http://arxiv.org/abs/1504.04455
In the era of rapidly increasing amounts of time series data, classification of variable objects has become the main objective of time-domain astronomy. Classification of irregularly sampled time series is particularly difficult because the data can not be represented naturally as a plain vector which directly can be fed into a classifier. In the literature, various statistical features derived from time series serve as a representation. Typically, the usefulness of the derived features is judged in an empirical fashion according to their predictive power.
In this work, an alternative to the feature-based approach is investigated. In this new representation the time series is described by a density model. Similarity between each pair of time series is quantified by the distance between their respective models. The density model captures all the information available, also including measurement errors. Hence, we view this model as a generalisation to the static features which directly can be derived, e.g., as moments from the density.
In the numerical experiments, we use data from the OGLE and ASAS surveys and demonstrate that the proposed representation performs up to par with the best currently used feature-based approaches. While the density representation preserves all static information present in the observational data, the features are only a less complete description. The density representation is an upper boundary in terms of information made available to the classifier. Consequently, the predictive power of the proposed classification depends on the choice of similarity measure and classifier, only. We therefore expect that the proposed method yields performance close to an optimal classifier. Due to its principled nature, we advocate that this new approach of representing time series has potential in tasks beyond classification, e.g., unsupervised learning.
S. Kugler, N. Gianniotis and K. Polsterer
Mon, 20 Apr 15
14/42
Comments: Submitted to MNRAS
You must be logged in to post a comment.