
Profile ID: LFWR-SCP-O-961018